June 2026
Model Performance Series
Astra AI on Ion Channels
Ion channels are among the most consequential — and treacherous — classes in drug discovery: validated targets across pain, epilepsy, and arrhythmia, home to the hERG cardiac anti-target, and structurally diverse, with drug sites buried in the conduction pathway. This report shows how Orbion's Astra AI Suite characterizes them from sequence alone — function, topology, PTM sites, and ligand/pore-block pockets — measured against public experimental ground truth across 2,700 ion-channel proteins.
0.97 AUROC Topology
56% Ligand Recall
F1 Up to 0.90 for PTMs
0.97 AUROC on TM Topology, Single-pass to 24-TM Giants
Per-residue TM prediction agrees with UniProt at AUROC 0.97 (AUPRC 0.91, F1 0.87) across the full topological range. The 24-TM voltage-gated Ca²⁺ channels are the hardest (0.82).
Recovers Known hERG Pore Blockers
The 56% aggregate masks the result that matters — a clean biological split. On ligand-gated channels the model localizes the pocket well; on voltage-gated channels including the hERG safety case it recovers the ligand identity (the worked example recovers E4031, BeKm-1, ErgTx-1) but the pore-block site is structurally diffuse — read the panel as a ligand hypothesis set, not a residue-contact map.
F1 Up to 0.90 on PTM Site Prediction
Per-residue sites across 39 classes; strongest on disulfide bonds (0.90) and N-linked glycosylation (0.88) — the extracellular modifications that govern channel folding and surface expression.
Have a question?
Whether you want to understand our models, test a specific protein, or just ask how Orbion applies to your research — send us your question and we'll get back to you. Or, if you'd prefer a live walkthrough, book a demo.
