Joshua Meier

This paper introduces a computational antibody design platform that generates drug-like antibodies with high developability and atomic accuracy, demonstrating its utility for challenging therapeutic targets.

This paper introduces Chai-2, a model that designs antibodies with a 16% hit rate, achieving success in 50% of targets without prior binders, and enables rapid validation in under two weeks.

Chai-1 matches AlphaFold3 on PoseBusters ligand poses (77% vs 76%), beats AlphaFold-Multimer 2.3 on low-homology interfaces, and stays competitive without MSAs on multimers and antibodies.

IgDesign designs antibody CDRs using deep learning and validates them in vitro, showing high success rates and improved affinities for multiple therapeutic antigens.

The paper develops deep learning models trained on synonymous variant datasets to predict and optimize protein expression in E. coli.

Deep generative models combined with high-throughput screening design de novo anti-HER2 antibodies without further optimization, yielding nanomolar binders functionally comparable or superior to trastuzumab.

Language models trained on high-throughput affinity data predict KD of unseen antibody variants across three orders of magnitude, and a "naturalness" score can be optimised alongside affinity.

A single pre-trained protein language model, ESM-1v, scores mutations zero-shot about as well as family-specific DeepSequence across 41 deep mutational scans, with no experimental labels or retraining.

The paper introduces a Neural Potts Model that predicts pairwise coupling matrices for protein families using amortized optimization, showing improved performance on low-depth multiple sequence alignments.

The paper trains a deep language model on 250 million protein sequences to learn biological representations that capture structural and functional properties.

One protein language model trained across many families on whole multiple sequence alignments, with interleaved row and column attention, reportedly outperforms unsupervised structure-learning methods by a wide margin.

Attention maps of the 650M-parameter protein language model ESM-1b, read out by a logistic regression fitted on 20 proteins, predict contacts better on average than Gremlin's alignment-based Potts models.

GUIDES designs CRISPR libraries for loss-of-function screens using RNA-seq and protein annotations to target exons and functional domains.