Today in Biology — Oct 16

[SOFIA] Welcome back to The Dish, everyone! So, things have been *buzzing* in the policy and funding world this week, and we've got some big news shaping how biology gets done, especially here in the US. The Department of Energy has just unveiled the first wave of awards for its new Genesis Mission.

[DANIEL] Hmm, Genesis Mission. That sounds suitably… foundational. What exactly is the DOE aiming for with this?

[SOFIA] Exactly! It's all about pushing what they call "super intelligence for science." We're talking massive investment in AI to accelerate scientific discovery, particularly in areas relevant to the DOE’s mission, which often means energy, materials science, and yes, also biology. They've just announced twelve new awardees, plus a new graduate fellowship program to train the next generation in this AI-science intersection.

[DANIEL] So, AI for science isn't new, but "super intelligence" suggests a step change. Are we talking about more compute power, novel algorithms, or a shift in how problems are framed? Because often, these initiatives are heavy on the promise and lighter on the truly disruptive methodological shift.

[SOFIA] That’s fair, Daniel. From what they’ve announced, it's really about integrating AI across the entire scientific workflow, from hypothesis generation to experimental design, data analysis, and even the synthesis of new materials or molecules. For engineering biology, this is huge. Imagine AI designing novel enzymes for biofuel production or optimizing microbial pathways for biomanufacturing — not just crunching data after the fact, but actively driving the discovery process. It’s about making the entire pipeline more intelligent and autonomous.

[DANIEL] That’s the dream, certainly. The challenge will be in validating those AI-generated hypotheses and designs. We've seen generative models produce fascinating outputs, but moving from a computationally plausible molecule to a biologically functional one with the desired properties often reveals the gaps in our predictive models. What kind of experimental validation frameworks are built into these Genesis awards?

[SOFIA] That's a great question, and the details on the experimental validation component for these specific awards aren't fully outlined in the initial announcement. But the overall push is certainly towards closing that loop. And speaking of designing novel biomolecules, *Science* magazine was just highlighting some discussion around RFdiffusion3. That's a new generative AI system that’s apparently designing proteins and nucleic acids atom by atom. The idea is to engineer biomolecules with really precise, complex functions.

[DANIEL] Atom by atom – that’s ambitious. Protein design has made strides, but achieving *de novo* function with atomic precision, especially for complex tasks, remains incredibly difficult. The proof will be in the pudding: how many of these computationally designed molecules actually fold as predicted, and then, how many perform the intended function *in vitro* and *in vivo*? Without robust experimental validation data attached to these claims, it’s still in the realm of impressive computational demonstrations rather than proven engineering tools.

[SOFIA] Agreed, the proof is always in the biological pudding. But the excitement is palpable, especially for us in engineering biology. If these tools mature, it completely changes how we approach designing new biological systems. And it’s not just the DOE pushing this. The NSF just laid out its vision for the "next generation of technologies" and support for American talent, tying into the White House's "Science: A New Golden Age" report. It’s a coordinated push across multiple agencies.

[DANIEL] A coordinated push is good, especially if it translates into sustained, predictable funding for fundamental research. But sometimes these grand pronouncements can lead to a scattering of resources rather than focused impact. What are the concrete programs the NSF is launching here?

[SOFIA] They're setting up a series of programs and initiatives to support that vision, though the specifics of each individual program are still emerging. It sounds like they're building out infrastructure and talent development alongside the direct research funding. And then, there’s ARPA-H, which just kicked off its BIOGAMI program.

[DANIEL] ARPA-H is always interesting. BIOGAMI, what's that one targeting?

[SOFIA] BIOGAMI is all about finding and fixing protein dysfunction at the earliest stages. So, think about diseases where misfolded or dysfunctional proteins are key players. This program wants to develop technologies to detect those problems super early and intervene. From an engineering perspective, that means developing new diagnostics and therapeutic modalities that can sense and correct subtle molecular errors. It’s a very targeted, problem-driven approach, which is classic ARPA-H.

[DANIEL] Early detection and intervention for protein dysfunction is a noble goal. The challenge there is often the signal-to-noise ratio in early disease states, and the specificity of any intervention to correct only the problematic proteins without off-target effects. It’s a high-risk, high-reward area, which is precisely what ARPA-H is designed for.

[SOFIA] Absolutely. So, we're seeing this strong, multi-agency current pushing towards AI-driven discovery, next-gen tech development, and very targeted biological interventions. It feels like a moment where the tools we're building are starting to intersect with some really ambitious national goals.

[DANIEL] It's certainly a lot of momentum. The key, as always, will be rigorous execution and ensuring that the scientific ambition is matched by equally robust experimental validation and a clear-eyed assessment of what these new tools can *actually* deliver.

[SOFIA] Always the voice of reason, Daniel! And on that note, we’ll be right back after the break.