bioRxiv2025
Proteins that bind to a target protein of interest, termed "binders," are essential components of biological research reagents and therapeutics. Target proteins present multiple binding surfaces with varying interaction potential. High-potential surfaces, or "hot spots," are experimentally identified as the most probable binding sites in de novo discovery campaigns. However, hot spots and their default binding modes do not always confer the desired specificity. Related proteins or isoforms often share similar hot spots, resulting in promiscuous binding. Interaction with a hot spot may also fail to elicit the intended biological outcome. Consequently, methods that direct de novo binder discovery toward targets with defined specificity are critically needed. We recently developed phage-assisted non-continuous selection of binders (PANCS-Binders), a selection platform with unparalleled speed and sequence-function fidelity that enables routine de novo binder discovery within days. However, because PANCS-Binder selections enrich variants based primarily on affinity, secondary screening is unlikely to identify binders to lesser hot spots because of the high likelihood of convergence. These alternative binding surfaces with weaker inherent interactions may possess desirable specificity profiles. Here, we develop PANCS-spec-Binders, which incorporates simultaneous selection and counterselection to control the specificity of enriched binders. We demonstrate PANCS-spec-Binders in two proof- of-concept applications: (1) discovery of isoform-selective binders that bind HRAS with >100-fold higher affinity than the highly related KRAS isoform, and (2) discovery of epitope-specific binders that either target or avoid the LIR interaction region of LC3B. PANCS-spec-Binders enables rapid identification of binders with defined specificity within days. SIGNIFICANCEAffinity reagents, termed "binders", are essential tools in research and therapeutic development. Binders generally require high specificity either at the selectivity level (binding only the target protein, but not related proteins) or at the epitope level (binding only at a specific surface on the target rather than another). While methods to discover binders in general have progressed, identifying binders with defined specificity features often requires extensive secondary screening and frequently results in failure. Here, we adapt our recently developed binder discovery platform to solve these two selectivity problems. By rapidly screening billions of variants, we can direct specificity between highly related proteins, direct binding to a specific epitope, and, because of the fidelity of our selections, identify binders that specifically avoid a defined epitope.