Brian A. Kuhlman

This paper introduces HBDesigner, a deep learning tool for designing hydrogen-bond networks in proteins, showing improved performance over existing methods.

This paper designs de novo proteins that bind DELE1, block its oligomerization, and inhibit the mitochondrial stress response in cells.

This paper uses AI to design protein masks that control TNF-family ligand assembly and activity, enabling conditional activation and improved therapeutic applications.

The paper introduces Caterpie, a live cell labeling system using split fluorescent proteins to enable high-fidelity identification of multiple cell populations with 97% accuracy.

This study compares mRNA and yeast display technologies for evaluating ML-designed miniprotein binders, finding mRNA display better enriches slow dissociating binders and achieves higher library coverage.

This study identifies PRAME as a therapeutic target in NUT carcinoma, showing its high expression and susceptibility to T-cell mediated cytotoxicity via PRAME epitopes presented by HLA Class I molecules.

This paper designs orthogonal TCR β interfaces to prevent mispairing in dual-TCR therapeutics, achieving high pairing fidelity and demonstrating enhanced potency in targeting cancer antigens.

The paper develops a computational platform to stabilize human VH domains for creating multispecific biologics with conditional activity.

The paper designs a modified tyrosinase enzyme with a substrate recruitment domain to selectively modify tyrosine residues in peptides, enabling efficient bioconjugation of cytotoxic molecules to antibodies.

This paper designs a water-soluble CD20 mimic using computational methods to enable binding studies and therapeutic development.

This study develops a tetravalent dengue vaccine using stabilized E dimers on liposomes, eliciting type-specific neutralizing antibodies without immune interference.

This study evaluates a stabilized DENV2 E homodimer vaccine, finding it elicits broader neutralizing antibody responses targeting multiple E protein domains compared to the wild-type monomer.

The paper benchmarks protein stability models on double point mutations and finds they fail to capture epistatic interactions, with some improvements from epistasis-aware models.

This paper introduces ThermoMPNN, a deep learning model for predicting protein stability changes using transfer learning and a large dataset, achieving competitive performance on benchmarks.

The paper designs de novo proteins that inhibit oncogenic Gq by embedding binding motifs, achieving high stability and potency without further optimization.

The paper presents a framework for ensuring reproducibility in bio-macromolecular modeling through automated benchmarks and a test server, demonstrating its utility for Rosetta and other tools.

This study uses optogenetic control of Set2 in yeast to show rapid, transcription-dependent H3K36me3 deposition and transcription-independent, stochastic demethylation.