G-SPRI: A Structure-Centric Graph Model for Comprehensive Prediction of Cancer Driver Events from Missense Mutations.

Area of Science:

• Computational Biology
• Genomics
• Structural Bioinformatics

Background:

• Interpreting personal genomes and identifying disease biomarkers requires accurate prediction of missense mutation functional impacts.
• Current *in silico* methods often rely on sequence or basic structural features, neglecting complex biophysical patterns in protein 3D structures.

Purpose of the Study:

• To develop a novel multilevel framework, G-SPRI, for enhanced prediction of missense mutation pathogenicity.
• To leverage atomic-resolution protein geometry and graph-based learning for improved variant interpretation.

Main Methods:

• Developed G-SPRI, a framework utilizing a novel alpha-shape protein graph to capture residue connectivity from atomic geometry.
• Integrated wild-type structural properties and mutation-specific perturbation signals from the Protein Data Bank (PDB).
• Employed graph-based learning for distinguishing pathogenic from benign missense variants.

Main Results:

• G-SPRI demonstrated improved pathogenicity prediction for individual mutations on a binary benchmark.
• Integrated with mutation recurrence, G-SPRI identified more cancer driver genes than state-of-the-art methods from over 2.3 million mutations.
• G-SPRI provided comprehensive evidence for pinpointing likely driver mutations and structurally susceptible regions within disease genes.

Conclusions:

• G-SPRI offers a powerful approach for functional interpretation of missense mutations using protein structural information.
• The framework enhances the discovery of disease-related genetic variants and cancer drivers.
• G-SPRI's ability to quantify pathogenicity and structural influence aids in understanding disease mechanisms.