Jeffrey A. Ruffolo

E1: Retrieval-Augmented Protein Encoder Models

The paper introduces Profluent-E1, retrieval-augmented protein language models that improve performance by conditioning on homologous sequences, achieving state-of-the-art results in fitness prediction and contact-map prediction.

Contextual protein and antibody encodings from equivariant graph transformers

The paper introduces a method to learn protein and antibody residue representations using equivariant graph transformers, showing structural context can determine amino acid labels and improve antibody design.

Engineering of CRISPR-Cas PAM recognition using deep learning of vast evolutionary data

This paper introduces Protein2PAM, a deep learning model that predicts CRISPR-Cas PAM specificity from protein sequences, enabling the design of Cas variants with tailored PAM recognition.

Scaling unlocks broader generation and deeper functional understanding of proteins

Scaling sparse Mixture-of-Experts protein language models up to 46B parameters on 1.5T curated tokens broadens viable protein generation across families and amplifies gains from laboratory preference alignment.

Adapting protein language models for structure-conditioned design

This paper introduces proseLM, a protein design method that adapts language models with structural and functional context, improving residue recovery and experimental performance in genome editing and antibody design.

Enhancement of antibody thermostability and affinity by computational design in the absence of antigen

This study uses DeepAb to design antibody variants with enhanced stability and affinity without antigen information, showing 91% and 94% improvements in thermal/colloidal stability and affinity, respectively.

Design of highly functional genome editors by modeling the universe of CRISPR-Cas sequences

Language models trained on over one million mined CRISPR operons generated Cas9-like effectors and matching guide RNAs, and an AI-designed editor is reported to edit the human genome precisely.

FLAb: Benchmarking deep learning methods for antibody fitness prediction

FLAb benchmarks deep learning models for predicting antibody fitness properties across six metrics, finding no model consistently correlates with all properties.

Flexible Protein-Protein Docking with a Multi-Track Iterative Transformer

GeoDock predicts protein-protein docked structures using a multi-track iterative transformer, achieving high success rates and fast inference speeds for rigid and flexible targets.

Hallucinating structure-conditioned antibody libraries for target-specific binders

The paper introduces FvHallucinator, a deep learning model that generates antibody CDR libraries conditioned on structure, improving binding affinity and enriching for therapeutic candidates.

Fast, accurate antibody structure prediction from deep learning on massive set of natural antibodies

The paper introduces IgFold, a deep learning method for fast and accurate antibody structure prediction, achieving results comparable to AlphaFold in under a minute.