Project Mutavault evaluates biological foundation models for DNA synthesis screening, exploring how their ability to predict protein structure, function, and interactions can harden screening against novel engineered and AI-designed biological threats.
About the project
This research project addresses a growing biosecurity risk created by advances in biological AI foundation models, whose ability to generate novel DNA and protein sequences will increasingly challenge key safeguards like DNA synthesis screening (Baker & Church, 2024; Hunter 2024). Building on pioneering work that demonstrated AI-enabled vulnerabilities in current screening methods (Wittmann et al. 2025), this project conducts an in-depth evaluation of state-of-the-art biological model capabilities for both (I) generating low-homology protein variants that could maintain function while evading sequence-based screening, and (II) accurately predicting biomolecular properties like structure, function, and binding from DNA or protein sequences, including for low-homology variants. The work also utilizes non-hazardous proteins as safe proxies in wet-lab experiments to characterize AI-generated protein stability, functional activity, and other relevant properties (Ikonomova et al. 2025), providing ground truth evidence towards understanding model capabilities across a range of biomolecular functions that are relevant to biosecurity. If successful, this project will provide a foundation for developing function-based screening methods that leverage powerful biological AI models to reinforce DNA synthesis screening and related safeguards against novel engineered and AI-designed biological threats.
References:
Baker, D, & Church, G (2024). Protein design meets biosecurity. Science, 383(6681), 349-349. https://doi.org/10.1126/science.ado1671
Hunter, P. (2024). Security challenges by AI-assisted protein design. EMBO reports, 25(5), 2168-2171. https://doi.org/10.1038/s44319-024-00124-7
Ikonomova, SP, Wittmann, BJ, Piorino, F, Ross, DJ, et al. (2025). Experimental Evaluation of AI-Driven Protein Design Risks Using Safe Biological Proxies. bioRxiv. https://doi.org/10.1101/2025.05.15.654077
Wittmann, B. J., Alexanian, T., Bartling, C., Beal, J., Clore, A., Diggans, J., ... & Horvitz, E., et al. (2025). Strengthening nucleic acid biosecurity screening against generative protein design tools. Science, 390(6768), 82-87. https://doi.org/10.1126/science.adu8578
Theory of change
Powerful foundation models trained on large biological datasets continue to scale in generative capabilities, and increasingly pose a risk to established biosecurity safeguards like DNA synthesis screening. At the same time, frontier language models may provide uplift that lowers the skill barrier for malicious actors to engineer new pathogens. Securing biotechnology against misuse in this context requires safeguards that leverage powerful AI capabilities for defensive purposes. This work lays the foundation for harnessing bio foundation models to improve the accuracy and sensitivity of sequence screening methods—particularly for challenging genes and proteins whose sequences may not resemble anything found in nature—providing a robust barrier to detect and deter misuse.
Your role
Under direction and guidance from the PI, mentees will pursue focused technical research in support of the project, and within a defined scope and objectives that are tailored to the individual’s background, interests, and level of time commitment. See the attached proposal for more details.
Prerequisites
Qualified mentees will have the following:
-
A background in bioinformatics, computational biology, structural biology, biochemistry, biophysics, protein engineering, biosecurity, AIxBio, AI/ML engineering, or a related field.
-
Prior technical research experience.
-
A good understanding of the basics of biomolecular sequence, structure, and function.
-
Some experience with biological AI models.
-
Proficiency with Python.
-
Strong critical thinking and creative problem-solving abilities.
-
Curiosity and a desire to understand the world.
-
The integrity and judgment to responsibly carry out sensitive research.
Time commitment
8+ hrs/wk
Application question(s)
Scenario: You’re the biosecurity officer in charge of screening orders at a small DNA synthesis company. While reviewing an order you encounter the following mystery sequence that didn’t return any close matches using the company’s internal sequence screening tool:
ATGGTTCCTGGTCATATTGTTTCTGAAGCTAATGATCAACAACAAGAACAAGATGTTGGTGATGTTGTTTGTTTTGATATTTCTAGAGGTGCTGTTATTAGATGGAATGGTTCTCCTCCTGAAACTGCTGTTGGTATTGTTGCTTTGGATCATGGTCCTGCTCAAGATTTTGTTTCTGTTTATATTGGTGAACCTGTTAAATTGGATGAAGTTAAATGGGCTCAAGTTCCTGGTCCTGGTGCTTCTGAACCTGATTGTTTTTCTCATTTGAAAATTAGATTGGTT
Part 1 (~10 min): Briefly describe the steps you would take to investigate the mystery sequence, with the goal of determining what it codes for and whether it should be flagged as a biosecurity threat. Include a short description of what information you expect to gain from each step, and how you would use that information in your investigation. A concise list is fine (i.e. 100-200 words total).
Part 2 (~50 min): Write a short python script that implements one or more steps from your answer in Part 1 using publicly available tools and libraries, and taking the sequence as an input. Share the script, the input(s), the outputs you got from the mystery sequence, and what (if anything) you can conclude from the outputs. This is meant to be a quick demo, so pick one or a few steps and focus on getting them to return a result, rather than implementing the entire process in detail. (And please don’t share any credentials in the code or inputs).
Note: It’s okay to use AI in completing this exercise, but if you do, briefly describe which models you used and how you used them.
About the mentor
Gary Abel is a scientist with a background in biochemistry, molecular biophysics, and DNA sequencing. He spent 15 years in biotech R&D before shifting last year to focus on biosecurity and addressing AI-bio risks. Gary is founder and Principal Scientist at Fourth Eon Bio, an independent research organization exploring how bio foundation models can improve DNA synthesis screening and other biological safeguards.