Drug Discovery: Advances and Innovations

SPONSORED BY DRUG DISCOVERY: Advances and Innovations Cancer Drug Discovery: Reaching for the High‑Hanging Fruit Five Key Advances Shaping Pharmacogenomics How Machine Learning Is Reshaping Drug Discovery Credit: iStock, Alkestida CONTENTS 5 Rethinking Synthetic Accessibility in Modern Drug Discovery 9 Discovering Drugs for Chronic Pain: Rising to the Challenge 13 How Machine Learning Is Reshaping Drug Discovery 16 Five Key Advances Shaping Pharmacogenomics 20 Offering Fresh Hope: Game‑Changing Drug Discoveries for Rare Diseases 25 Rethinking Target-Based Drug Discovery: Challenges, Innovations and the Next S-Curve 31 Cancer Drug Discovery: Reaching for the High‑Hanging Fruit 34 mRNA Nanoparticles Offer New Hope for Female Infertility Treatment DRUG DISCOVERY: ADVANCES AND INNOVATIONS 3 TECHNOLOGYNETWORKS.COM FOREWORD The core aim of drug discovery is to translate biological insight into medicines that improve and save lives. In recent years, the landscape of drug discovery has evolved rapidly. Emerging tools such as AI and machine learning are helping researchers analyze vast biological datasets and accelerate key stages of the discovery pipeline, enabling faster and more informed decision-making. At the same time, breakthroughs in areas such as pharmacogenomics, drug delivery, and next-generation therapeutic modalities are empowering scientists to explore novel strategies to address chronic pain, cancer, rare diseases, and reproductive health conditions. This eBook brings together a collection of insights that highlight the scientific advances, technological innovations, and collaborative efforts shaping modern drug discovery. Through expert perspectives and real-world examples, it explores how researchers are overcoming long-standing challenges and opening new pathways toward therapeutic innovation. The Technology Networks editorial team Application Programming Interface Connect SYNTHIA® Retrosynthesis Software API with your preferred cheminformatics software for high-throughput pathway design & molecular synthesis assessment. Discover More www.synthiaonline.com/api Expedite in silico drug discovery The Life Science business of Merck operates as MilliporeSigma in the U.S. and Canada. MK_AD13159EN Ver. 2.0 04/2026 © 2026 Merck KGaA, Darmstadt, Germany and/or its affiliates. All Rights Reserved. Merck, the vibrant M, Sigma-Aldrich and SYNTHIA are trademarks of Merck KGaA, Darmstadt, Germany or its affiliates. All other trademarks are the property of their respective owners. Detailed information on trademarks is available via publicly accessible resources. 5 DRUG DISCOVERY: ADVANCES AND INNOVATIONS Rethinking Synthetic Accessibility in Modern Drug Discovery In today’s drug discovery landscape, the bottleneck is no longer generating ideas—it’s deciding which ideas are worth pursuing. With the rise of combinatorial chemistry and generative AI models, researchers can now design millions of candidate molecules in silico. But there’s a catch: many of these compounds are, in practical terms, impossible or prohibitively complex to synthesize. So how do you separate the promising from the impractical, before investing valuable time and resources? This is where synthetic accessibility becomes a decisive factor. The hidden cost of “virtual” molecules Virtual screening pipelines are incredibly powerful, but they often overlook a critical dimension: can the molecule actually be made? A compound might show excellent predicted binding affinity, ideal ADMET properties, and novel structural features, yet still require over 20 synthetic steps, exotic reagents, or unfeasible reaction pathways. Without early-stage filtering, teams risk prioritizing molecules that stall in the lab. The solution lies in integrating synthetic feasibility directly into the design loop. A new lens: Synthetic accessibility scoring Synthetic accessibility scoring provides a fast, quantitative way to estimate how difficult it is to synthesize a molecule. Instead of relying on intuition or manual retrosynthetic analysis, scientists can now assign a score to each candidate, instantly. The SYNTHIA® Synthetic Accessibility Score (SAS) reframes this challenge by answering a simple but powerful question: How many steps would it take to make this molecule from commercially available building blocks? The result is a single score between 0 and 10: ∙ 0–2  Easy to synthesize (or already commercially available) ∙ 3–6  Moderately complex ∙ 7–10  Difficult or potentially infeasible This transforms synthetic feasibility into something actionable—enabling rapid “make/no-make” decisions at scale. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 6 From SMILES to insight in seconds The workflow is intentionally simple. Chemists input molecules using SMILES notation, the standard text-based representation of chemical structures. These are submitted—individually or in batches—to a cloud-based API. In return, each molecule is assigned a SAS. What makes this powerful is scale. Millions of molecules can be processed daily, and thousands in a single query. This means synthetic feasibility can be evaluated: ∙ During virtual screening ∙ Alongside docking or QSAR predictions ∙ Before committing to synthesis Instead of being a late-stage constraint, synthesis becomes an early design parameter. Under the hood: Learning chemistry from chemistry Behind the scenes, the SAS model leverages advances in deep learning—specifically, graph convolutional neural networks (GCNNs). Unlike traditional descriptor-based models, GCNNs operate directly on molecular graphs: ∙ Atoms become nodes ∙ Bonds become edges ∙ Structural relationships are learned dynamically The architecture combines: ∙ A directed message passing neural network (D-MPNN) for molecular representation ∙ A feedforward neural network (FNN) for prediction Crucially, the model isn’t trained on abstract labels—it learns from real retrosynthetic routes generated by SYNTHIA®’s planning engine. This grounds predictions in practical chemistry rather than theoretical approximations. Making sense of the score One of the challenges in synthetic accessibility is handling extreme complexity. Molecules requiring many steps can quickly skew predictions. To address this, the scoring system uses a smoothing function: ∙ For simpler molecules, the score behaves almost linearly with the number of steps ∙ For highly complex molecules, scores are compressed toward the upper limit (10) 12 10 8 6 4 2 0 0 5 10 STEPS 15 20 25 FIGURE 1: DEPICTION OF SMOOTHING FUNCTION APPLIED TO SCORES. NOTE, THAT OVER SMALL AND MODERATE VALUES (X-AXIS), THE SYNTHETIC ACCESSIBILITY SCORE (Y-AXIS) BEHAVES CLOSE TO LINEAR. IN OTHER WORDS, THE RETURNED SCORE CORRESPONDS TO THE NUMBER OF SYNTHETIC STEPS PREDICTED BY THE MODEL. FOR A HIGHER NUMBER OF PREDICTED SYNTHESIS STEPS (AROUND 10 OR ABOVE), THE RELATED SCORE IS SMOOTHENED SUCH THAT THE RETURNED VALUE IS STILL CLOSE TO (AND NOT GREATER THAN) 10. THIS ALLOWS TO RE-SCALING ALL CONSIDERED CASES TO [0, 10] INTERVAL. This keeps the scale intuitive while preserving resolution where it matters most—especially when distinguishing between “hard” and “very hard” molecules. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 7 When chemistry defies intuition: Real-world examples Even experienced chemists can be surprised by synthetic accessibility when subtle structural factors come into play. Case study 1: Simpler than it looks A derivative of sulfamethoxazole (Figure 2)—despite appearing more complex—receives a much lower SAS score (~1.0) than the parent drug (~4.0). Why? Because it is closer to readily available building blocks and requires fewer synthetic transformations. Complexity in structure doesn’t always translate to complexity in synthesis. Case study 2: The cost of protection In contrast, the N-Boc derivative of adrenaline (Figure 3) scores higher (~8.4) than adrenaline itself (~7.6). Here, the added protecting group introduces unnecessary synthetic steps. Although common in lab workflows, such modifications can increase overall complexity when viewed from a retrosynthetic perspective. These examples highlight a key insight: Synthetic accessibility is not just about structure—it’s about pathways. FIGURE 2: CHEMICAL STRUCTURES OF MOLECULES FOR SULFAMETHOXAZOLE USE CASE. FIGURE 3: CHEMICAL STRUCTURES FOR ADRENALINE USE CASE. O H3C H3C CH3 CH3 H3C O N OH OH HO NH OH OH HO SAS = 8.399 SAS = 7.631 O O O O N S SAS = 1.038 SAS = 4.051 HN O O S NH H3 C H3 C O N HN NH2 H3 C TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 8 Scaling decisions across the pipeline Integrating synthetic accessibility into discovery workflows unlocks several advantages: ∙ Smarter prioritization: Focus on compounds that are both biologically promising and synthetically feasible. ∙ Faster iteration cycles: Eliminate impractical candidates early. ∙ Improved collaboration: Align computational chemists and synthetic chemists with shared metrics. ∙ Data-driven decision-making: Replace subjective judgments with reproducible scoring. Because the system is delivered via a scalable API, it can be embedded directly into existing cheminformatics platforms, enabling seamless, automated decision-making. A note on limitations Like all machine learning models, synthetic accessibility predictions depend on the data they were trained on. For molecules that fall outside the model’s applicability domain—for example, highly novel or unusual structures—predictions may be less reliable. This doesn’t diminish the value of SAS, but it does reinforce an important principle: Use synthetic accessibility scores as a guide, not a substitute, for expert judgment. Toward “design for synthesis” Drug discovery is evolving from a linear process into an integrated, feedback-driven system. In this new paradigm, molecules are no longer just designed for potency or selectivity—they are designed for synthesis from the very beginning. By embedding synthetic accessibility into screening workflows, tools like SYNTHIA® SAS enable a shift from: ∙ Can we make this?  asked too late to ∙ Should we make this?  answered immediately And in a field where time, cost, and success rates are tightly linked, that shift can make all the difference. 9 DRUG DISCOVERY: ADVANCES AND INNOVATIONS Discovering Drugs for Chronic Pain: Rising to the Challenge Kerry Day, PhD Drug discovery is a lengthy and expensive process; however, computational modeling and AI now offer unprecedented opportunities for faster discovery. A field still in urgent need of new treatments, however, is chronic pain. Chronic pain, defined as pain persisting for more than three months, is estimated to affect almost one-third of the global population. It leads to reduced quality of life, disability, and psychological comorbidities such as depression and anxiety, while having major economic, healthcare system, and societal impacts.1 Current treatments aim to block the pain signal at all levels of transmission from the periphery to the primary somatosensory cortex area of the brain, where pain is perceived.1 Many chronic pain medications have an unspecific mechanism of action, depressing the central nervous system to cause side effects such as reduced mobility and impaired memory.2 While opioids are effective for chronic pain relief, there is a human cost to their use. Opioid associated mortality has quadrupled since the turn of the century, with adverse side effects including tolerance, addiction, and increased pain sensitivity.1 With limited efficacy and a common theme of adverse side effects, there remains a significant unmet clinical need for effective treatments for chronic pain. Dr. Christopher L. Robinson, a regenerative and pain physician-scientist at The Johns Hopkins University School of Medicine, explained why drug discovery and development for chronic pain is so challenging: “Pain is not just a single target or defect. There are numerous Credit: iStock/BitsAndSplits TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 10 pain signaling pathways, receptors, and underlying causes of pain. Some of the aforementioned are involved in other critical processes, so inhibiting one of them can lead to off-target effects, but does not mean we are not trying.” Also addressing the challenge of chronic pain treatment, David Bennett, a professor of neurology and neurobiology, and Steven Middleton, a postdoctoral researcher, both at the University of Oxford, added: “Chronic pain is a very heterogeneous condition likely reflecting multiple mechanisms at play, and not all drugs will work for all types of chronic pain. That's why detailed patient phenotyping when recruiting for clinical trials is really important.” “Pain-relief can engage the reward system, so it is then an added challenge to develop new and effective analgesics that have no side effects, such as addiction,” they continued. Despite these significant challenges, scientists worldwide are exploring new mechanisms and targets for pain relief. Non-opioid alternatives Many emerging treatments focus on modulating opioid receptors to achieve the same pain relief as opioids without the adverse side effects. One such area of research explores the activation of different opioid receptors, specifically the delta-opioid receptor (DOR), as opposed to the main opioid receptor, mu. DOR is widely distributed across pain-processing areas, but is also present in the amygdala, which is responsible for emotional processing, so it may modulate the emotional association of pain.1 Molecules such as LIM Kinase, RSG4, and GPCRs all aim to increase expression of DOR and are either in preclinical trials or under investigation.1 The mechanism of action of RSG4 is to prolong the natural analgesic effects of endogenous opioids, removing the need for exogenous sources and their unwanted side effects. Another molecule that enhances naturally occurring endogenous opioids is granulocyte-colony stimulating factor (G-CSF), which increases local neutrophils, leading to a greater secretion of endogenous opioids for pain relief. Sodium channel inhibitors Robinson, Bennett, and Middleton, all experts in this field, believe that selective sodium channel inhibitors, such as the recently FDA-approved drug Suzetrigine, are the most significant development in pain drug discovery in recent times. In the central nervous system (CNS), sodium channels are essential for neuronal excitability and signal transmission. Certain subtypes of sodium channels (e.g., Nav1.7, Nav1.8, Nav1.9) have specific roles in pain signal transmission, offering potential drug target options.3 Suzetrigine, a Nav1.8 channel blocker, is the first non-opioid drug for moderate to severe acute pain to be approved in decades, and studies are ongoing to assess its use in chronic pain.3 Gene therapy An alternative approach in emerging pain treatments is gene therapy. In preclinical trials, DNA vectors have shown success inhibiting leukocyte elastase to reduce neuropathic pain in mice.1 Two treatments in development have also exhibited beneficial effects for other diseases, offering the possibility of dual therapy. Although these early-stage therapies require further preclinical trials and human studies to assess safety and efficacy, they offer new avenues for chronic pain treatment development. Pain-related genes The recent discovery from Bennett’s laboratory of a pain-related gene (SLC45A4) unveils a new therapeutic target and pathway in chronic pain. Their preclinical studies indicate a role for polyamines in chronic pain, while clinical studies show increased levels of polyamines in rheumatoid arthritis and inflammation.4 Optimistic about the future, Bennett and Middleton remarked, “If our future research continues to support SLC45A4 and polyamine regulation as a key drug target, TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 11 then it is possible patients suffering from chronic pain may one day see these new drugs in clinics, but we have many hurdles to tackle first.” Cannabinoid receptors For the first time, scientists have designed a compound that binds to cannabinoid receptor type 1 (CB1), eliciting analgesia without the unwanted psychoactive effects.5 Using 3D computational modeling to identify a novel cryptic pocket (a binding area in the receptor that appears transiently), the researchers from Stanford University and Washington University then rationally designed a compound to bind selectively and not trigger unwanted secondary effects in the CNS. “The drug we designed here holds a great promise; it reduces dependency on opioids for chronic pain and offers a non-addictive alternative,” remarked Vipin Rangari, a scientist at Washington University School of Medicine and lead author of the paper. Targeting novel pain pathways Researchers at the University of Aberdeen recently discovered a new pain pathway for targeting chronic pain. This pathway centers around the concept of sng pain (the Taiwanese term for soreness related to muscle acid buildup) and is related to excessive amounts of glutamate release in muscles, leading to permanent activation of pain nerves.6 Their findings put a spotlight on ASIC3 as a new drug target for tissue acidosisassociated chronic pain, which is common in conditions such as rheumatoid arthritis, delayed onset muscle soreness and fibromyalgia. Psychedelics Following successful trials of psychedelics for the treatment of depression and anxiety, they are now under investigation for use in chronic pain.7 This is an emerging area of interest, and more research is needed to determine if this class of drugs could be beneficial in chronic pain. Many chronic pain treatments target symptoms. How can the root cause of such a complex problem be treated? Robinson believes the answer may lie in regenerative medicine and stem cell-based treatments: “We are now able to grow tissues and organs in the lab, and in the near future, we will one day regrow your own cartilage from your own cells by utilizing induced pluripotent stem cells (IPSCs), not to be confused with embryonic stem cells. So, it is you curing you via the beauty of IPSCs-based science.” Harnessing innovative technologies in chronic pain management and drug discovery With new treatment options for chronic pain scarce and still under development, scientists have turned to innovative technologies such as virtual reality (VR), wearable medical technologies and AI. Virtual reality distracts the patient from the perception of pain and facilitates neural reprocessing or retraining the brain to interpret pain differently. VR studies in chronic lower back pain demonstrate pain reduction, whereas wearable medical technology research reveals reduced depression and opioid use.8 Such wearable technology also gives clinicians insight into daily physiological data readouts in chronic pain, aiding their understanding of the condition. The unique ability of AI to interrogate large volumes of data is accelerating drug discovery in terms of target identification, virtual screening, biomarker discovery, drug repurposing identification, and prediction of pharmacokinetic properties and toxicity.9 “[…] small molecule drug discovery is a feat in itself, but AI-driven drug modelling is expediting the processes by narrowing down the hits,” explained Robinson. However, AI faces challenges of data scarcity, limited training options, a lack of standardization, and intense resource requirements.9 Despite these, AI’s rapid advancement offers hope of faster drug discovery and a net benefit for health research in the future. Using computational modeling to identify cryptic pockets offers a novel method for discovering therapeutic targets that would have otherwise remained hidden. Computer-aided drug discovery applications can TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 12 significantly reduce the number of candidate molecules to evaluate, therefore expediting the process.10 “[…] exciting advancements in drug discovery include rapid advancements in structural biology, specifically cryoEM. Target structures from cryoEM can be used to rationally tune and discover molecules that are effective in their therapeutic purpose while avoiding off-target or side effects,” explained Evan O’Brien, an assistant professor at The Johns Hopkins University School of Medicine. “Further, next-gen screening approaches are rapidly replacing ‘conventional’ library screenings. Ultra-large ‘barcoded’ libraries of small molecules allow for fast discovery of new lead compounds, and computational docking approaches (again leveraging the above experimental advances in cryoEM) allow for computers to do the high-throughput discovery for us.” Machine learning can also be coupled with proteomic profiling, enabling the identification of protein biomarkers in chronic pain.11 After decades of little progress in chronic pain, research aided by technological innovations enabling a new era of drug discovery, revealing new mechanisms, therapeutic targets and compounds. It is hoped these breakthroughs will soon lead to treatments for millions of chronic pain sufferers worldwide. REFERENCES 1. Patel NP, Bates CM, Patel A. Developmental approaches to chronic pain: A narrative review. Cureus. 2023;15(9):e45238. doi: 10.7759/ cureus.45238 2. Zakka T, Papler H, Pêgo-Fernandes PM. Chronic pain: A big challenge. Sao Paulo Med J. 2024;142(1):e20231421. doi: 10.1590/1516-3180.2024.1421.131223 3. Chen R, Liu Y, Qian L, et al. Sodium channels as a new target for pain treatment. Front Pharmacol. 2025;16:1573254. doi: 10.3389/ fphar.2025.1573254 4. Middleton SJ, Markusson S, Akerlund M, et al. SLC45A4 is a pain gene encoding a neuronal polyamine transporter. Nature. 2024. doi: 10.1038/s41586-025-09326-y 5. Rangari VA, O’Brien ES, Powers AS, et al. A cryptic pocket in CB1 drives peripheral and functional selectivity. Nature. 2025;640,265– 273. doi: 10.1038/s41586-025-08618-7 6. Lee CH, Lin JH, Lin SH, et al. A role for proprioceptors in sngception. Sci Adv. 2025;11:eabc5219. doi: 10.1126/sciadv.abc5219 7. Robinson CL, Fonseca ACG, Diejomaoh EM, et al. Scoping review: The role of psychedelics in the management of chronic pain. J Pain Res. 2024;17,965–973. doi: 10.2147/JPR.S439348 8. Slitzky M, Yong RJ, Bianco GL, Emerick T, Schatman ME, Robinson CL. The future of pain medicine: Emerging technologies, treatments, and education. J Pain Res. 2024;17 2833–2836. doi: 10.2147/JPR.S490581 9. Zhang K, Yang X, Wang Y, et al. Artificial intelligence in drug development. Nat Med. 2025;31(1):45-59. doi:10.1038/s41591-024- 03434-4 10. Shah M, Patel M, Shah M, Patel M, Prajapati M. Computational transformation in drug discovery: A comprehensive study on molecular docking and quantitative structure activity relationship (QSAR). Intelligent Pharmacy. 2024;2(5),589-595. doi: 10.1016/j. ipha.2024.03.001 11. Chen L, Kelleher E, Meng R, et al. Diagnosis, prognosis, and drug discovery for chronic widespread pain: A large proteogenomic study. Adv Sci. 2025;e07691. doi: 10.1002/advs.202507691 MEET THE INTERVIEWEES: Dr. Christopher L. Robinson is a physician-scientist and assistant professor exploring regenerative medicine and stem cell-based treatments for chronic pain at the Johns Hopkins University School of Medicine. Dr. David Bennett is a professor of neurology and neurobiology researching pain and the response of the nervous system to injury at the University of Oxford. Dr. Steven Middleton is a postdoctoral researcher investigating neural injury at the University of Oxford. Dr. Vipin Rangari is a scientist at Washington University School of Medicine working in pain research whose interests include opioid agonists and cannabinoids. Dr. Evan O’Brien is an assistant professor at The Johns Hopkins University School of Medicine specialising in G-protein coupled receptor (GPCR) structure & dynamics for drug discovery. 13 DRUG DISCOVERY: ADVANCES AND INNOVATIONS How Machine Learning Is Reshaping Drug Discovery Katie Brighton Drug discovery pipelines are notorious for being costly, slow, and failure-prone, leading to AI and machine learning becoming more commonplace to accelerate progress and improve outcomes. Currently, machine learning in drug discovery centers around data-rich stages, which provide plentiful data for algorithm training. However, parts of the pipeline that generate less data could also benefit from machine learning. Technology Networks spoke to Dr. Daniel Reker, an assistant professor of biomedical engineering at Duke University, about his work on pairwise molecular learning, which enables better computational decisionmaking in data-scarce scenarios. In this interview, Reker discusses how pairwise molecular learning opens up new avenues in drug discovery, including for first-in-class drug candidates, and explores what happens when machine learning is integrated into automated laboratories. Q: How would you describe the role machine learning plays in modern drug discovery today—and where does it still fall short? A: Machine learning is actively reshaping drug discovery across multiple stages of the pipeline, and we see widespread adoption from pharma and biotech as well as interest from tech companies and numerous startups. The majority of these efforts currently focus on target identification, lead generation, and clinical trials. While it's still too early for definitive assessments, early readouts suggest computational approaches have accelerated timelines and modestly improved success rates, which could be significant given how costly, slow, and failure-prone drug discovery is. Credit: iStock/Just_Super TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 14 However, the current impact of machine learning concentrates heavily on data-rich stages that leverage high-throughput screening, genomics, and large-scale clinical datasets to enable training and fine-tuning of complex algorithms. Substantial progress can still be made in addressing data-scarce drug discovery challenges like lead optimization, safety, and formulation development. These stages rely on low-throughput experiments, such as complex synthesis, material characterization, and in vivo animal studies, but they represent critical decision points that determine the fate of drug candidates. Innovations in novel experimental platforms and robust computational algorithms are poised to enhance these decisions with potentially even stronger benefits to reduce cost and failure rates compared to what we have seen so far, ultimately positioning the community to bring more and better therapies to patients. Q: Could you explain a little more about what pairwise molecular learning is? A: Pairwise molecular learning transforms the traditional machine learning task into a contrastive problem where the algorithm directly compares two molecules rather than evaluating each one independently. Essentially, instead of asking the computer, “What is the potency of molecule A?” we transform the question to “Which of these two molecules is more potent?” This enables combinatorial data augmentation, creating millions of molecular comparisons from just hundreds to thousands of original datapoints. In simple terms, we give deep neural networks different perspectives on the same underlying data to enhance training efficiency. This allows us to train cutting-edge deep learning architectures on datasets of as few as 100–1000 compounds, which is where a lot of the real-world pharmaceutical decision-making around critical properties like drug safety, metabolism, and pharmacokinetics happens—these are expensive to measure experimentally but essential for advancing the best candidates. We believe pairwise learning will enable the community to unleash the predictive power of deep neural networks for these data-scarce but high-value decision points. Q: What kind of avenues in drug discovery does pairwise molecular learning open up? A: Pairwise molecular learning opens several exciting avenues in drug discovery. First, it enables more accurate computational molecular optimization by directly predicting which chemical modifications will improve critical drug properties like safety, metabolism, and potency. This helps medicinal chemists prioritize which compounds to synthesize next, saving time and resources. Second, this pairwise augmentation approach enables better computational decision-making in data-scarce scenarios. This is particularly valuable for properties like drug safety, metabolism, and formulations—critical decision points where experimental data is limited and expensive to generate. It can also enhance predictive performance on novel and challenging drug targets where little knowledge has been accumulated so far, thereby providing an opportunity for machine learning to better support the identification of first-in-class therapies. This capability is further strengthened algorithmically by pairwise learning's ability to incorporate bounded or incompletely characterized datapoints that are normally discarded from modeling efforts. While insufficiently characterized for direct inclusion in traditional models, these datapoints still provide important perspectives and contrast to stronger candidates. Third, our data suggests the algorithm excels at identifying genuinely novel molecules. By learning the impact of molecular changes rather than simply identifying analogues of known compounds, it avoids the memorization problem common in complex algorithms and pushes the algorithm to focus learning on relationships and patterns. In our proof-of-concept data, this enables more drastic structural modifications during optimization, with strong potential to further enhance safety and efficacy of drug candidates. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 15 Q: What are the biggest gains you’ve seen from combining machine learning with automated labs, and where are the remaining bottlenecks? A: The biggest gains from combining machine learning with automated labs that I have seen stem from creating truly adaptive experimental design loops. In the machine learning community, we call these “active learning workflows” to indicate that the predictive algorithm is directly involved in the data acquisition and can request the most informative and valuable datapoints. Our work and others have shown that such “active learning” setups can potentially reduce the required data for decision-making by up to 90% and enable better predictive models by directly addressing biases in the data. These setups have helped us to identify new drug candidates using fewer datapoints as well as identifying new nanoparticle formulations that enhance the efficacy and safety of medications with greater accuracy. A major remaining bottleneck in this deployment of such feedback loops centers around automation infrastructure and algorithmic robustness. Most highthroughput screening platforms are optimized for scale at the cost of flexibility, for example, relying on rapidly screening pre-defined compound libraries rather than enabling adaptive cherry-picking of individual experiments suggested by algorithms. Additionally, several of the critical experiments such as material characterization or even in vivo studies are difficult to integrate into these automated workflows. We believe these feedback cycles are most impactful in truly low-data scenarios—like early-stage projects with under 100 datapoints. But building predictive models and enabling them to decide which datapoint to acquire next remains challenging even for the most data-efficient computational approaches. We're addressing this through pairwise learning methods as well as other new active learning developments including yoked learning, where algorithms are paired to work together. There's substantial room for further innovation in automation architecture and experimental design strategies to maximize the impact of integrated laboratories on drug discovery. MEET THE INTERVIEWEE: Dr. Daniel Reker is a tenure-track assistant professor in the Department of Biomedical Engineering at Duke University and a Duke Science and Technology Fellow. He received his PhD from the Institute of Pharmaceutical Sciences at ETH Zurich and conducted his postdoctoral work in drug delivery as a SNSF Postdoctoral Fellow with Bob Langer and Giovanni Traverso at MIT. At Duke, he is building an integrated experimental and computational laboratory to conduct active machine learning campaigns with applications in drug discovery and drug delivery. 16 DRUG DISCOVERY: ADVANCES AND INNOVATIONS Five Key Advances Shaping Pharmacogenomics Bree Foster, PhD Pharmacogenomics is the study of how genetic variation influences individual responses to medications. This field has gained significant momentum since the completion of the Human Genome Project and the advent of high-throughput sequencing. To date, researchers have catalogued over 69,000 distinct single nucleotide variants (SNVs) and 200 structural variants (SVs) across more than 200 pharmacogenes.1 These advancements have enabled the discovery of numerous drug–gene associations, which now inform pharmacogenetic labeling and clinical guidelines for more than 100 medications.2,3 As pharmacogenomics matures, new technologies are rapidly reshaping how we understand the genetic basis of drug response. From sequencing innovations that decode hard-to-map genes to machine learning models that predict variant effects at scale, these tools are helping researchers pinpoint how genetic variation drives individual responses to therapy. Crucially, these insights are now feeding back into drug discovery pipelines and informing more tailored treatment strategies. This listicle explores five advances – spanning sequencing, AI, real-world data integration, and multiomics – that are driving the next generation of pharmacogenomics research and accelerating its clinical impact. 1. Long-read sequencing Long-read sequencing (LRS) is a key technology for pharmacogenomics as it is now revealing previously hidden structural variants, complex haplotypes, and pharmacogenetically relevant alleles that were undetectable with short-read technologies. Unlike traditional sequencing, LRS provides contiguous reads Credit: iStock/Katy L. Pack TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 17 that span tens of kilobases, allowing researchers to characterize complex genetic regions in a single assay. This enhanced resolution allows researchers to better characterize genes involved in drug metabolism and improve predictions of individual drug response. For example, the highly polymorphic enzyme cytochrome P450 2D6 (CYP2D6) processes between 20-30% of commonly prescribed drugs but is extremely challenging to genotype accurately due to its complex structural variations, including gene duplications, deletions, and hybrid alleles. LRS enables comprehensive mapping of the CYP2D6 locus, capturing these variations in full, thereby improving phenotype prediction and guiding safer, more effective drug dosing.4,5 Several platforms now provide high accuracy reads (up to 99.9%) with improved error correction, making them powerful tools for genome assembly, structural variant discovery, and clinical pharmacogenomics applications.6,7 As costs continue to fall and throughput rises, LRS is likely to become a go-to approach for uncovering hidden pharmacovariants and improving personalized medicine. 2. Integrating electronic health records and biobanks Electronic health records (EHRs) have become widespread over the past decade as, besides supporting and improving diagnosis, clinical decisions, and treatment coordination, they provide new data analytics opportunities.4 EHRs commonly encompass patient demographics, medical history, drug prescriptions, and, in some cases, laboratory results, radiological images, and wearable device data.8 Modern biobanks, such as the UK Biobank (UKB), All of Us Research Program, FinnGen and the Million Veteran Program, offer access to rich datasets that integrate whole-genome or whole-exome sequencing, longitudinal EHRs, medication records, and detailed phenotypic and survey data. These resources allow researchers to move beyond traditional trial-based pharmacogenomics and explore drug response across diverse populations and care settings. The All of Us Research Program has already demonstrated clinical value with pharmacogenomic insights such as DPYD genotyping for fluoropyrimidine toxicity. This has helped to refine dosing guidelines and improve patient safety by distinguishing between variants that truly impact chemotherapy response and those that do not.9 The UKB has also enabled discovery and replication of pharmacogenomic signals at scale, particularly for adverse drug reactions and prescribing behaviors. For example, associations between drug maintenance dose and 9 PGx genes were tested in 200,000 UKB participants by assigning individuals to a metabolizer class (e.g., poor, intermediate, and normal) based on their genotype. The study revealed known CYP2C9 and a novel CYP2C19 variant as determinants for warfarin dosage.10 EHR-linked biobanks accelerate the translation of research findings into actionable pharmacogenetic guidelines, guiding more precise prescribing decisions across populations. As integration with genomics becomes more routine, these real-world data ecosystems are helping bridge the gap between genetic discovery and personalized treatment planning. 3. Polygenic risk scores While traditional pharmacogenomics often focuses on single gene–drug interactions, many treatment outcomes, such as efficacy, side effects, and toxicity, are shaped by a complex interplay of multiple genetic variants. Polygenic risk scores (PGS) offer a way to capture this complexity by aggregating the effects of thousands of variants into a single predictive metric. PGS have already been used to predict response to sulfonylureas in type 2 diabetes, lurasidone efficacy in schizophrenia and to identify heart failure patients most likely to benefit from beta-blockers.11-13 In some cases, PGS rival monogenic mutations in risk prediction while also being more broadly applicable across populations. Clinical translation of PGS is already underway. For example, the Centre for Familial Breast and Ovarian Cancer in Cologne, Germany, offers CanRisk. This CE-certified web tool integrates family history, rare TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 18 pathogenic variants in cancer susceptibility genes, PGS, lifestyle, hormonal, and clinical factors as well as imaging data to predict breast and ovarian cancer risks and estimate the likelihood of carrying pathogenic variants in specific genes. 4. Machine learning and AI Pharmacogenomics generates vast, multidimensional datasets from genomic sequences and transcriptomes to clinical outcomes. Extracting actionable insights from this data requires computational approaches that can detect subtle, non-linear patterns. Deep learning algorithms have already been used in various aspects of genetics and pharmacogenomics to improve the understanding of genetic variation and its impact on drug response. For example, a deep learning model called Hubble.2D6 was developed to predict the functional impact of CYP2D6 haplotypes directly from DNA sequence data.14 Similarly, a neural network was trained on a cohort of breast cancer patients to predict CYP2D6-mediated tamoxifen metabolism.15 However, both studies were limited, and further research on a broader set of genes is needed. Future improvements rely on better training datasets, such as those from deep mutational scanning. By combining structural predictions with experimental data from these scans, models like AlphaMissense can refine their predictions based on validated functional impacts, increasing reliability in clinical and research settings.16 Expanding training datasets to include a broader range of variants, particularly those linked to diverse phenotypes and populations, will enhance the model's generalizability. Incorporating multiomics data, like transcriptomics and proteomics, may further improve the ability to predict the pathogenicity of variants in complex biological contexts, advancing personalized medicine and gene therapy strategies. 5. Integrating multiomics data Numerous factors beyond genetics influence how medications are absorbed, distributed, metabolized, and eliminated, impacting both efficacy and the risk of adverse effects. Integrating multiple layers of biological information, including genetics, epigenetics, metabolomics, proteomics, nutrition, and microbiome data, offers a comprehensive approach to optimizing therapeutic outcomes.17 Key advancements include: ∙ Single-cell multiomics technologies, which reveal how gene expression and chromatin states vary within cell subpopulations in response to treatment. This is crucial for tackling heterogeneity in cancer and immune diseases. ∙ Spatial omics, which helps map where pharmacogenomic effects occur in tissue context, uncovering why drugs may work in some cell niches but not others. ∙ Proteogenomics, enabling researchers to link genetic variants to altered protein networks and downstream drug targets, supporting rational drug design and repurposing. By integrating these data types, scientists can identify regulatory variants, post-translational modifications, and context-specific pathways that modulate drug response. For example, one study analyzing tumors from patients resistant to immune checkpoint inhibitors combined transcriptomic, epigenomic, and spatial profiling to reveal immune-suppressive states that were undetectable through genomic data alone.18 These findings highlight the importance of incorporating diverse omics approaches to better understand, and ultimately overcome, therapy resistance and adverse effects. Despite their potential, multiomic approaches are still difficult to implement clinically due to high costs, technical complexities in data integration and limited validation through clinical outcomes. However, as analytical tools improve and more outcome-linked datasets become available, these approaches are poised to become foundational in next-generation precision pharmacotherapy. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 19 Looking Ahead Seventy years of methodological development have transformed pharmacogenomics from an emerging science into an interdisciplinary area of research that is key to the implementation of personalized medicine. As we move towards a future where pharmacogenomics becomes an integral part of clinical decision-making, advances – such as LRS, multiomics, and machine learning – are setting the stage for more personalized, precise, and effective healthcare. The next generation of pharmacogenomics promises to deliver treatments that are not only informed by genetic data but also by the complex interactions within the multiscale biological network, offering the potential to minimize adverse drug reactions and maximize therapeutic efficacy. REFERENCES 1. 1. Ingelman-Sundberg M, Nebert DW, Lauschke VM. Emerging trends in pharmacogenomics: from common variant associations toward comprehensive genomic profiling. Hum Genom. 2023;17(1):105. doi: 10.1186/s40246-023-00554-9 2. 2. Weinshilboum RM, Wang L. Pharmacogenomics: Precision medicine and drug response. Mayo Clin Proc. 2017;92(11):1711- 1722. doi: 10.1016/j.mayocp.2017.09.001 3. 3. Table of pharmacogenomic biomarkers in drug labeling. FDA. September 23, 2024. Accessed July 22, 2025. https:// www.fda.gov/drugs/science-and-research-drugs/tablepharmacogenomic-biomarkers-drug-labeling 4. 4. Ingelman-Sundberg M. Pharmacogenetics of cytochrome P450 and its applications in drug therapy: the past, present and future. Trends Pharmacol Sci. 2004;25(4):193-200. doi: 10.1016/j. tips.2004.02.007 5. 5. Ammar R, Paton TA, Torti D, Shlien A, Bader GD. Long read nanopore sequencing for detection of HLA and CYP2D6 variants and haplotypes. Published online May 20, 2015. doi: 10.12688/ f1000research.6037.2 6. 6. Mantere T, Kersten S, Hoischen A. Long-Read sequencing emerging in medical genetics. Front Genet. 2019;10. doi: 10.3389/ fgene.2019.00426 7. 7. Pollard MO, Gurdasani D, Mentzer AJ, Porter T, Sandhu MS. Long reads: their purpose and place. Hum Mol Genet. 2018;27(R2):R234-R241. doi:10.1093/hmg/ddy177 8. 8. Dinh-Le C, Chuang R, Chokshi S, Mann D. Wearable health technology and electronic health record integration: Scoping review and future directions. JMIR mHealth uHealth. 2019;7(9):e12861. doi: 10.2196/12861 9. 9. Turner AJ, Haidar CE, Yang W, et al. Updated DPYD HapB3 haplotype structure and implications for pharmacogenomic testing. Clin Transl Sci. 2024;17(1):e13699. doi: 10.1111/cts.13699 10. 10. Auwerx C, Sadler MC, Reymond A, Kutalik Z. From pharmacogenetics to pharmaco-omics: Milestones and future directions. HGG Adv. 2022;3(2):100100. doi: 10.1016/j. xhgg.2022.100100 11. 11. Li JH, Szczerbinski L, Dawed AY, et al. A polygenic score for type 2 diabetes risk is associated with both the acute and sustained response to sulfonylureas. Diabetes. 2020;70(1):293- 300. doi: 10.2337/db20-0530 12. 12. Li J, Yoshikawa A, Brennan MD, Ramsey TL, Meltzer HY. Genetic predictors of antipsychotic response to lurasidone identified in a genome wide association study and by schizophrenia risk genes. Schizophr Res. 2018;192:194-204. doi: 10.1016/j.schres.2017.04.009 13. 13. Lanfear DE, Luzum JA, She R, et al. Polygenic score for β-blocker survival benefit in European ancestry patients with reduced ejection fraction heart failure. Circ Heart Fail. 2020;13(12):e007012. doi:10.1161/CIRCHEARTFAILURE.119.007012 14. 14. McInnes G, Dalton R, Sangkuhl K, et al. Transfer learning enables prediction of CYP2D6 haplotype function. PLoS Comput Biol. 2020;16(11):e1008399. doi: 10.1371/journal.pcbi.1008399 15. 15. van der Lee M, Allard WG, Vossen RHAM, et al. Toward predicting CYP2D6-mediated variable drug response from CYP2D6 gene sequencing data. Sci Transl Med. 2021;13(603):eabf3637. doi: 10.1126/scitranslmed.abf3637 16. 16. Cheng J, Novati G, Pan J, et al. Accurate proteome-wide missense variant effect prediction with AlphaMissense. Science. 2023;381(6664):eadg7492. doi: 10.1126/science.adg7492 17. 17. Shaman JA. The future of pharmacogenomics: Integrating epigenetics, nutrigenomics, and beyond. J Pers Med. 2024;14(12):1121. doi: 10.3390/jpm14121121 18. 18. Wen J, Wang Y, Wang S, et al. Genetic and transcriptional insights into immune checkpoint blockade response and survival: lessons from melanoma and beyond. J Transl Med. 2025;23(1):467. doi: 10.1186/s12967-025-06467-6 20 DRUG DISCOVERY: ADVANCES AND INNOVATIONS Offering Fresh Hope: GameChanging Drug Discoveries for Rare Diseases Kerry Taylor-Smith Rare diseases affect a very small portion of the population, typically less than 1 in 2,000 people. Around 80% of rare diseases have a genetic origin, and are often biologically complex, meaning that they are not very commercially attractive to the pharmaceutical industry.1 But the people living with these life limiting conditions need treatment to ease or treat their symptoms, so how are drugs developed for such conditions? Patient‑led organizations have stepped in to fund the costly, early‑stage academic development of treatments for rare diseases. Two such examples are the AT Society and Cystic Fibrosis Trust, both of whom are supporting researchers on the frontline of developing new drugs to treat the symptoms of ataxia telangiectasia (AT) and cystic fibrosis (CF). Combined strategies for patient tailored treatment for AT There are an estimated 200 cases of AT in the UK; this rare, inherited condition is complex and life limiting, and leaves children practically wheelchair-bound by the age of 10. It is sometimes referred to as a “multi-system” disorder because it affects several different organs or systems within the body, but there is a lot of variability between individuals and no two people with AT will have the exact same symptoms.2,3 Symptoms are not immediately apparent at birth; they tend to develop slowly over time until the age of seven or eight when they worsen more quickly. Ataxia or lack of coordination is usually the first to appear, followed Credit: iStock/ TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 21 by difficulties with speech and swallowing, and then fatigue, low weight, and slow growth. Around two-thirds of individuals will have a weakened immune system, making them more susceptible to frequent coughs, colds, and infections of the throat, ear, sinuses, and sometimes lungs. There is also a greatly increased risk of cancer; roughly 25% of people with AT will develop cancer, usually lymphomas or leukemia in children, and solid tumors such as breast cancer and cancer of the esophagus in adults.3 AT is caused by a mutation in a single gene—the ATM gene, which is involved in DNA repair, cell division, and maintaining genetic stability. The mutation prevents the ATM protein from being produced. Understanding how this deficit causes the different symptoms of AT would make it easier to find drugs to intervene in these processes.4 Gene therapy might be the most obvious choice, but the ATM gene is a quite large—66 exons and more than 150 kb of DNA —making it unsuitable for traditional gene therapy using viruses. To overcome this issue, Dr. James Dixon and his research group at the University of Nottingham Biodiscovery Institute have been focusing on advanced non-viral delivery technologies, developing peptide-based nanoparticles that can safely and efficiently deliver large therapeutic genes and gene editing tools into hard-to-reach tissues like the brain. “This is particularly important for AT, where neurodegeneration drives many of the most devastating symptoms and where conventional viral gene therapies are unsuitable due to the size of the ATM gene and safety concerns,” explained Dixon. “By combining innovative nanoparticle chemistry with state-of-the-art gene editing approaches, we aim to restore ATM activity either by correcting faulty genes or by providing new, functional copies in affected cells.” Dixon’s work, which includes collaborators from the UK, USA, and EU, is helping to develop new drugs for AT “by creating a realistic and scalable pathway to a permanent, disease-modifying therapy rather than symptomatic treatment. Our platform is designed to be flexible, allowing repeated or single-dose administration, low immunogenicity, and precise control over where and how therapeutic genes act.” And the technology doesn’t just apply to AT; it can be adapted for other genetic and neurodegenerative disorders, accelerating the broader development of next-generation nucleic-acid medicines. “Ultimately, this research lays the groundwork for first-in-human trials and offers real hope of long-term benefit, improved quality of life, and potentially a cure for people living with AT,” Dixon added. Dixon’s team are also developing antisense oligonucleotide (ASO) strategies as a complementary and alternative therapeutic route. “ASOs offer a powerful way to modulate ATM expression or correct specific splicing defects without permanently altering the genome,” Dixon said. “This approach could be especially valuable for patients with particular mutation classes or as an earlier-stage intervention while gene therapy approaches continue to mature. Importantly, the same brain-penetrating nanoparticle technologies we have developed for gene delivery can also be adapted to deliver ASOs efficiently to the central nervous system, creating a unified delivery platform for multiple therapeutic modalities.” Dixon believes this combined strategy could offer flexible, patient-tailored treatments for AT. “Permanent ATM gene augmentation or editing offers the possibility of a one-off, lifelong correction, while ASO-based therapies provide a reversible, adjustable alternative that may reach the clinic more rapidly for some patients.” By developing and validating both approaches in parallel, the research “maximizes the chances of delivering effective new drugs for AT and ensures that emerging therapies can be matched to patient need, disease stage, and long-term safety considerations,” Dixon said. And the work could make “a profound difference” to the lives of those with AT by “addressing the disease at its root cause rather than only managing symptoms,” Dixon added. “By restoring ATM function in affected cells—particularly TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 22 in the brain, where neurodegeneration drives loss of movement, speech, and independence—these approaches aim to slow, halt, or even prevent the progression of the most debilitating aspects of the condition.” If successful, this could mean children with AT retain their mobility, coordination, and communication for longer, which would improve independence, education, and social participation. In the longer term, Dixon believes therapies based on ATM gene replacement, gene editing, or ASO could reduce the need for repeated hospital visits and invasive supportive care by offering durable or even permanent correction with a single or limited number of treatments. “This would not only improve quality of life for patients, but also significantly reduce the emotional, physical, and financial burden on families and caregivers,” Dixon said. “Importantly, by developing multiple complementary treatment strategies, this work increases the likelihood that effective therapies can be tailored to different patients, disease stages, and mutation types— bringing real hope of longer, healthier lives for people living with AT.” Small molecule drugs for CF CF is well-known—it’s the most common genetic disease in the Caucasian population—but is considered rare, affecting just 10,000 people in the UK. Around one in 25 people carry a faulty CF gene, usually without knowing, and if two carriers have a child, that child has a one-in-four chance of having CF.5,6 “CF is an inherited autosomal recessive condition,” explained Dr. Maya Desai, a retired consultant respiratory pediatrician at Birmingham Women's and Children's NHS Foundation Trust. “The most common CF-causing genetic variants occur in white populations although CF can be found in almost every other ethnicity.” CF is caused by a defect in the cystic fibrosis transmembrane conductance regulator (CFTR) gene and the protein it produces; over 2,000 specific gene mutations have been identified so far.5 Defects in the gene “result in chloride channel dysfunction in many cells of the human body,” said Desai. This causes a thick, sticky mucus to collect in the lungs and digestive system of individuals with CF. “Main manifestations are recurrent chest infections and the progression to bronchiectasis and pancreatic insufficiency which results in failure to thrive and malnutrition,” Desai added. Most people are treated with tablets or treatments focused on preventing and treating the results of the defect, but more recently therapies have targeted the faulty CFTR protein and its production.5,6 The CFTR protein is an attractive therapeutic target, and small molecule CFTR modulators like Kaftrio® (ivacaftor/tezacaftor/elexacaftor), Symkevi® (tezacaftor/ ivacaftor), and Orkambi® (lumacaftor/ivacaftor) have been developed to target the underlying cause of CF by helping the CFTR protein work more effectively. The modulators help regulate the flow of water and chloride in and out of cells, and when chloride moves more normally, mucus in the lungs and other organs becomes thinner and less sticky.7,8 “Small molecule drugs interact with the process of CF gene expression and protein production to either increase gene expression or improve the protein folding and function. Current modulators are either CFTR correctors or potentiators, usually given in combination. By increasing the amount of functioning protein available, either by increasing production or improving its function, the chloride channel can work more normally as it does in healthy people,” Desai said. These modulators were developed through a long process of developing candidate molecules and refining their properties using in vitro models, Desai said. Firstly, functional assays that examine chloride channel activity were developed to allow scientists to evaluate new treatments. Then, high-throughput screening rapidly tested thousands of compounds in a short space of time in search of potential therapeutic “hits,” which can be further developed for clinical use.5 TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 23 There are currently five CFTR modulators approved for use by the NHS in England and these treatments work for around 90% of people with CF, but research is aimed at developing new treatments for the remaining 10%.5,8 But for now, standard care involves “antibiotics for infections, chest physiotherapy, nebulized mucolytics, pancreatic enzyme supplements, vitamins, hospital admissions, multidisciplinary care delivered in specialist centers,” Desai said. “The modulators are given in addition. In time, some elements of standard care will be reduced.” Offering hope Living with a rare disease is confusing, worrying, and exhausting. These diseases are complex, and it’s not that they are poorly understood, rather that treatments and cures might not be as readily available as they are for other diseases. Drug development can be costly and lengthy, and pharmaceutical companies lean more towards drugs that are profitable and will benefit a greater number of individuals. However, treatments for rare diseases should not be overlooked. Thanks to charities like the AT Society and Cystic Fibrosis Trust, vital research into what causes these diseases and how their progression can be slowed or symptoms lessened can continue, offering hope for patients that one day, a drug to treat their rare disease might become a reality. REFERENCES 1. What is a rare condition? NHS England Digital. https://digital.nhs. uk/data-and-information/publications/statistical/rare-conditionregistration-statistics/updated-to-2022/what-is-a-rare-condition. Published 2025. Accessed January 23, 2026. 2. For professionals. AT Society. http://atsociety.org.uk/forprofessionals/. Accessed January 26, 2026. 3. How does AT affect people? AT Society. https://atsociety.org.uk/ about-a-t/how-does-a-t-affect-people/. Accessed January 26, 2026. 4. The ATM gene and protein. AT Society https://atsociety.org.uk/ about-a-t/the-atm-protein/. Accessed January 26, 2026. 5. Hine C, Nagakumar P, Desai M. Small molecule drugs in cystic fibrosis. Arch Dis Child Educ Pract Ed. 2022;107:379-382. doi: 10.1136/archdischild-2020-319009 6. Cystic Fibrosis FAQ, Cystic Fibrosis Trust. https://www. cysticfibrosis.org.uk/what-is-cystic-fibrosis/faqs. Accessed January 26, 2026. 7. How is Cystic Fibrosis treated? Cystic Fibrosis Trust. https://www. cysticfibrosis.org.uk/what-is-cystic-fibrosis/cystic-fibrosis-care. Accessed January 26, 2026. 8. Modulators, Cystic Fibrosis Trust. https://www.cysticfibrosis.org. uk/what-is-cystic-fibrosis/cystic-fibrosis-care/treatments-andmedication/modulators. Accessed January 26, 2026. MEET THE INTERVIEWEES: James Dixon is an associate professor of Stem Cell & Gene Therapy Technologies in the Division of Regenerative Medicine & Cellular Therapies (RMCT) at the University of Nottingham Biodiscovery Institute (BDI). Maya Desai was a consultant respiratory pediatrician at Birmingham Children’s Hospital from 2002 to 2024. She was clinical lead of the pediatric CF network, delivering care to around 300 children and young people with CF. She is a Trustee at the CF Trust. Click here to view the full infographic The development of pharmaceutical products relies on the latest technological developments. Just as any structure requires solid foundations, innovations in biotechnology support the entire industry. Biotechnology is the foundation of drug development Recombinant DNA technology This cornerstone of modern biotechnology provides a means to combine DNA from different sources to create novel therapeutic proteins and peptides. Cell culture systems These systems allow the growth of cells outside their natural environment, providing controlled conditions for producing biologics at scale. This includes the use of CHO cells for antibody production. Monoclonal antibody production It is now possible to develop highly specific antibodies that target disease mechanisms with unprecedented precision. Gene editing tools CRISPR-Cas9 and similar technologies enable targeted modifications to correct genetic defects, such as those in sickle cell disease. 25 DRUG DISCOVERY: ADVANCES AND INNOVATIONS Rethinking Target-Based Drug Discovery: Challenges, Innovations and the Next S-Curve Laura Elizabeth Lansdowne Drug discovery has long relied on two key strategies: target-based and phenotypic screening. Recombinant technologies and genomics opened the door to modern target-based drug discovery (TBDD), allowing researchers to identify key genes implicated in disease and screen large compound libraries against defined molecular targets. As sequencing of the human genome edged closer to completion, the number of potential targets grew rapidly, fueling enthusiasm for target-led approaches. This was mirrored in regulatory outcomes. For example, between 1999 and 2013, 69% of firstin-class regulatory approvals (40% small molecules, 29% biologics) made by the US Food and Drug Administration came from target-based discovery. However, as the approach was adopted more widely, its limitations became more visible. Many drug discovery programs that looked promising early on struggled to translate later in development. A recent systematic review of over 30,000 studies spanning 150 years found that only ~ 9% of approved small-molecule drugs were truly discovered through target-based methods, and even those that were often relied on off-target mechanisms for their therapeutic effects. Credit: iStock/PollyW TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 26 By the mid-2000s, experienced drug discovery leaders started to question if the industry’s pivot towards target-driven strategies had introduced new bottlenecks. Among them was Dr. David Brown, who has decades of experience in pharmaceutical research and biotech leadership, serving at several top pharmaceutical companies and contributing to multiple successful therapeutic programs. Drawing on S-curve theory (Figure 1), Brown argued in 2007 that after an initial period of rapid progress, target-based discovery had entered a productivity plateau. As one industry manager observed at the time: “For the past decade, the pharmaceutical industry has experienced a steady decline in productivity, and a striking observation is that the decline coincided with the introduction of TBDD.” In this article, we examine ongoing challenges shaping TBDD and highlight how new technologies and interdisciplinary approaches are helping to address them. The enduring challenge of target validation Even targets backed by solid mechanistic reasoning can fail once they reach patients. The field has become fixated on selecting a single target to treat a disease, even though human biology is far more complex. A drug acting on one target often cannot capture that complexity. Recombinant systems add further limitations: they don’t reliably reflect human physiology, so a target that behaves well in a cell-based assay or model organism may not translate effectively. This gap between preclinical promise and real therapeutic effect – known as the “valley of death” – remains one of the major bottlenecks in drug discovery. “In the 1990s, with genomic data increasingly available, pharmaceutical companies changed almost entirely to target-based approaches and abandoned most phenotypic research. That was a gamble that only partly paid off, and in retrospect, I think we can say that a more balanced approach would have been more productive for the industry,” noted Brown. Time Quantity of Users S-Curve Innovators Early Adopters Early Majority Late Majority Induction Payback Obsolescence Laggards FIGURE 1: S-CURVE MODEL OF TECHNOLOGY ADOPTION AND PERFORMANCE MODES (INDUCTION, PAYBACK, OBSOLESCENCE). ADAPTED WITH PERMISSION FROM DAVID BROWN. CREDIT: TECHNOLOGY NETWORKS. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 27 “Most drugs act on multiple targets, so optimizing for a single target has undoubtedly reduced the rate of breakthrough inventions. Most drugs optimized for a single target later fail in clinical trials, after many years of effort and expenditure,” he added. Brown’s decades in the field have taught him that human bias plays a major role in how targets are selected, often leading to failures downstream. On top of that, the industry tends to focus on the same targets. Artificial intelligence (AI), however, offers a new approach with reduced bias. “We can let the data point us to the right targets (plural) and the right molecules. My latest drug, HLX-1502, for the treatment of Neurofibromatosis 1 tumors came from this approach. It is currently in a Phase 2a clinical trial in the USA. We can even let the data design molecules for us,” he said. When considering TBDD risk, it can be broadly split into two categories. “Validation risk” reflects the uncertainty that modulating a target will produce a meaningful therapeutic effect. “Technical risk” relates to the practical challenge of finding a molecule that can safely reach, engage, and modulate the specific target in humans. Both risks can derail pipelines, even late in development. Dr. Rachael Dickman, lecturer in drug discovery at University College London, highlighted some of the key issues she faces: “Challenges in peptide discovery in my experience are mostly regarding formulation/ delivery and stability. With the use of peptide library technologies (e.g., phage display and mRNA display), it is now possible to find ligands with high affinity (low nM or pM) for a desired target much more quickly, but whether the identified peptides are sufficiently ‘druglike’ to reach the target tissue or have long enough halflife to be therapeutically useful is still challenging.” Robust validation is essential to “stress-test” a target before significant time and money are invested in advancing a compound. She noted that “use of tool compounds to demonstrate that target modulation has the desired effect on relevant biomarkers and pharmacology, both in vitro and in vivo, gives additional confidence that the target is likely to be druggable.” Her lab is developing peptide drugs for infections and other diseases, particularly those driven by antibiotic resistance. To do this, they synthesize complex cyclic peptides and design analogues via solidphase chemistry, then explore their structures and mechanisms using NMR spectroscopy and a range of biophysical tools. Emerging tools and modalities in targetbased drug discovery More recently, new tools and approaches are emerging with the potential to reshape the field. For example, in peptide discovery, advances in library technologies and rational chemical modifications are helping researchers overcome stability and tissue-delivery challenges. Meanwhile, computational methods enable rapid in silico screening and optimization of candidate molecules, reducing both time and cost. Dr. Avner Schlessinger, professor of pharmacological sciences at the Icahn School of Medicine at Mount Sinai, leads a lab focused on streamlining drug discovery through computational chemistry and AI to study disease pathways. He explained the transformative power of AI: “Traditional medicinal chemistry explores chemical space through manual, iterative optimization ‒ a process that is often slow and heavily dependent on human intuition.” “AI dramatically accelerates and expands this process, particularly when combined with expert-driven insight. In our center, we routinely use deep-learning-based methods to address biological problems, such as in molecular docking and active learning, where we virtually screen libraries with billions of synthetically accessible compounds, guided by AI models that improve with each iteration.” TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 28 By integrating patient-derived multiomics datasets, network-based AI models, and structure-based design, his team can now identify disease-relevant targets that are both genetically implicated and druggable: “Using these approaches, we have identified novel targets for various central nervous system disorders and cancer, and have applied AI-driven drug discovery and structural biology to develop promising tool compounds.” “Since our focus is on the preclinical stage rather than advancing compounds directly into the clinic, we prioritize the development of high-quality tool compounds that rigorously validate a target’s function in disease-relevant models. This helps de-risk the biology early and supports downstream therapeutic investment,” he added. “AI enables fast, iterative feedback loops between target identification, molecular modeling, and experimental validation. This dramatically accelerates the timeline from target discovery to hit optimization, which was previously a slow and resource-intensive process,” Schlessinger said. While this approach can increase confidence and help identify the most promising targets, it’s worth noting that it does not fully overcome the inherent limitations of a single-target approach. It’s important we keep the complexity of disease front of mind: multiple targets and factors often influence a drug’s efficacy and safety. Improving the success of TBDD requires leaving this single-target mindset behind ‒ it isn’t helpful to become too fixated on one gene or protein. Approaching the next S-curve: rethinking productivity in drug discovery Using Brown’s S-curve framework, we can reflect on productivity and look to predict the possible next phase of drug discovery. The initial surge in performance following the switch from phenotypic to target-based approaches eventually plateaued, partly because the field tried adopting the new technology before it had fully matured. “Target-based strategies started to become feasible at scale in the 1980s with the invention of fast protein liquid chromatography, which greatly accelerated the isolation of individual biochemicals at purity levels suitable for testing drug molecules (Early Adopters). Then, it accelerated in the 1990s (Early Majority), when genomic data began to appear. The obsolescence phase has been with us since the mid-to-late 2000s. We needed new approaches by that time, but we’ve had to wait until now,” explained Brown. As a result, there’s been a gap of almost a generation between the second and third S-curve (Figure 2). The timing of S-curves matters just as much as their shape: “The key point is that any industry needs a new S-curve to kick in before or at the time an old one begins to taper off at the top of the S, with diminishing productivity (Obsolescence Mode). If there’s a gap between one S-curve and the next, an industry is in trouble; its productivity languishes. Unfortunately, that has happened in the pharmaceutical/biotech industry.” Figure 2, which shows the three waves of drug discovery productivity, illustrates this dynamic. Brown has witnessed all three phases: “During my career of over 50 years inventing new medicines, I have seen three waves: two completed S-curves and a third underway – it’s now at the end of the flat induction phase and beginning to pay off. But there has been a gap between the second and third waves. Just as the first wave ran its course by the late 1980s, the second plateaued by about 2010. We needed a new third wave to be in payback mode, but it was actually only partway through the bottom line of the third S-curve, between the ‘Innovators’ and ‘Early Adopters’ phases.” Brown now believes the field is moving beyond the early adoption phase of the third S-curve (Figure 2, final green data point) and entering the growth/payback phase. In doing so, we’ll be able to compensate for the plateau and decline (second orange data point) at the end of the second S-curve. This transition also brings a shift in methodology. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 29 Brown thinks the coexistence and reintegration of multiple drug-discovery approaches is key to this new wave: “The first wave of phenotypic drug discovery was largely replaced by the second wave of TBDD and now a third wave of AI-based multiomics methods is coming into play. Actually, I think all three approaches have their uses if used appropriately.” Both Dickman and Schlessinger recognize that drug discovery is entering a phase of major change. Dickman noted, “I think new modalities are especially important ‒ as we try to develop therapies for increasingly complex diseases (e.g., neurodegenerative diseases, cancers) new modalities will be needed to effectively modulate targets.” “AI is driving a paradigm shift in drug discovery by enabling systems-level targeting and rapid iteration. It’s now possible to identify not just individual targets but contextspecific vulnerabilities such as those unique to particular disease states, cellular environments, or genetically defined patient subpopulations,” said Schlessinger. Brown believes AI/machine learning using multiomics data is key: “This is our best chance to improve on the awful record of recent decades, in which the majority of preclinical and clinical projects have failed after years of work and huge amounts of money spent, and often it’s because of the single target that was selected in the very first step of the drug discovery process.” Collaboration across academia, startups and industry As the field works to overcome long-standing bottlenecks in TBDD and push toward the payback phase of the next S-curve, collaboration becomes even more important. Academic researchers tend to have the freedom to take on higher-risk projects and explore novel mechanisms, whereas industry professionals bring expertise in clinical development and experience navigating regulatory requirements to produce medicines at scale. Innovation Window Innovation Window Innovation Window Revenue & Product Version & Time Plateau & Decline Maturity Point Early Adoption Original Product or Strategy New Product or Growth Strategy New Product or Growth Strategy Growth Growth Growth FIGURE 2: SUCCESSIVE WAVES OF DRUG DISCOVERY PRODUCTIVITY. ADAPTED WITH PERMISSION FROM DAVID BROWN. CREDIT: TECHNOLOGY NETWORKS. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 30 However, a key driver of current innovation is techbio startups. These companies, which are often funded by venture capital at a scale beyond academic labs, are at the forefront of developing new tools, modalities, and approaches in drug discovery. They bridge the gap between early-stage research and laterstage industry development. Dickman highlighted the value of this synergy: “From an academic perspective, collaborating with industry in drug discovery is hugely valuable, not only because of the expertise across drug development, but also because it helps focus research efforts, e.g., towards something which is potentially commercially viable and therefore may eventually be of benefit to patients.” Schlessinger added that these partnerships are strengthened through training pathways: “Many of our PhD students and postdocs gain experience in translational AI and drug discovery through internships with startups or industry, creating a longterm, bidirectional flow of talent and knowledge between sectors.” Conclusion Advances in chemical biology, structural methods, multiomics, and AI-driven modeling are reshaping how researchers identify and validate targets. New therapeutic modalities, particularly peptides and targeted protein degradation, are expanding what counts as “druggable”. At the same time, there now seems to be more emphasis on building stronger biological evidence before design begins, helping to reduce avoidable failures later in development. Hopefully, these tools and lessons will translate into a more consistent flow of effective new drugs, particularly for complex diseases. “Used together, these approaches [AI and multiomics] provide an alternative approach to drug discovery that bridges chemistry and biology, removes human bias and lets the data tell us what should work. This is the future, the third wave, the third S-curve,” concluded Brown. MEET THE INTERVIEWEES: Dr. David Brown has over 50 years of experience in medical research and leadership in the pharmaceutical and biotechnology industry. He has held senior roles at Zeneca, Pfizer, GlaxoWellcome and Hoffman La-Roche, and served as President and CEO of Cellzome AG. At Pfizer, he co-invented Viagra and led its development to proof of clinical efficacy, and contributed to the discovery of Relpax® (eletriptan HBr), indicated for acute migraine headaches in adults. Brown is an entrepreneur and advisor, cofounding Healx Ltd, where he serves as a non-executive board member, as well as Crescendo Biologics. He earned his PhD in chemistry from the University of Bristol. Dr. Rachael Dickman is a lecturer in drug discovery at University College London (UCL). She earned her PhD from UCL in 2018 under the supervision of Professor Alethea Tabor. Dickman’s research focuses on developing peptide drugs to treat infections and diseases. Her group studies antimicrobial compounds from nature to address antibiotic resistance, synthesizes complex cyclic peptides and analogs and investigates their structure and mechanisms using NMR and biophysical methods. She also works on synthesizing unusual amino acids and exploring peptide-based treatments for neurodegeneration. Dr. Avner Schlessinger is a professor of pharmacological sciences at the Icahn School of Medicine at Mount Sinai in New York City, director of the AI Small Molecule Drug Discovery Center, and codirector of the Disease Mechanisms and Therapeutics (DMT) Training Area at Mount Sinai. He earned his BSc in biology and chemistry from Tel Aviv University and completed his PhD in biochemistry and molecular biophysics at Columbia University. Schlessinger’s research focuses on improving and automating drug discovery by integrating computational chemistry and artificial intelligence to characterize disease pathways. 31 DRUG DISCOVERY: ADVANCES AND INNOVATIONS Cancer Drug Discovery: Reaching for the High‑Hanging Fruit Joanna Owens, PhD In recent decades, there’s been unprecedented progress in cancer drug discovery – from the original targeted kinase inhibitors exploiting molecular weaknesses in cancer cells, to the breakthrough of antibody-based therapeutics and checkpoint inhibitors.1 Despite this progress, many of the most important mutated genes in cancer cells remained undruggable.2,3 Today, there are drugs in the clinic and many more in development against these intractable targets.2,3 In this article, we explore some of the latest approaches being used to target these infamous molecules. Turning KRAS into a cell-surface antigen About 10 years ago, Dr. Charles Craik was in the audience at UCSF when his colleague, chemical biologist Dr. Kevan Shokat, presented the progress they were making with developing inhibitors against Kirsten rat sarcoma viral oncogene homolog (KRAS). They had developed one of the first compounds that could block the common KRAS mutation, G12C – the nowapproved cancer drug sotorasib. “But the problem that occurs with all these drugs is cancers develop resistance, so we thought, let’s start thinking ahead of the game,” Craik recalled. Craik’s lab specializes in generating unique conformationally select antibodies that you can’t normally generate from a hybridoma. “We wondered, would the covalently bound KRAS-inhibitor complex show up on the surface of the cell and be detectable by an antibody?” Credit: iStock/Nadezhda Buravleva TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 32 By combining the strengths of his and Shokat’s labs, they forged a way toward an antibody-based KRAS therapeutic that could overcome resistance and benefit a large population of patients. The first step was to prove the hypothesis that modified KRAS peptides could make it through the proteosome, be delivered onto major histocompatibility complex (MHC) alleles and be displayed on the cell surface. They started by synthesizing two KRAS peptides each with a KRAS inhibitor covalently bound to the cysteine group and found that these peptides readily formed functional MHC class I complexes to form “haptens” that could be recognized by an antibody.4 “Once we saw this was feasible, we knew we had opened up so many possibilities,” said Craik. “Now, when resistance to KRAS inhibitors develops in a patient, we have a handle to go after.” However, a further challenge with intracellular antigens is their density on the cell surface. “KRAS is not an abundant protein in the first place and once it’s loaded onto the MHC complexes, there’s probably fewer than 1,000 receptors per cell – and not on every cell,” said Craik. To address this, the team is exploring targeted radiolabeled antibodies to develop a highly potent payload for antigen-bearing cancer cells. “In theory, you can target one receptor with one radionuclide, and because of the bystander effect, the drug will anchor on a single cell and hit all the surrounding tumor cells.” An antibody carrying the beta emitter lutetium has shown the principle works in animals and now they’ve shown in animal models that actinium – an alpha emitter – might be even more effective and safe.5 “Alpha emitters are extremely potent, but the radiation only travels around five cell lengths, so they might be considered much safer than other types of radionuclides.” Opening the door to targeting other intracellular antigens Beyond solving resistance for patients taking KRAS inhibitors, this strategy also addresses another common issue in cancer drug discovery – the challenge of finding and targeting cancer-specific antigens versus cancerassociated antigens. “If drugs aren’t targeting a tumor-specific antigen, we’ll always have a narrow therapeutic index between hitting the cancer cell versus the same antigens on normal cells,” said Craik. “There are actually very few truly tumor-specific antigens and most of them are intracellular, like KRAS.” This approach opens the door to targeting other intracellular oncogenes that have previously been impossible to reach with antibody-based therapeutics. By using any irreversible inhibitor that binds to the target and stays in the tumor cells – regardless of whether it blocks its activity – it becomes feasible to label any target making it detectable on the cell surface. Making quantum leaps in cancer drug discovery KRAS is also the initial focus of Dr. Igor Stagljar and Dr. Alán Aspuru-Guzik, at the University of Toronto Aspuru-Guzik is developing advanced algorithms supported by quantum computing to identify prototype drugs and has already generated several compounds targeting KRAS mutations. Meanwhile, Stagljar’s lab has developed the mammalian-membrane two-hybrid (MaMTH) assay, a live cell-based platform that enables the detection of interactions between membrane-bound proteins and their partners and adapted this into a drug discovery platform capable of identifying smallmolecule inhibitors and activators of cancer-causing mutations in a wide range of membrane proteins. 6,7 “With MaMTH, you can monitor the interaction between wild-type and mutant KRAS and their effector proteins and assess whether candidate molecules selectively inhibit these interactions.” Stagljar’s team had built a comprehensive toolbox of KRAS mutants over many years, allowing them to rapidly test whether the compounds identified through quantum computing/ artificial intelligence (AI) were specifically targeting KRAS, delivering answers within just a few weeks. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 33 “Even with the breakthrough of the covalent inhibitors and now pan inhibitors, these compounds aren’t performing that well clinically,” said Stagljar. “We need much better and more potent clinical molecules for all the other KRAS mutants.” The new quantum-computing/AI-enabled approach uses generative machine learning (ML) models, which can understand the underlying distribution of atoms and bonds in a dataset and then construct new molecules with predefined properties.8 Although classical ML algorithms have significantly aided advances in drug discovery, combining these approaches with quantum ML algorithms can enhance target space exploration, selecting compounds out of the trillion molecules in the cloud or using generative AI to generate candidate molecules predicted to fit. “With classical ML approaches, the positive hit rate was very low – less than 1% depending on the target,” said Stagljar. “But the power of these new algorithms has made it possible to fit molecules into cryptic pockets.” This is crucial for a challenging target like KRAS which has a very smooth surface. Stagljar’s team tested 14 compounds identified by their hybrid quantum-classical generative model in combination with generative AI and found that two of them selectively inhibited mutant KRAS in the assay, while the others primarily affected wild-type KRAS.8 “This preclinical drug discovery stage usually takes between four and six years,” said Stagljar. “But using quantum computing and AI, we’ve shortened this timeframe to about three to four months.” Their work continues to investigate KRAS, but the team is now evolving the approach for other difficult targets such as transcription factors, ubiquitin ligases, some phosphatases, and other small GTPases. “This is where we need to collaborate because the computational tools get us the starting points, but we need to validate those small molecule hits in living cancer cells using specific functional assays, which we can do very rapidly in our lab,” said Stagljar. “I believe this is the future of cancer drug discovery: test quickly, fail fast and move forward with the most promising candidates.” REFERENCES 1. Falzone L, Salomone S, Libra M. Evolution of cancer pharmacological treatments at the yurn of the third millennium. Front Pharmacol. 2018;9:1300. doi: 10.3389/ fphar.2018.01300 2. Huang L, Guo Z, Wang F, Fu L. KRAS mutation: from undruggable to druggable in cancer. Signal Transduct Target Ther. 2021;6(1):386. doi: 10.1038/s41392-021-00780-4 3. Whitfield JR, Soucek L. MYC in cancer: from undruggable target to clinical trials. Nat Rev Drug Discov. 2025. doi: 10.1038/s41573-025- 01143-2 4. Zhang Z, Rohweder PJ, Ongpipattanakul C, et al. A covalent inhibitor of KRAS(G12C) induces MHC class I presentation of haptenated peptide neoepitopes targetable by immunotherapy. Cancer Cell. 2022;40(9):1060-1069.e7. doi: 10.1016/j.ccell.2022.07.005 5. Pandey A, Rohweder PJ, Chan LM, et al. Therapeutic targeting and structural characterization of a sotorasib-modified KRAS G12C-MHC I complex demonstrate the antitumor efficacy of hapten-based strategies. Cancer Res. 2025;85(2):329-341. doi: 10.1158/0008-5472.CAN-24-2450 6. Petschnigg J, Groisman B, Kotlyar M, et al. The mammalianmembrane two-hybrid assay (MaMTH) for probing membraneprotein interactions in human cells. Nat Methods. 2014;11(5):585- 592. doi: 10.1038/nmeth.2895 7. Saraon P, Snider J, Kalaidzidis Y, et al. A drug discovery platform to identify compounds that inhibit EGFR triple mutants. Nat Chem Biol. 2020;16(5):577-586. doi: 10.1038/s41589-020-0484-2 8. Ghazi Vakili M, Gorgulla C, Snider J, et al. Quantum-computingenhanced algorithm unveils potential KRAS inhibitors. Nat Biotechnol. 2025. doi: 10.1038/s41587-024-02526-3 MEET THE INTERVIEWEES: Professor Charles Craik is a Principal Investigator at UCSF whose research focuses on defining the roles and the mechanisms of enzymes and other challenging proteins in complex biological processes and on developing technologies to facilitate these studies. Dr Igor Stagjlar is a group leader at the University of Toronto whose primary interests are in the identification and characterization of the interaction partners of membrane proteins associated with disease states, to obtain a better understanding of their molecular function, and in the identification of novel drugs and drug targets for use in therapeutic treatments. 34 DRUG DISCOVERY: ADVANCES AND INNOVATIONS mRNA Nanoparticles Offer New Hope for Female Infertility Treatment Izzy Hirst Impaired fertility is the inability to achieve pregnancy after 12 months or more of unprotected, regular intercourse. It impacts more than 1 in 10 women aged 15–49 in the United States, and in one-third of infertile couples, the cause is attributed to female factors. In Assisted Reproductive Technology (ART), eggs and embryos are manipulated to achieve fertility. Though ART does not treat the underlying cause, it can bypass certain roadblocks, for example, facilitating embryo placement within the uterus in an individual with fallopian tube defects. As a result, over 8 million babies have been born through ART since its inception in 1978. However, this option is neither suitable nor accessible for all couples. ART is associated with significant financial and emotional burdens, particularly as success is dependent on many factors, so couples cannot foresee the number of cycles required for conception. Furthermore, some couples cannot conceive via ART, as it does not address the root cause of infertility and will not be of benefit in cases that cannot be overcome with manipulation, such as a lack of viable eggs or inadequate uterine function. The latter, uterine factor infertility, is reported to account for up to 16.7% of cases in which couples face difficulty conceiving. Technology Networks spoke with Dr. Laura M. Ensign, principal investigator at Johns Hopkins Medicine, and Dr. Saed Abbasi, lead study author and research associate at Johns Hopkins University School of Medicine, regarding a lipid nanoparticle (LNP) mRNA delivery system they have developed to treat uterine and endometrial causes of infertility. Credit: iStock/Fabian Montano TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 35 Q: How did you become aware of the gap in treatment options for infertility, and what inspired you to carry out research in this area? Laura M. Ensign (LME): I became aware of the tremendous gap in treatment options for infertility through my research and entrepreneurial activities. While the process of ART has certainly evolved and become more effective, it was alarming to me that women often had to go through numerous rounds of painful, expensive, and stressful procedures to get pregnant. When Abbasi proposed targeting mRNA LNPs to the endometrium, I was very excited about the potential to improve the effectiveness of ART and quality of life for patients. Saed Abbasi (SA): Our new technology can be considered as an add-on to enhance ART and increase the likelihood of success by improving embryo attachment to the endometrium. While ART failures can stem from various problems, including egg and sperm quality, almost 50% of embryos transferred to the uterus fail to implant, highlighting the significant role of uterine and endometrial factors in ART failures. Q: In your research, you give examples of gynecological conditions, such as endometriosis, that may lead to infertility. Can you explain how they cause infertility and how mRNA nanoparticle technology aims to counteract this? SA: A key step in achieving a successful pregnancy is the implantation of a viable embryo in the endometrial lining of the uterus. Normally, the endometrium secretes a wide range of cytokines and hormones that thicken and prepare it for embryo implantation. These factors can be lacking or reduced when there is infection or inflammation. What causes an inflamed uterus? Examples of conditions that cause inflammation within the uterus include trauma, endometriosis, and structural conditions such as fibroids and adhesions, seen in Asherman’s syndrome, amongst others. Our technology aims to improve endometrial function by restoring secreted cytokines and hormones to normal levels using mRNA encoding for these factors. Q: By refining your approach, you successfully delivered mRNA to the endometrium, achieving therapeutic levels without systemic adverse effects. What informed the adjustments you made, and which changes were most important? SA: First, we used mRNA instead of the standard form of protein infusions (i.e., recombinant proteins). Estrous cycle The estrous cycle primarily occurs in nonprimate mammals, such as cows, dogs, and rodents. This is a recurring, hormone-driven, reproductive cycle in which the uterine lining is reabsorbed. Cycles may occur once a year, during specific seasons, or multiple times a year, depending on the species. Menstrual cycle The menstrual cycle primarily occurs in humans and primates, such as apes and monkeys. This is also a recurring, hormone-driven reproductive cycle, though the uterine lining is shed, leading to menstruation. Cycle duration ranges from approximately 24–45 days. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 36 mRNA is like a prodrug; it can only be converted into a therapeutic protein after it is taken up by the target cells. This will naturally reduce protein levels in body parts without cells, such as bodily fluids, making systemic drug distribution and adverse effects lower. Second, our LNP designs contain a surface signal that acts like a zip code to direct the packaged mRNA specifically to endometrial cells. Taken together, mRNA was specifically delivered to endometrial cells by LNPs, where the mRNA was converted to therapeutic protein only inside endometrial cells with much-reduced access to other tissues. Q: In your experiment, aiming to mimic fertility-reducing structural changes, you used an endometrial injury model in mice. How did you develop the model, and did you discover any limitations? LME: While it is often true that we are limited by the preclinical models and tools available to validate new therapeutics, this is particularly true in many areas of women’s health. However, we were fortunate to have a prior report of the ethanol-induced injury model in mice, which was a model of thin endometrium. Abbasi was very careful about validating the model, including characterizing the change in endometrial structure and the reduction of successful embryo implantation, while also ensuring that his LNPs were able to successfully deliver mRNA to the damaged endometrium. There are obviously differences in the structure of the uterus and the function of the endometrium in mice. However, the endometrium in both humans and mice utilizes cell surface proteins called integrins to act as a “dock” for the embryo to attach, which is why Abbasi engineered the LNPs to bind to integrins. Further, he was conscientious about scouring the literature to determine when integrin expression in humans was most closely mimicked in mice, and to design experiments with the highest likelihood of translation to human biology. Q: What considerations will be needed to translate this technology to human studies successfully? LME: The human and mouse endometria are very similar in structure, and both undergo hormone-dependent remodeling. Like humans, the mouse endometrium decidualizes to prepare for embryo implantation, and many of the cellular markers that are expressed are conserved between the two species. Thus, while the mouse estrous cycle is much shorter than the menstrual cycle and does not include menstruation, there are many structural and functional similarities, making the mouse a commonly used model for endometrial research. We will need to ensure that the targeting approach provides equal benefit in delivering therapeutic mRNA to the endometrium while minimizing off-target delivery in humans. Additionally, granulocyte-macrophage colony-stimulating factor (GM-CSF) may not be the ideal therapeutic protein in all cases, though the versatility of using an mRNA-based approach allows for selecting other protein targets for testing. mRNA technology is like changing pizza toppings when you need a new therapeutic protein, rather than remaking the entire dish from scratch as you would with recombinant proteins. TECHNOLOGYNETWORKS.COM DRUG DISCOVERY: ADVANCES AND INNOVATIONS 37 Q: What potential impact could this technology have in the future? SA: Now, we can change the mRNA sequence to provide instructions for making an endless list of therapeutic proteins. Although our study focused on delivering GM-CSF mRNA, some patients may benefit more from other types of cytokines, growth hormones, or a personalized combination. This is a huge advantage of using mRNA over recombinant proteins, in which each protein must be synthesized and extensively purified before it can be dosed in humans. mRNA technology is like changing pizza toppings when you need a new therapeutic protein, rather than remaking the entire dish from scratch as you would with recombinant proteins. Additionally, our technology can be explored to treat other endometrial disorders, not only infertility, but also endometrial cancers and painful conditions such as endometriosis. LME: I am particularly excited about the potential to treat a range of conditions and disorders that affect the endometrium, as well as the potential to engineer LNPs that can target other tissues in the female reproductive tract. There are many unmet needs in women’s health, and research and technology development often lag behind in these areas. MEET THE INTERVIEWEES: Dr. Saed Abbasi is a research faculty member at the Johns Hopkins University School of Medicine. His work focuses on drug delivery technologies for enabling targeted therapeutics and correcting genetic disorders. His research pioneered the use of mRNA for brain gene editing and contributed to the development of safer COVID-19 vaccines, as well as advanced targeted therapies for cancer, liver, and the endometrium. Before joining Johns Hopkins, he earned his PhD in pharmaceutical sciences from Hokkaido University, Japan. Dr. Laura M. Ensign is the Marcella E. Woll professor of ophthalmology and the vice chair for research in the Wilmer Eye Institute at the Johns Hopkins University School of Medicine. She has secondary appointments in chemical and biomolecular engineering, biomedical engineering, physiology, pharmacology and therapeutics, gynecology and obstetrics, infectious diseases, and oncology. Ensign’s research focuses on characterizing biological barriers in health and disease to design more efficacious formulations for prophylactic and therapeutic drug delivery, and has already led to the translation of pharmaceutical products that improve clinical management of human disease. DRUG DISCOVERY: ADVANCES AND INNOVATIONS 38 TECHNOLOGYNETWORKS.COM CONTRIBUTORS Bree Foster, PhD Bree Foster is a science writer at Drug Discovery News. She holds a PhD in comparative and functional genomics from the University of Liverpool and enjoys crafting compelling stories for science. Izzy Hirst Izzy graduated from the University of Hull with a bachelor's degree in biomedical sciences. She then completed a two-year master’s degree in physician associate studies at Hull York Medical School. Izzy joined Technology Networks in January 2026 and covers topics such as drug discovery and biopharma. Katie Brighton Katie joined Technology Networks in January 2022 as a scientific copywriter. She holds a master’s by research degree in molecular and cellular biology and a bachelor's degree in biochemistry from the University of Leeds. Kerry Day, PhD Kerry started her career in a physiology laboratory, obtaining a PhD in respiratory research before spending 10 years in academia. Now, motivated by the belief that science should be transparent and accessible, Kerry uses written communication skills acquired in her career to communicate science to a worldwide community. Kerry Taylor-Smith Kerry Taylor-Smith is a science writer who covers everything from climate change and the environment, to health and wellness, to material science, astronomy and space. She has a BSc in Natural Sciences from the University of Bath and over a decade of experience writing for various scientific and consumer publications. Joanna Owens, PhD Joanna Owens holds a PhD in molecular toxicology from the University of Surrey. She has over 20 years’ experience writing about a wide range of scientific topics in biosciences, pharmaceuticals and biotechnology. Laura Elizabeth Lansdowne Laura Lansdowne is the managing editor at Technology Networks, she holds a first-class honors degree in biology. Before her move into scientific publishing, Laura worked at the Wellcome Sanger Institute and GW Pharma.