PD-L1 Small-Molecule Discovery
Nearly every major PD-L1 small-molecule dimerization-inducer campaign—Bristol Myers Squibb, Incyte, Gilead, Arbutus Biopharma—converged on closely related biphenyl cores, owing to the constraints of the tight protein–protein interface. The surrounding patent landscape was crowded.
I led the InterX/NeoTX redesign effort that produced a novel phenyl-benzothiophene chemotype with low-nanomolar biochemical activity and a binding mode confirmed by X-ray crystallography. Using custom cheminformatics scripting, I then extended the design into sp³-enriched spirocyclic scaffolds intended to reduce aromaticity and improve developability.
Manuscript in preparation.
Human Mediator and Transcriptional Regulation
How intrinsically disordered activation domains of transcription factors tune RNA polymerase II activity through coactivators remains one of the central unresolved problems in gene regulation.
I built a structural model of this coupling from large-scale AlphaFold 3 co-folding of human activation domains with the Mediator Tail, performed at the practical size limit of current structure-prediction workflows and tested against cryo-EM, biochemical, and functional data. The model shows how diverse transcription factors route regulatory information through Mediator to set Pol II initiation, providing a structural entry point into disease-relevant transcriptional programs.
The work rests on an AI-assisted reasoning and structural-bioinformatics infrastructure I built: a hierarchical agent system for code development, data-quality review, adversarial hypothesis testing, and synthesis across the full evidence base.
Manuscript in preparation.
Promoter Recognition and Opening
My earlier work on bacterial transcription developed as a continuous experimental and conceptual program. Using SELEX, I discovered a previously unrecognized promoter element; the same experiments produced a minimal protein–DNA system suitable for crystallography (Molecular Cell, 2006).
The co-crystal structure it enabled, together with accompanying biochemistry, overturned the long-standing view that RNA polymerase recognizes the −10 element as intact duplex DNA. The enzyme instead captures conserved bases flipped out of the double helix — a mechanism now described in textbooks (Cell, 2011).
A second assumption held that the polymerase clamp must open to admit promoter DNA and close only after loading. I combined stopped-flow fluorescence kinetics with reporters on the enzyme and on DNA, using antibiotic inhibitors to lock the clamp in distinct conformations and resolve its successive motions in real time. Clamp closure proved to be an obligatory step during melting, revising the classical open–load–close model (Science, 2017).
I later integrated the structural and kinetic evidence into the steric-filter model, a unified framework for promoter opening by cellular RNA polymerases (TiBS, 2024).
