Ali Madani

FLIP2 adds seven protein fitness datasets and 16 splits for enzymes, light-sensitive proteins, and interactions; simple ridge and zero-shot baselines often rival fine-tuned protein language 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.

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 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.

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.

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.

This paper shows that deep learning language models can generate functional artificial proteins across families, with activities similar to natural proteins and structures resembling natural folds.

One attention head of a protein BERT model tracks 3D contacts (28% vs 1.3% background); other heads target binding sites, and deeper layers attend to higher-level properties.

A 1.2B-parameter autoregressive model trained on 280M tagged sequences completes held-out proteins with near-native Rosetta energies and ranks high-fitness GB1 variants without any GB1 training.