A Multi-Step Computational Workflow for Screening and Prioritizing SHP2-Binding Molecules

Background/Objectives: SHP2 (PTPN11) is a key regulator of RAS/MAPK signaling and a well-validated target in cancer and developmental disorders. Designing ligands for its catalytic site is challenging due to the pocket's intrinsic flexibility and the presence of conserved structural water molecules critical for ligand recognition, which limits traditional discovery approaches. This study aimed to systematically identify and prioritize novel SHP2-binding candidates using a computational strategy that accounts for these challenges. Methods: An integrative computational workflow was applied, combining water-aware docking, large-scale virtual screening of 714,409 compounds, MM/GBSA binding free-energy analysis, AI-driven chemical space modeling using ChemBERTa, and microsecond-scale molecular dynamics (MD) simulations. The high-resolution catalytic PTP domain of SHP2 structure was analyzed to identify conserved water molecules (W711, W716, W726, W776) essential for reproducing the crystallographic binding mode of the reference ligand 3LU. Candidates were prioritized based on docking scores, physicochemical criteria, structural inspection, MM/GBSA energetic profiles, and occupancy of distinct chemical space regions. Results: Seven compounds were selected. SwissADME analysis confirmed favorable drug-likeness and GI absorption, with no BBB permeation. ChemBERTa embeddings revealed substantial structural novelty relative to known SHP2 inhibitors. 1 μs molecular dynamics simulations suggested stable binding of compound 4 (2-(3-methyl-2,6-dioxopurin-7-yl)acetate) and persistent interactions with the conserved water network. MM/GBSA evaluation subsequently highlighted its energetically coherent profile. Conclusions: The workflow prioritizes compound 4 as a promising and structurally innovative SHP2-binding candidate. This integrative strategy provides a generalizable approach for targeting proteins with flexible pockets, critical water networks, and limited scaffold diversity, offering a roadmap for challenging computational ligand-prioritization projects.