Tong Chen

AI for Drug Discovery | PhD @ UPenn Chatterjee Lab

[Pranam Chatterjee @pranamanam]

Input a protein sequence + the residue positions to target → potent motif-specific binders in seconds! 🪄 Introducing our newest version of moPPIt! Experimentally validated for domain- and IDR-specific binding, as well as receptor inhibition and CAR-T cell design! 🧫

📜: https://t.co/d2ei8r04FP
🤗: https://t.co/ZbV51OrN1i

🧵👇

[Pranam Chatterjee @pranamanam]

Introducing PeptiVerse 🚀, our open-source platform for therapeutic peptide property prediction. We support WT and modified SMILES inputs, and can predict solubility💧, permeability🔬, hemolysis🩸, non-fouling👯, half-life⏱️, tox ☠️, and binding affinity🔗 -- try it out!

🤗: https://t.co/6kjSJFEGtp
📜: https://t.co/UMhGawlPJi

🧵👇

[Pranam Chatterjee @pranamanam]

Real-world validation time for one of our lab's favorite new therapeutic peptide design algorithms: MOG-DFM! 💥

As a reminder, our multi-objective-guided discrete flow matching (MOG-DFM) algorithm, developed by my brilliant PhD student @TongChen321, learns Pareto-efficient transport in discrete sequence space, re-weighting token velocities through rank-directional scoring and adaptive hypercone filtering to balance competing properties like affinity, solubility, half-life, hemolysis, and non-fouling. ⚖️ Best of all, guidance happens at inference time using black boxed predictors! ⬛

@adaptvbio tested 24 of our MOG-DFM-designed peptides, and 6 bound cleanly to their target, FcRn, a key autoimmune target that recycles IgG. ♻️ We're synthesizing these 6 now in my lab @Penn and doing property testing before moving into antagonism assays and autoimmune animal studies! 🧪 ➡️ 🧫 ➡️ 🐁

It’s so gratifying to see designs from theory to bench so quickly. We're super grateful to @julian_englert, @CotetTudor, and the @adaptvbio team for their collaboration and spotlight! I highly recommend working with them if you're doing anything binder design -- it was so fun! 🙌

Also, if you want to use MOG-DFM for your targets, check out our paper (which was awarded a Spotlight at the ICML @genbio_workshop) and model on @huggingface:

📜: https://t.co/KcV1aU7PDD
🤗: https://t.co/cHAdy6xfZz