Mark J. S. Kelly

This paper introduces a hybrid physics-deep learning framework for designing small-molecule binding proteins, achieving experimental success in binding affinity and structural accuracy.

This study characterizes AcrIF11, an ADP-ribosyltransferase that inhibits CRISPR-Cas systems with high potency and specificity, and shows its reversibility through macrodomain enzymes.

This paper designs dynamic proteins with switch-like behavior using deep learning, validated by structure solving and simulations showing controllable conformational changes.

This paper designs a protein to bind naphthalenediimides and evaluates its ability to generate reactive species for proximity labeling.

This paper identifies the requirements for efficient endosomal escape by ZF5.3, a mini-protein, showing pH-induced unfolding and lipid-BMP interaction are critical for its function.

This paper introduces Frame2seq, a structure-conditioned masked language model for protein sequence design that generates novel, stable proteins with low sequence identity to native sequences.

The paper investigates how the GTPase Ran/Gsp1 achieves functional multi-specificity through biophysical and genetic perturbations, revealing that its effects are linked to kinetic parameters of the switch cycle rather than direct interaction sites.