Phil Leung

My doctoral advisor @UWMolES / @UWproteindesign (and current scientific advisory board membe @Xaira_Thera) David Baker won the nobel prize!! So stoked and extremely determined to deliver on the promise of this tech! nobelprize.org/prizes/chemist…

Glad to see our antibody design paper finally out in @Nature (and congrats to lead authors and everyone involved )! At @Xaira_Thera we are excited about pushing antibody design further to bind harder targets and make drugs for unmet medical needs. Paper: nature.com/articles/s4158…

Congrats @adam_broerman @FloPraetorius @UWproteindesign and team!

Showcase of #proteindesign possibilities+why they matter! - minibinders (rigid fusions; composability with other denovo designs) - tunable biophysical properties (modulate k_off, flexibility, strain) - #crispyshifty hinges (allosteric control of binding kinetics to native target)

We searched for Bigfoot using CRISPR. DNA storage meets CRISPR search: AI maps the data, Cas9 slices it, nanopores read it. Multiplexed molecular search with semantic retrieval across a database of DNA-encoded images. Off-targets = a feature, not a bug. tinyurl.com/mr34c7fh

Thrilled to share our latest publication in Nature: nature.com/articles/s4158…. This work reflects a fantastic collaboration—special thanks to David Baker and @TimothyPJenkins. Hoping it sparks attention to this neglected health issue and drives solutions in the years ahead!

Bonus Flo reunion pic! @FloPraetorius

What an immense privilege to celebrate the #NobelPrize with David, former colleagues from @UWproteindesign , current colleagues at @Xaira_Thera and @RosettaCommons family!

So cool to see #CrispyShifty hinges applied this way! Congrats @adam_broerman

Excited to announce our new #proteindesign strategy for allosterically controlling the kinetics of protein-protein interactions! Read on for cool applications in cytokine signaling, biosensing, and protein circuits. biorxiv.org/content/10.110…

I guess errors are expected. But it does highlight the need for journals to require depositing source code before publication.

⚛️ I got a new job ⚛️ After 5 wonderful years at @UWproteindesign and @UWMolES, I’m headed to Palo Alto for a postdoc at @StanfordUChem under Todd Martinez. SUPER excited to dive into deep learning for quantum chemistry and explore the Bay Area!!

Nobel Laureate David Baker on the phone with fellow awardees Demis Hassabis and John Jumper. The trio discussed how their teams have inspired one another and what the future may hold.

#CrispyShifty proteins mentioned - highlight of the announcement for me 😁 @ArvindSPillai1 @FloPraetorius

My doctoral advisor @UWMolES / @UWproteindesign (and current scientific advisory board membe @Xaira_Thera) David Baker won the nobel prize!! So stoked and extremely determined to deliver on the promise of this tech! nobelprize.org/prizes/chemist…

Congrats!!

@pranamanam Sure, structure prediction for IDRs is not well posed. That's not the original point you were making though 😅

@definitelyphil I guess for two state proteins, sure! I’m just unsure about any protein with IDRs

I'm going to keep saying this until someone proves me otherwise: AlphaFold(2, 3, 10000, whatever) or any structure prediction tool will not generalize to protein conformations if its training is optimized to "ground truth" structures in the PDB! Crystallized or cryo-EM structures are NOT ground truth: they are artificially-constrained, single conformations in a physiologically irrelevant environment. 👎 Perturbing these structures via MD or diffusion won't magically generate possible alternate conformations either. Those resulting poses are also artificial and likely wrong. This amazing paper, I believe, supports my point. This is also my gripe about training language models with structural tokens -- those are biased pieces of knowledge that inhibit design and modeling of conformationally and functionally diverse proteins operating in complex cellular environments. 🙄 However, I have to give credit to the ESM3 team for balancing loss on sequence, structure, and function during training. If you want to work on static proteins, go ahead, use these tools, use structure -- you'll probably get some working designs or semi-trustworthy models. 🤷‍♂️But we need a more first-principle, unbiased approach to modeling conformationally dynamic structures. I know a lot of people disagree with me on this. I like pLMs because at least we're not making assumptions (MLM is as basic as you get), and some signal should be trustworthy if evolution is indeed meaningful, but I agree that we need better training paradigms that correct evolutionary inconsistencies (can't be using ESM-2 forever). We'll keep pushing on this for out-of-distribution but important target classes. 😬 Alright, that's my rant for the week. We have some exciting works coming out soon though, so hope you'll stay tuned to our work. 🤗 Paper: nature.com/articles/s4146… (Beautiful) Benchmarking Code/Results: github.com/ncbi/AF2_bench…

@pranamanam FRET, DEER, EM and x-ray crystallography of the two states: science.org/doi/full/10.11…

@definitelyphil Phil, how do you know this? Like do you know that the change it's predicting is real? I think it's the question of what is ground-truth again, and whether the conformational states we are observing is just random sampling of an overfit latent space.

@DdelAlamo @srivsesh @anthonygitter @KevinKaichuang @pranamanam @DaveJuergens @sokrypton @avapamini @json_yim @amyxlu @rohitsingh8080 @proteinrosh @JustasDauparas @thisismadani @HannesStaerk Did anyone grab the original tweet? I can't see it and I'm dying of curiosity

@srivsesh @anthonygitter @KevinKaichuang @pranamanam @DaveJuergens @sokrypton @avapamini @json_yim @amyxlu @rohitsingh8080 @proteinrosh @JustasDauparas @definitelyphil @thisismadani @HannesStaerk *or rather, if all the inputs (structure, masking, etc) are identical. I can't think of a different explanation off the top of my head

Love this article by @schubertcm @geekwire on what we're trying to do here in #Seattle at @Xaira_Thera @JustasDauparas @naterbennett0 geekwire.com/2024/inside-th…