Muneeb Sultan

Our paper on combing Docking with DELs is finally out in @JCIM_JCTC (and as one of the cover articles!) !

Enhanced Diffusion Sampling: We develop a framework for efficient rare event sampling and free energy calculation with diffusion models. We introduce Metadynamics and Umbrella Sampling for diffusion models. @MSFTResearch #MachineLearning #MD #Biology arxiv.org/abs/2602.16634

@zavaindar Agree on most things but disagree that the US should be ceding ground on follow-on molecules. Almost all things in the clinic have some issue that 2nd gen+ solve and giving up on them all together is leaving trillions on the table.

on china and biotech: 1) we regularly see founders and execs in our tech portfolio trade in salary for increased equity. across biotech in the US, we've seen for a decade+, the opposite behavior. convince a salaried person to work 9-9-6, the same hours those at the leading AI labs and those in China are doing... it's near impossible. culturally, biotech in china is faster and cheaper 2) ironically, a decades long trend of conservatism - consensus herding around the same small set of "validated" targets - may well be western biotech's achilles heels. as long as we're all crowding around the same targets (PD1, TL1A, GLP1, etc..) china will inevitably finds ways to develop the same medicine within the same well trodden modality (small molecule or antibody) hitting the same clinical goal post with a fraction of the time and money. the answer here isn't "western biotech is dead", but rather we've lost our way. we used to be inventors and cowboys. manifest destiny, so to speak. until we (the west) regain comfort and confidence again charting the frontiers of biology, chemistry, and science, then we'll be chasing the same commoditized known biology against cheaper and faster competitors this isn't localized to the private sector, but instead a corollary of self inflicted damage by cuts to NIH, attacks on academia, and the thwarting of world class immigrant talent that 20 years ago would have dreamt of being educated and working in the US, and now is often far more welcome in Europe and China. we're seeing it across our friends in academia and across the broader private sector. American exceptionalism is premised on being a global melting pot of the world's .001% smartest and most ambitious people. The moment we lose that we become just another nation state, just another economy 3) this is not the end of western biotech. the two sides can and will continue to learn from each other. increasing numbers of US biotech are ideating in the west, running research in China, and early clinical trials in China and Australia before coming back to America for larger registrational studies. China knows they have to ultimately cater to a western audience (see: BeOne for example). This means using the Chinese market as a stepping stone to a global market financially backstopped by the US and Europe 4) we cant rest on an assumption that China will fast follow and America will have a monopoly on invention. Did we learn nothing from deepseek last year? Have we not looked at the obvious performance, battery, design, and technology advantages of say BYD over Tesla? we have to lean into innovation and invention, knowing that any prior hegemony we may have had is now momentarily a jump ball, up for grabs 5) the fact this is described as a "spicy topic" where folks are wary to confront the issue head on and instead sharing their honest opinions only in hushed corners is itself part of the problem. without having honest, data driven conversations about what's happening the rhetoric turns very quickly to team-cheering dogma which is neither a strategy nor solace for global capital markets. i'm happy to take advantage of these market dislocations as an investor, though as a society we should seek to have open honest candid conversations that right the ship much much sooner than later

Things are finally looking more optimistic in biotech after a long time. Hoping sentiments continue to improve. #JPM2026

@DdelAlamo Fair, some of it is probably also that the pre training objective of PLMs, the reconstruction metric for IF models and the parametrized physics functional forms have little to do with the properties of ABs like aggregation. We need better/different pre training recipes for them.

@mmsltn The germline bias is only learned by PLMs AFAIK, not inverse folding models and physics based models. So that can't be the full story

I wonder how much of these results are due to the quirks of antibodies specifically, and how much is due to the reasons outlined in the Paper With The Greatest Graphical Abstract Of All Time

I do wonder if the fastest way for US biotech to get regulatory parity to China is for US regulators to require all international/ex-China rights to explicitly require Taiwan.😁

For folks in preclinical research, including many friends trying to raise, going through RIFs or trying to find new roles, I see you and wish you all the best navigating this multi year biotech bear market and the turbulence ahead. In the end, it will be okay.

Kudos to Jason for talking about this. It absolutely boggles my mind that we don't have more biotech execs/CEOs talking about this every week or lobbying Congress/State assemblies to jump in with legislative changes to FDA/MFN/IP/tax laws to make American R&D competitive.

I fear stories like this will become increasingly common and more grim. Competition is fine (and good) but currently American biotech is hamstrung by bloated CMC/GMP reqs for ph 1, high clinical dev costs, and no good mechanism to get FIH data quickly.

@owl_posting @leashbio But it’s not really a fair comp. At that point. Though I think how leash is approaching it at scale makes a lot of sense and hopefully things will improve with scale

@owl_posting @leashbio It’s an interesting question if it’s fair to ask boltz to generalize to DEL outputs. A non binder might be a non binder because of the dna label or a binder might be a stronger than expected binder because of the loss of entropy enforced by the linker. Both would hurt boltz perf

An ML drug discovery startup trying really, really hard to not cheat owlposting.com/p/an-ml-drug-d… on the 12-person, Utah-based startup @leashbio, their culture of rigor, and the many ways small molecule models accidentally learn the wrong thing

To 🇺🇸 with ❤️, 18 years, 2 degrees, multiple biotechs and a startup later, I finally got my citizenship last week. 🇺🇸 remains the only place where a kid from 🇵🇰 could get access to such incredible opportunities. Feeling thankful to everyone that helped me get here.

@_judewells @adaptyvbio

Ok let’s go @adaptyvbio binder design competition: this time designing proteins to neutralise the Nipah virus. Lots of great de novo ML binder design tools out there now, but this year I’m submitting an entry from TEAM HUMAN, seeing if pure rational design can win against the machines.

@owl_posting going from double digit to single digit nM is about 1/2 -1 kcal which could be from a single AA substitution. Doing 600 variants per epitope gives you a lot more coverage within a very constrained search space.

@owl_posting Thats mostly coming from using low throughout (chai) vs high throughput (JAM-2) for the testing. IIUC Jam is designing a single YSD library across targets with around 20 epitopes x 600 variants per target while chai is doing around 10 per epitope.

one final interesting bit of context is that while overall capabilities aren’t super dissimilar between the two models (both can create binders to a hard-to-drug class of receptors, most of which are developable), JAM-2 seems clearly better today this isnt something i naively appreciated before i looked at the affinities; JAM-2 binders are sub-nM/single-digit nM almost everywhere, Chai-2 binders are mostly double-digit/triple-digit nM’s