Welcome to Q-Torsion

To explore the potential of quantum methods in solving biological problems, we developed a quantum-inspired SVM using a kernel derived from quantum feature mapping principles. On balanced phi (φ) and psi (ψ) torsion angle datasets, it modestly outperforms the classical SVM.

This app lets you test both models using protein structures. DSSP provides ground truth secondary structure and torsion angles, which are then classified as allowed or disallowed by each of the two classifiers.

In a generative capacity users can generate 1,000s of high-confidence allowed φ/ψ samples from the quantum model by specifying a seed.

Interested in trying the quantum kernel? Start with any protein structure.

Enter a PDB_Chain ID (e.g., 1hv4_A) or an AF ID preceded by 'AF-' (e.g., AF-A0A021WW64)

Or upload a single-chain CIF file:

Run inference

Figure: Protein backbone torsion angles (φ, ψ) describe the conformational flexibility of residues and are fundamental to structural integrity. Experimental measurements from DSSP provide ground-truth distributions, which are embedded into Hilbert-like feature spaces and used to fit a quantum-inspired kernel-based classifier. The resulting decision boundary is visualized with green indicating allowed regions, red indicating disallowed regions, and yellow highlighting zones of transition between these states.