NeuroGeno3D: A unified clinical decision support system integrating structural biology, molecular interpretation, and clinical evidence for precision neuro-oncology.
Abstract
Translating genetic variations into clinically useful molecular interpretations in neuro-oncology remains difficult, requiring physicians to examine many databases and resources. We created NeuroGeno3D, a browser-based clinical decision support platform that combines structural bioinformatics, molecular pathology, and clinical evidence in a single interactive process. The platform integrates canonical protein sequence validation, interactive 3D WebGL visualization, and real-time AI-assisted structural prediction with ESMFold for protein sequences up to 500 amino acids. Detected variations are analyzed by a deterministic WHO CNS categorization support engine and linked to downstream signaling pathways using Reactome and KEGG databases. NeuroGeno3D also incorporates ClinVar annotations, PubMed literature, and appropriate clinical guidelines from NCCN, CNS, AANS, and EANO into a sequential pre-, intra-, and post-operative workflow. Predefined hotspot variations are evaluated in median 1.12 s (IQR: 1.02-1.26 s) under typical settings, while custom protein sequences take median 6.30 s (IQR: 5.70-7.15 s) to process. In retrospective clinical evaluation across 140 adult diffuse glioma cases, NeuroGeno3D achieved an overall diagnostic accuracy of 92.1 % (95 % CI: 86.4 %-96.0 %), supported by excellent usability scores (mean System Usability Scale score: 88.9 ± 4.8/100, Grade A + ). The platform also provides structured, evidence-based reports that bring together molecular discoveries, structural visualization, pathway context, and treatment relevance in a single interface. By combining these components, the platform enables rapid literature exploration for rare genomic variants, supporting research alongside clinical interpretation, and lays the groundwork for future potential real-time intraoperative molecular diagnostic and decision-support procedures.