Protein Structure & Prediction

LIVE

Protein Variant Landscape

A UniProt-style scatter plot of every observed variant for the current gene, plotted on a position × amino-acid-substitution grid. Horizontal axis: protein position. Vertical axis: substituted amino acid (21 rows: 20 residues plus stop). Colored domain bands run across the background. Each observed variant is a dot color-coded by pathogenicity class.

PRESR heatmap toggle (STXBP1-only)

Fills the background with per-cell pathogenicity predictions for every possible missense substitution — for STXBP1 that is 11,142 cells — shaded red (predicted pathogenic), amber (uncertain), or green (predicted neutral).

3D Protein Viewer

Backed by 3Dmol.js, loaded on demand. Representations: cartoon, licorice, surface, backbone, cartoon+stick. Coloring: spectrum, hydrophobicity, polarity, pLDDT confidence. Measurement mode for distances and angles. Optional ghost wild-type overlay behind the mutant.

ESMFold structure prediction

On-demand prediction via NVIDIA NIM. Preset proteins (PEX10, TCF4, DYRK1A, CHD2, and others) for quick-load. Per-residue pLDDT confidence coloring. Cascading fallback on the NIM endpoint check. Requires an NVIDIA NIM API key.

Explain Mode

Click any residue for a two-tab card. Simple tab: parent-friendly description, gene-specific glossary, distance to variant site. Deep tab: pLDDT, B-factor, solvent accessibility, PRESR score, nearby neighbors, structural descriptors.

Why it exists

The same viewer serves the researcher and the caregiver who wants to understand what a residue actually does. The landscape puts observed and predicted data in the same visual space.