Structure Prediction and Omics informed Classifier (SPOC) Tool

This tool calculates compact SPOC (cSPOC; Schmid and Walter) scores and other confidence metrics (pDOCKQ, ipTM, avg_models) for user-uploaded AlphaFold multimer (AF-M) predictions. Results are confidential*.

The tool will only process binary predictions of human proteins. For the SPOC score to be valid, predictions should involve structures generated with AF-M default settings (1 ensemble, 3 recycles) to match the training regimen for SPOC.

Drag and drop data files for one or more protein pairs (up to 50) below. For every pair to be analyzed, please supply a total of 6 AF-M output files:

3 ".PDB" structure files
e.g. PROTEIN1-PROTEIN2_unrelaxed_rank_001_alphafold2_multimer_v3_model_4_seed_000.pdb
3 ".JSON" Predicted Angstrom Error (pAE) files
e.g. PROTEIN1-PROTEIN2_scores_rank_001_alphafold2_multimer_v3_model_4_seed_000.json

If more than 3 models per complex are uploaded, the tool randomly analyzes 3 models to mirror the classifier's training regimen.

*We track the number of structures submitted per user, but not which structures were submitted, and structures and results are immediately deleted after analysis.

Loaded files Loaded pairs

So far users have run 388 SPOC jobs and analyzed 6,722MB of data.

Submit analysis job

You will recieve an email with a CSV file of the results once this job is run. Jobs are run in the order that they are recieved and will be completed as soon as possible. Your job is expected to complete in ~5 minutes.

Job name

Results email

By submitting this job you acknowledge that this platform is only for research purposes and that any reported scores and analysis should be interpeted carefully. A high SPOC score is not definitive proof that an interaction is real nor is a low score proof that an interaction is not real.