Today in Biology — Sep 13

[SOFIA] Alright, so big news this week straight out of the Department of Energy: they are really doubling down on open AI models for science. This is part of the Genesis Mission, which we've talked about before – it's this huge push to accelerate scientific discovery with AI.

[DANIEL] Hmm, and "open" in this context is pretty key, isn't it? It suggests a move away from proprietary, black-box systems towards something more collaborative and auditable, which is crucial for scientific reproducibility. But it also means navigating the complexities of data sharing and potential misuse.

[SOFIA] Exactly. And it’s not just talk; we're seeing actual awards. Lawrence Livermore National Lab is leading ten projects under this Genesis Mission, and the University of Washington is also heavily involved. For us in engineering biology, this is huge. Think about designing new enzymes, predicting protein folding for novel functions, or optimizing metabolic pathways in non-model organisms – all of that relies on massive datasets and complex simulations. If we have open AI tools that can chew through that data faster and more effectively, it changes the game for how quickly we can iterate and build.

[DANIEL] It certainly lowers the barrier to entry for groups without their own dedicated AI development teams. My question, though, is how "open" these models will truly be. Will the underlying architectures and training data be fully accessible, or just the inference engines? The utility for actual scientific advancement depends heavily on that level of transparency. Without insight into the training data, we risk perpetuating biases or generating biologically irrelevant predictions.

[SOFIA] That’s fair, Daniel. It’s definitely something to watch. But the sheer investment signals a real shift in how the DOE sees AI integrating into fundamental science. And speaking of shifts, the NSF is also overhauling its funding approach to align more with White House priorities. We don't have specifics yet on how new initiatives will be funded, but staff are apparently concerned about a more streamlined approach.

[DANIEL] "Streamlined" often translates to prioritizing certain areas, which can leave others underfunded. It’s always a balance for agencies like the NSF: supporting truly foundational, curiosity-driven research versus directing funds toward national strategic goals. For engineering biology, this could mean more emphasis on applications tied to energy, defense, or health, potentially at the expense of more speculative, long-shot basic science that might not have an immediate clear application.

[SOFIA] Right, it’s a push and pull. And then there's ARPA-H, which just launched what they're calling the world's first bid to build FDA-authorized clinical AI for cardiovascular care. That’s a very specific, very high-stakes application.

[DANIEL] That’s a significant move, given the regulatory hurdles for any medical AI. Getting FDA authorization for clinical AI is a huge validation step, and it points to a future where AI isn't just a research tool but a direct part of patient care. The rigor required for that kind of approval will undoubtedly influence how these models are developed and validated. I’d be very interested in seeing their proposed validation protocols and how they address issues like algorithmic bias in diverse patient populations.

[SOFIA] Absolutely. It’s the kind of thing that could really accelerate the translation of AI models into actual clinical tools, which would then feed back into how we think about engineering biological systems for diagnostics and therapeutics. And then, there’s this interesting bit of news that came out last week: the NIH has reached an agreement to use part of its budget to pay for Department of Defense research.

[DANIEL] That’s a notable collaboration. It suggests an increasing overlap in research priorities, perhaps in areas like biodefense, battlefield medicine, or even technologies with dual-use potential. But it also raises questions about mission creep for the NIH, whose primary focus is typically civilian health. We’ll need to see how they delineate those funding lines.

[SOFIA] Yeah, that’s a new one. It blurs some traditional lines, doesn't it? It’ll be interesting to see how that collaboration actually manifests in terms of specific projects and how it impacts the broader funding landscape for biological research. A lot of these shifts seem to be pointing towards more targeted, mission-driven science, often with AI at the core.