Ahh, I know I've ignited a big sequence vs. structure debate (unintentionally, I promise -- I'm very agreeable!) 🔥 But just wanted to share the good news that the @NIH @NIGMS recently awarded my lab the R35 MIRA Award to develop isoform-specific therapeutics to undruggable, unstructured targets using generative language models! 🌟 We're excited to take this challenge on over the next 5+ years, and are also very grateful for the trust placed in us to go after these (currently) incurable diseases! 🙏 Read more about our recent progress and plans here: pratt.duke.edu/news/chatterj…

Super excited to share our new protein language model: FusOn-pLM, developed by my wonderful first-year PhD student @SophieVincoff and her team of amazing undergrads!! 🎉 Sure, we know AlphaFold3/RFDiffusion and other structure prediction/design methods are great -- but what about for conformationally disordered proteins? 🤔 Well, you don't get more disordered than fusion oncoproteins (the core drivers of multiple pediatric cancers) which are large, heavily disordered, and have no druggable binding pockets -- no chance that either structure-based methods or small molecules get to these. 😖 As you probably expect from us ➡️ protein language models to the rescue!! 🦸 But wait, wait 🫷, no pLM has been trained on these "fused" sequences...at least not until now! 👉 Here, we tune the query weights of the final 11 layers in the SOTA ESM-2-650M model on over 35k fusion oncoprotein sequences! We specifically introduce a novel MLM strategy, employing our SaLT&PepPr (nature.com/articles/s42003-0…) binding-site probability predictor to focus masking to interaction-mediating amino acids. By attending to these key residues, our final embeddings outperform all others on both fusion oncoprotein-specific benchmarks and disorder prediction tasks (like CAID), while maintaining distinct representations of fusion sequences to their head and tail proteins! 🌟 So what's next? We can use FusOn-pLM embeddings for fusion oncoprotein-targeted design -- more updates to come on that front! 😉 In the meantime, please feel free to read our paper and try out our model on @huggingface! 🤗 Paper: biorxiv.org/content/10.1101/… Model: huggingface.co/ChatterjeeLab…