Biased Fab phage-display libraries were designed to determine whether inhibitory CDR H3 motifs from potent anti-matriptase antibodies could be transferred to target homologous serine proteases. Using reverse-binding and substrate-like H3 motifs from parental clones A11 and E2 as templates, six synthetic libraries with 1010 diversity were constructed. Selection against matriptase identified sixteen inhibitors with sub-100~nM potency, representing 100,000-fold improvement over circularized H3 loops alone. Selection against TMPRSS2, a serine protease implicated in viral entry and prostate cancer with 43% sequence identity to matriptase, yielded binders with micromolar inhibitory potency. Selection against urokinase plasminogen activator (uPA, 35% identity) identified binders that adopted a substrate-like CDR H3 binding mode in our structural models. Across all reference structures, including the separately identified uPA inhibitor AB2 (PDB: 9PYF, deposited with this work), benchmarking of five co-folding methods and rigid-body docking showed that co-folding consistently achieved acceptable to high quality DockQ scores, outperforming traditional docking and capturing the recognition of key active site determinants. Ensemble predictions of mutational binding energy changes (DeltaDeltaG) using these models identified key paratope-epitope interactions, with predictions validated through mutagenesis. This work establishes a framework integrating biased antibody libraries with computational structure prediction and analysis for targeting conserved protease epitopes.
Toward terminological clarity in digital biomarker research
Digital biomarker research has generated thousands of publications demonstrating associations between sensor-derived measures and clinical conditions, yet clinical adoption remains negligible. We identify a foundational


