DEVELOPMENT OF STRATEGIC SCRIPTS FOR IDENTIFYING PROTEIN PARTNERS AND EXOSITES: A CASE STUDY WITH HUMAN Α-THROMBIN

Human α-thrombin (factor IIa) is a trypsin-like serine protease involved not only in blood coagulation, but also in cellular processes including inflammation, cytoprotection, and apoptosis. Its multifunctional nature is largely attributed to two positively charged surface regions, termed exosite I and exosite II, which mediate interactions with a wide variety of substrates, cofactors, receptors, and inhibitors. Owing to its well-characterized interaction network, α-thrombin was selected as a model to validate a computational pipeline designed to identify functional exosites and predict protein partners from target-directed phage display data. Phage display selections were carried out against PMSF-inhibited factor IIa, followed by next-generation sequencing (NGS) to uncover 8,516 unique peptide sequences. Peptides were ranked based on enrichment metrics generated by the Target-Specific Analysis Tool (TSAT), and the 1,000 highest-ranked sequences were aligned with proteins previously identified as thrombin interaction partners, including antithrombin III, GPIb-α, fibrinogen, protein C inhibitor, and prothrombin fragment 2. Structural features, including secondary structure and solvent accessibility (DSSP), were incorporated into a custom Python pipeline to eliminate incompatible alignments and minimize false-positive predictions. AlphaFold2 was used to model the structure of high-scoring peptides, which were subsequently submitted to blind molecular docking against α-thrombin using GOLD. The predicted hotspots were in strong agreement with experimentally validated thrombin interaction interfaces, indicating that phage display-enriched peptides retain molecular features relevant to protein–protein recognition and can be used to identify functional exosites. Beyond recovering established interaction regions, the approach also identified previously uncharacterized surface patches that represent candidate binding sites for future experimental investigation. These findings validate the computational workflow as a robust strategy for exosite identification, laying the groundwork for future enhancements such as graph-based structural representations to refine interface mapping and prediction specificity.

This work was supported by Fund. Amparo Pesq. Estado São Paulo (FAPESP #2024/03185-6, #2020/08615-8) and by Conselho Nac. Des. Cient. Tecnológico (CNPq #309940/2019-2).