To solve this, Birnbaum and Keating present PottsMPNN, a sequence design model that learns a sequence-energy landscape from MSAs. PottsMPNN outperforms other sequence design models and is a drop-in alternative to ProteinMPNN. Code: https://t.co/k8j16OGBio
(2/2)
Here’s a new preprint from the Keating Lab: https://t.co/vUyQs3cWH1
Foster Birnbaum and Amy E. Keating demonstrate that sequence design models, such as ProteinMPNN, are limited because they were trained only on native sequences. (1/2)
#ProteinDesign#StructuralBiology
We’re expanding our digital footprint!
To keep up with the latest protein design research, lab news, and updates from the Keating Lab, follow us over on Bluesky: https://t.co/llPgQCGfSG.
See you there! #ProteinDesign#StructuralBiology
A self-supervised approach aligning protein sequence and structure spaces enables efficient binder screening with only backbone structural information — a powerful asset for early-stage protein binder design.
🔗 https://t.co/u9hfX7Fz1Y
To learn more, read out the publication in PRX Life and use the publicly available code (https://t.co/TwPz6UqYpQ). Thanks to Foster Birnbaum, Saachi Jain, Aleksander Madry, and Amy E. Keating for this work! 3/3
New work from the lab! Check out RLA (https://t.co/qVp8ROiwgl), a contrastive-learning approach that assesses sequence-structure compatibility by aligning sequence and structure machine learning representations! RLA can successfully filter protein binder designs. 1/3 🧶
RLA has been tested on several benchmark sets, including several design libraries of miniprotein designs for a variety of protein targets. For all but two targets, filtering with RLA results in a higher success rate after subsequent AF2-based filtering. 2/3
First twitter thread🧵and also my first BioRxiv preprint! I’m excited to finally release my undergrad work into the world: combining GNNs, Potts models, and Tertiary Motifs (TERMs) for protein design!
See the preprint here: https://t.co/aiVoRA8g6S
1/
Our first keynote speaker: Dr. Amy Keating (@keating_lab). Interested in protein interaction specificity, Dr. Keating highlights the power of data-driven computational exploration of protein interactions. Dr. Keating was our student choice of #PEC2022 and we are ecstatc to host!
Fast, reliable, computational methods for designing protein-binding peptides would be immensely useful. As a first step, we show that tertiary structural motifs from the PDB can be used to reconstruct known peptide structures and generate new ones. https://t.co/5tiUlY5wF5 (1/9)
Scientists in @keating_lab designed a screening method to probe how short stretches of amino acids called SLiMs selectively bind to certain proteins, and distinguish between binding partners with similar structures. I covered this recent work for MIT News: https://t.co/AzT4N5Wcs6
Excited to share our newest work! We describe a surprising mechanism behind how a short linear motif binding domain achieves interaction specificity.
https://t.co/NrPm2I6T5e
@KevinKaichuang@KeatingLab Hi Kevin! Thanks for sharing our review. We're not quite sure who nabbed the original @KeatingLab twitter handle, but we assure you that we're the real thing, and they're the impostor :)
Review of data-driven protein design, including structure, sequences, and high-throughput functional datasets.
Vincent Frappier @KeatingLab
https://t.co/clioPkZWnN
Read about the amazing accomplishments of alum MIT biologist and president of The Protein Society, Prof. Amy Keating (PhD ’98 Houk/García-Garibay groups) https://t.co/N0pbs4wJZf @houk1000@GaribayLab@MITBiology@ProteinSociety