Perturbation Proteomics Is Revealing How Drugs Really Work
Perturbation proteomics captures protein responses to drugs and other cellular interventions.
Understanding how drugs interact with biological systems remains one of the biggest challenges in drug discovery and precision medicine.
While genomic and transcriptomic technologies have transformed our ability to study disease, they do not always explain what happens inside cells after a therapeutic intervention.
Proteins often tell a different, and more complete, story.
This is where perturbation proteomics is making an impact. By capturing how proteins respond to a defined perturbation, researchers can gain deeper insights into drug mechanisms, identify biomarkers of response, and uncover new therapeutic opportunities.
Among the researchers helping to push this field forward are Prof. Ruedi Aebersold, a professor emeritus of molecular systems biology at ETH Zürich who works on quantitative and systems proteomics, and Dr. Tiannan Guo, the director of the Westlake Center for Intelligent Proteomics, whose work focuses on perturbation proteomics, AI, and large-scale proteomic modeling.
Despite advances in drug discovery, important questions remain about how therapies interact with complex biological systems.
“Today we generally lack a deep understanding of the biology of the drug target and the relationship between the target and the clinical condition,” explained Aebersold.
Recent progress in mass spectrometry (MS), data analysis, and AI are helping to address this, enabling researchers to map cellular responses to perturbations at scale and depth.
Technology Networks spoke with Aebersold and Guo about the rise of perturbation proteomics, its role in understanding drug responses and mechanisms of action, and how new technologies could help influence the future of precision medicine.
The foundations of perturbation proteomics
The concept behind perturbation proteomics is straightforward: apply a defined perturbation and measure how the population of proteins responds.
These perturbations can take many forms, including drug treatments, CRISPR-mediated gene edits, RNA interference experiments, nutrient changes, environmental stressors, or natural genetic variability within a population.
Although transcriptomic technologies can reveal which genes are being expressed, gene expression does not always predict protein abundance or activity; focusing on proteins, therefore, provides a more comprehensive view of what is actually going on in the cell vs what could be. Proteins are also subjected to a range of processes after they are produced.
“Foundational principles in molecular biology taught us that biological functions are catalyzed and controlled predominantly by proteins,” said Aebersold. “Intense research has also shown that the relationship between transcript profiles and protein abundance profiles is complex and poorly predictive, especially if one intends to predict protein abundance profiles from transcript profiles following perturbations.”
“Fortunately, proteomic technologies have advanced to a stage where such measurements are now feasible,” Aebersold added.
High-throughput MS approaches, including data-independent acquisition, tandem mass tag workflows, and sequential window acquisition of all theoretical mass spectra, allow researchers to quantify thousands of proteins across hundreds or thousands of samples. At the same time, advances in automation are improving throughput, making large-scale perturbation studies possible.
However, researchers are no longer limited to measuring protein abundance alone.
“The state of a protein manifests by different modalities (location, interactions, post-translational modifications (PTMs), structure, etc.); there is no single technique that, by itself, provides the totality of the desired information. However, there now exists a palette of techniques, each of which is capable of measuring a different indicator of protein state,” said Aebersold.
These developments build on a series of advances in quantitative and systems proteomics over the last two decades and are now being applied to perturbation-based drug studies.
“To extract the functionally most relevant information from perturbation experiments, proteins are clearly the mechanistically most informative class of molecules.” — Prof. Ruedi Aebersold.
Linking proteomic signatures to drug responses
One of the most compelling applications of perturbation proteomics is in drug discovery.
Protein Research Workflows: From Expression to Analysis
This guide provides practical guidance on protein expression, sample preparation, purification, electrophoresis, western blotting, immunoassays, mass spectrometry, and cryo-EM.View How To Guide
When a drug enters a cell, it triggers a cascade of molecular events, and measuring these changes can help researchers determine how a drug works, identify unintended effects, and understand why some patients respond differently from others.
“Drug perturbation-based proteomic datasets are uniquely valuable because they capture the direct molecular response of cells following drug treatment,” said Guo.
“Perturbation-based approaches promise to increase the knowledge of both drug targets and their relationship to the clinical condition, because they provide insights into how the system adapts as a whole, at a molecular level, to a wide range of perturbations,” Aebersold added.
The ability to observe cellular adaptation in real time offers an advantage over more indirect approaches.
“Proteomic profiling after such perturbations provides the most relevant and direct readout of these effects,” said Guo. “Traditional transcriptomic or genetic perturbation datasets do not directly measure the specific cellular changes induced by a given drug.”
These datasets are proving valuable for mechanism-of-action (MoA) discovery. By comparing protein signatures generated by different compounds, researchers can infer whether a drug is acting on its intended target, whether off-target effects are occurring, and whether compounds with similar signatures may share biological mechanisms.
The approach has been strengthened by studies generating large-scale perturbation datasets. Researchers have begun creating dose- and time-resolved proteomic maps that capture how proteins and PTMs change across multiple drug concentrations and exposure times.
Aebersold believes that this level of mechanistic understanding can improve drug discovery efforts: “It seems obvious that in comparison to screening efforts that lack biological context, detailed mechanistic understanding of the targeted system in the state before and after treatment will increase the likelihood of identifying suitable targets and lead compounds,” he said.
The growth of these datasets is also fueling AI development.
Drug Repurposing Unlocks Novel Therapeutic Approaches
This listicle highlights real-world success stories in drug repurposing and emerging strategies involving genomics, clinical data, network biology, and AI to identify new uses for existing medicines.View Listicle
“I think the most important recent development is high-throughput proteomics with deep coverage,” said Guo. “The speed of proteome profiling has greatly increased, allowing us to capture proteomic states under 1000s of different perturbations. This is extremely important for generating sufficient data for building deep learning models to understand how a drug works.”
“The relationship between AI and proteomics is multifaceted, and these advances will allow a more complete understanding of proteomic dynamics in biological samples,” explained Guo.
Researchers are also integrating proteomic signatures with phenotypic measurements such as cell viability, proliferation, or treatment sensitivity. The result is a more comprehensive picture of how molecular changes translate into biological outcomes.
“Without drug-specific perturbation data, it is impossible to accurately predict drug sensitivity or mechanism of action.” — Dr. Tiannan Guo.
Applications of perturbation proteomics in precision medicine and drug discovery
As perturbation proteomics datasets expand, researchers are beginning to explore applications beyond MoA studies.
One promising area is biomarker discovery. Patterns of protein expression and PTMs could help identify which patients are most likely to respond to a therapy, offering a route towards more personalized treatment strategies.
Another opportunity lies in drug repurposing. If two compounds generate similar perturbation signatures, they may influence related biological pathways even if they were developed for different indications, helping researchers identify unexpected therapeutic uses for existing medicines and bypassing years of early development.
The field is also moving towards integrated approaches that combine proteomics with transcriptomics, metabolomics, phosphoproteomics, spatial biology, and single-cell technologies. Together, these datasets could provide a systems-level view of drug action.
“It is likely that in the next 5–10 years the generation of synergistic multi-proteomic data sets from the same set of perturbed samples will accelerate,” said Aebersold.
However, significant challenges remain. Data generation is becoming faster, but producing large, standardized datasets remains costly.
Transferring Established HPLC Methods in Pharmaceutical QC
This case study shows how a pharmaceutical manufacturing site evaluated the transfer of established HPLC methods, with equivalent results obtained across three routinely tested products.View App Note / Case Study
“I see major challenges in establishing at public sector or private sector research centers the capability to systematically generate multimodal proteomic data from aliquots of large numbers (hundreds to thousands) of perturbed and phenotyped samples and to establish at the same time the computational/AI capabilities to integrate the data into predictive models. As most of the elements of such a platform exist at present in distributed forms at different institutions, meeting the challenge is more a financial and organizational issue than a scientific or technical one,” explained Aebersold.
“The analysis of the ensuing data and their integration into a multimodal model that represents the mechanistic state of the system studied will require AI,” Aebersold added.
Guo similarly highlighted current limitations, noting that “the main challenges are still limited throughput and the immaturity of current methods, which result in relatively high costs. In addition, the available data volume is insufficient to fully support large-scale analyses.”
Despite these challenges, researchers see considerable potential in combining perturbation proteomics and AI. Guo recently co-authored research exploring how AI could support applications ranging from protein identification to the creation of “virtual cells” that model biological responses computationally.
Mapping the future of drug discovery
Perturbation proteomics is rapidly becoming a foundation of drug response research. By revealing how proteins change following therapeutic intervention, the approach provides a mechanistic view of biology that is often inaccessible through transcriptomics alone.
The convergence of high-throughput proteomics, multiomics integration, spatial biology, and AI is now creating opportunities to build comprehensive maps of cellular response. This could help researchers predict drug efficacy, understand resistance mechanisms, identify biomarkers, and accelerate the development of new therapies.
“Progress in this area will be exciting to follow because it will likely transform experimental medicine,” said Aebersold.
“In the field of proteomics, the next 10 years could see transformative changes if we are able to establish a proteome foundation model,” Guo added. “Similar to how AlphaFold revolutionized protein structure prediction, a foundation model for proteomics could predict the spatial and temporal dynamics of the proteome. Such a model would fundamentally change the way we understand, predict, and personalize drug responses.”
If that vision becomes reality, perturbation proteomics could help researchers answer the question: Which therapies are most likely to work, for which patients, and why?
Add Technology Networks as a preferred Google source to see more of our trusted coverage.
