Arthur Deng

PhD student @ Stanford

[Cancer Grand Challenges @CancerGrand]

📜New in @ScienceMagazine: not one but two papers from team MATCHMAKERS!

The labs of David Baker @UWproteindesign and Chris Garcia @StanfordEMED published work that could revolutionise immunotherapy.

The Baker lab unveil a generative AI pipeline to design precise binders for antigen-MHC complexes—paving the way for more accessible immunotherapies.

🔗Read more: https://t.co/iOr6dpvpL1

[Arthur Deng @arthurdeng0205]

In wonderful collaboration with @karstenhouse_14 , @WUFang40615703, @SebastianThrun, K. Christopher Garcia, and @brianltrippe

[Arthur Deng @arthurdeng0205]

Predicting how mutations affect protein binding is key for drug design—but deep learning tools lag behind physics.

StaB-ddG closes the gap: combining stability models + smart pretraining to match FoldX accuracy at >1000× the speed.
Paper: https://t.co/HFJVw5nSKX
Code: https://t.co/Y3z4EAsjbH

@brianltrippe and @karstenhouse_14 will be presenting this paper at #ICML2025 next week in Vancouver
Contact: [email protected]

[Arthur Deng @arthurdeng0205]

Evaluations: past deep learning predictors claim “state-of-the-art” but we find they underperform a >20-year-old approach (FoldX). The issue:

1. leaky train/test splits
2. baselines without computationally expensive recommended settings

We fix these issues and find that StaB-ddG is the first deep learning ΔΔG predictor to match FoldX's accuracy while being >1,000× faster.