Software

Learned Graph Network Model for Protein Dynamics

LGNM predicts per-residue flexibility, normal modes, and conformational dynamics using graph neural networks combined with quantum-inspired neural PDE operators. It is designed as a fast, learned alternative to elastic network models and full molecular dynamics simulation.

Users can submit a PDB structure and receive predicted B-factors, mode decompositions, and dynamic flexibility profiles within seconds.

Cancer Biomarker Identification via Graph-Ontology Web Server

CABIgoWEB is a web server for interpretable cancer biomarker discovery. It integrates Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), and GraphSAGE with Gene Ontology schematics, Knowledge Graph Embeddings, and Protein-Protein Interaction networks.

The server allows researchers to submit expression data and receive ranked biomarker candidates alongside attention-based interpretability maps that surface the molecular subnetworks driving predictions.

Privacy-first document tools, entirely client-side

A collection of document utilities — PDF merge/split, OCR, format conversion — that run 100% in the browser. No files are ever uploaded to a server. Built to be a lightweight, privacy-respecting alternative to online PDF tools.