Enhanced computational strategies for categorizing HER2 kinase domain variants of uncertain significance through structural and functional analysis.

The HER2 (human epidermal growth factor receptor 2) kinase domain plays a pivotal role in receptor-mediated signaling and is frequently mutated across various cancer types. Variants of uncertain significance (VUS) within this domain pose significant challenges for clinical interpretation. In this study, a comprehensive computational analysis was conducted on 97 HER2 kinase domain variants comprising 25 pathogenic and 72 VUS entries using a panel of 13 predictive algorithms, including PredictSNP, PolyPhen-2, SIFT, and CADD. A subset of 32 variants (11 pathogenic and 21 VUS) was prioritized for in-depth structural and functional evaluation, based on concordant deleterious predictions from ≥ 9 algorithms. Conservation profiling via ConSurf and Align-GVGD revealed that these variants predominantly localize to highly conserved and functionally critical residues, underscoring their potential biological relevance. Thermodynamic stability profiling using I-Mutant indicated that the majority of prioritized variants exerted destabilizing effects on the protein's tertiary structure, consistent with their pathogenic annotations. Extended molecular dynamics simulations (200 ns) revealed noticeable deviations in root mean square displacement (RMSD), root mean square fluctuation (RMSF), and solvent-accessible surface area (SASA), accompanied by reductions in intramolecular hydrogen bonding and global structural compactness, highlighting conformational destabilization. Stereochemical integrity, assessed through Ramachandran plot analysis, was compromised in specific variants. At the same time, perturbations in hydrophobic core interactions and secondary structure elements further underscored disruptions in core packing and local conformational stability. Notably, the R816P variant demonstrated minimal perturbation to structural dynamics, highlighting the necessity of integrating time-resolved simulations with sequence-based pathogenicity predictions.