We've updated our Privacy Policy to make it clearer how we use your personal data. We use cookies to provide you with a better experience. You can read our Cookie Policy here.

Advertisement

How Is AI Shaping the Future of mRNA Therapeutics?

3D illustration of a purple mRNA strand on a purple and blue background.
Credit: iStock.
Listen with
Speechify
0:00
Register for free to listen to this article
Thank you. Listen to this article using the player above.

Want to listen to this article for FREE?

Complete the form below to unlock access to ALL audio articles.

Read time: 4 minutes

Interest in mRNA therapeutics has surged in recent years, driven by their ability to target a wide range of diseases and their potential for personalized medicine. As more of these therapies move through clinical trials, the pressure to streamline design, testing and manufacturing processes has intensified.


Advanced artificial intelligence (AI) technologies are playing a pivotal role in extracting insights from complex biological datasets, accelerating discovery cycles and optimizing sequence design. But as AI capabilities expand, what new possibilities lie ahead?


To learn more about how AI is advancing the mRNA drug discovery pipeline and what we can expect to see in the future, Technology Networks recently spoke with Dr. Davide De Lucrezia, vice president and general manager at Officinae Bio, part of Maravai LifeSciences. In this interview, De Lucrezia also discusses how Officinae Bio’s AI-powered platform is addressing current challenges and mitigating drug development bottlenecks.

Anna MacDonald (AM):

AI is breaking down boundaries of what’s possible across drug discovery, development and manufacturing. From your perspective, what are a few novel, useful advancements of AI and machine learning in drug discovery? 


Davide De Lucrezia, PhD (DDL):

When I began my career in the early 2000s, the field of nucleic acid therapy was beginning to incorporate high-throughput sequencing and analytics. Labs using these technologies were generating tons of data, but the data didn't provide actionable insights. This dilemma sparked my interest – ever since, I have wanted to understand how we can identify hidden patterns behind large datasets to increase their utility in research.


Now, those in the drug discovery phase have introduced AI and machine learning to extract more usable information from data. Multiomics datasets are hard to navigate as humans, so many scientists turn to machine learning to extract patterns and help inform and assess drug design. In past decades, machine learning was mainly used to optimize processes. AI and machine learning are no longer only optimizing what exists but advancing sequence design and supporting incredible acceleration of the discovery cycle.



AM:
What are some common challenges developers encounter in the DNA and RNA drug discovery phase?

DDL:
One of the challenges faced by developers is optimizing the entirety of a sequence for an application. A therapy isn’t just based on the gene of interest – considering the entire sequence can dramatically and holistically improve a therapeutic.

If we were to compare a DNA or RNA sequence to a symphony orchestra, much of the attention has traditionally gone to the first-chair violinist – the gene of interest. However, for peak performance, one must compose for the entire orchestra – untranslated regions, chemical modifications, lipid nanoparticles, etc.

Today, scientists are shifting focus to the full sequence to improve therapeutic efficacy. For mRNA-based therapeutics, we are focused on non-coding regions that enable tissue-specific targeting or modifications that sustain prolonged expression in vivo. We need a comprehensive view of the system for true optimization. This is where machine learning technology is truly valuable – it helps scientists handle the complexity of a system-level approach to drug discovery and development.


AM:
Do challenges in drug development vary depending on geographic location or national regulatory agencies?

DDL:

Challenges remain mostly the same, but one element that does vary is the relationship between data and privacy in different countries. One may naturally think that if they could include real patient data in the machine learning approach, they could design more effective therapeutics both at a population level and a personalized level. However, across different regions, countries and cultures, we see a wide range of perceptions about the use of patient data in therapeutic development.



AM:
Can you share how you are developing and incorporating AI-enabled RNA and DNA design platforms to address those challenges?

DDL:

In the discovery phase, timelines are condensed, and there's a lot of pressure to move quickly. When engineering a system with hundreds of thousands of different possibilities, it is key to invest the available time and resources where it's going to count the most.


For those in mRNA therapeutics, getting from sequence design to testable mRNA is often the slowest step – creating a bottleneck in the design/build/test cycle.


We address this with an integrated, machine learning-driven platform that accelerates and improves RNA design and synthesis with two proprietary algorithms. The first helps scientists design optimal sequences by predicting features that enhance translation, durability and tissue specificity. Once the design is selected, our second algorithm streamlines the build phase, accelerating synthesis workflows.


While most algorithms are prescriptive, our platform uses active learning – a feedback loop that continuously refines its recommendations based on experimental data. This means our models get smarter over time, better understanding sequence-function relationships and enabling scientists to make faster, more informed design choices.



AM:
From your perspective, how can the industry address the issue of access to AI-enabled RNA and DNA design platforms?

DDL:
Throughout my career, I always believed that biology was programmable. Current industry efforts to engineer biology are fully tailored to the specific company’s needs – they’re irreproducible, hard to scale and almost impossible to transfer to the rest of the community. It’s been a goal of mine to really develop a toolbox to make biology programmable in a reliable manner – making ingenuity accessible and affordable for all. 

Digital transformation and eventual integration of AI into nucleic acid design and development are key steps toward addressing these challenges, but organizations and leaders in the field first need to align on AI’s role and importance. 
Going back to accessibility, for an AI-based technology to be truly accessible and support the drug discovery timeline and projects overall, it must be easily integrated into the team’s existing technology. Requiring special equipment or staff to support the AI tech’s operations provides cost and logistical barriers for small, early-stage biotech companies.


AM:
Are there any risks involved with AI-enabled sequence optimization in comparison to traditional sequencing technologies? 

DDL:
There is potential to “overfeed” AI models with just one set of data. If companies use small datasets or large datasets from highly homogeneous experiments, the model risks either underperforming due to limited data or becoming overfitted to a narrow set of conditions. The effectiveness of AI models depends on the quality, quantity and diversity of the data they are trained on. However, by sharing data in a feedback loop, companies can help continuously improve the model's performance. 

Instead of being conservative or defensive, I believe industry members can and should focus on building transparent relationships with clients and ensuring that data is properly anonymized to protect their intellectual property and interests. This approach benefits everyone, as even small biotech companies can leverage the data network effectively in their drug discovery and development.



AM:
As AI tools become more integrated into nucleic acid drug discovery, do you foresee a shift in the skillsets needed for scientists in the field?

DDL:
To be clear, I do not think AI will replace the ingenuity of researchers, but it can help researchers to dramatically accelerate the discovery cycle time. The discovery cycle phase is really based on trial and error, and because of that, it is inherently slow.
 
AI and hardware automation allow research teams to re-channel human ingenuity toward more complex, high-value activities rather than tedious yet necessary tasks in drug discovery (cloning, data collection, etc.). Within the nucleic acid therapy space, more biologists can now focus on the “design” part of the cycle rather than the “build” phase.

With this in mind, the widespread inclusion of AI in drug discovery will certainly require a shift in how we train lab staff to use newer technologies. As researchers move to other valuable tasks, they will also need to be able to retrieve and analyze the data gathered through AI-based and automated tools.


AM:
Looking ahead 5–10 years, how do you envision AI shaping the future of personalized mRNA-based medicines?

DDL:
I expect AI optimization to play a large role in increasing therapeutic safety. In gene therapy, a key goal is to localize action to the right tissue at the right time. Available AI technologies allow scientists to better identify tissue-specific targets and design therapeutic candidates accordingly, ensuring efficacy while minimizing off-target effects. 

If AI technology continues to increase tissue specificity and timing, we can potentially create more potent drugs with fewer side effects in the next decade or so. It’s my hope that AI can help bring more effective and accessible therapeutics to patients faster, which is what this line of work is all about.