AlphaFold2, OpenFold, Protein Language Models and Beyond | Nazim Bouatta

Показать описание

Title: Single-sequence protein structure prediction using language models from deep-learning

Abstract: AlphaFold2 and related computational systems predict protein structure using deep learning and co-evolutionary relationships encoded in multiple sequence alignments (MSAs). Despite high prediction accuracy achieved by these systems, challenges remain in (1) prediction of orphan and rapidly evolving proteins for which an MSA cannot be generated; (2) rapid exploration of designed structures; and (3) understanding the rules governing spontaneous polypeptide folding in solution. Here we report development of an end-to-end differentiable recurrent geometric network (RGN) that uses a protein language model (AminoBERT) to learn latent structural information from unaligned proteins. A linked geometric module compactly represents Cα backbone geometry in a translationally and rotationally invariant way. On average, RGN2 outperforms AlphaFold2 and RoseTTAFold on orphan proteins and classes of designed proteins while achieving up to a 106-fold reduction in compute time. These findings demonstrate the practical and theoretical strengths of protein language models relative to MSAs in structure prediction.

~

Chapters:

00:00 - Intro
12:11 - Overview of Machine Learning and Protein Modelling
20:05 - Overview of AlphaFold2: Strengths, Limitations and Remaining Challenges
29:08 - Introducing OpenFold
46:22 - RGN2 - Single Sequence and Language Model
55:11 - Q+A

Valence Labs

Рекомендации по теме

Комментарии

These technical explanations are difficult for me to understand as a non-techie, but they help provide a clearer prediction of what is actually going on than the headlines, which are often written by people who don't understand the technology themselves, and then commented on Reddit by people who don't understand it either. And what is going on is actually more remarkable than what many of the headlines (and posters) say. RGN2's abilities are insane.

squamish

I was expecting a technical detailing, but its more on manager's
level.

naomifridman

AlphaFold2, OpenFold, Protein Language Models and Beyond | Nazim Bouatta

AlphaFold2, OpenFold, Protein Language Models and Beyond | Nazim Bouatta

Meta ESM-2 Fold - AI faster than Alphafold 2

Nazim Bouatta | Machine learning for protein structure prediction, Part 2: AlphaFold2 architecture

How to use Google Colab Alphafold2 and ChimeraX

Using Protein Language Models For Drug Discovery

Nazim Bouatta | Machine learning for protein structure prediction, Part 3: AlphaFold2 and OpenFold

Mohammed AlQuraishi - OpenFold: Lessons and insights from rebuilding and retraining AlphaFold2

Initial reaction to AlphaFold2 right after CASP14

OpenFold: Lesson Learned and Insights Gained from Rebuilding and Retraining AF2

AlphaFold2 reveals commonalities and novelties in... - Nicola Bordin - 3DSIG - Abstract - ISMB 2022

OpenFold

OpenFold: Lessons Learned and Insights Gained From Rebuilding and Retraining AlphaFold2

BERTology Meets Biology: Interpreting Attention in Protein Language Models (Paper Explained)

OpenFold: retraining AlphaFold2 yields insights on its learning mechanisms & generalization capa...

Language Models Generate New Proteins! Here's How.

DeepMind solves protein folding | AlphaFold 2

Nazim Bouatta | Machine learning for protein structure prediction, Part 1: Algorithm space

GCC2023 Day 3 Keynote - Kate Michie - Alphafold2 and the Age of Deep Learning

AlphaFold vs OpenFold - accelerating time to result in protein folding

Protein Folding with Hugging Face and Python

ESM: Protein Structure Prediction with LLM

Proteins as language: NLP, Machine Learning & Protein... - Dan Ofer - MLCSB - Poster - ISMB 2022

State-of-the-Art Estimation of Protein Model Accuracy using AlphaFold

ProteinBERT: a universal deep-learning model of... - Dan Ofer - Function - Abstract - ISMB 2022