Home Research Services
Online Services

StoPred Release History

Changes to the web server, newest first
arrow_backBack to StoPred

history Releases

Assembly-state scores, adjustable thresholds and templates current

October 2026
  • Assembly-state scores, run automatically. Besides the stoichiometry, every job reports a homo-oligomer score for one sequence (monomer-aware model) or a co-assembly score for 2–6 distinct sequences. The separate single-sequence model choice is gone.
  • Co-assembly model. Trained with negatives that include one unrelated member swapped into a real complex, so a set with a protein that does not belong scores lower. For 3–6 sequences, every pair and every leave-one-out subset is scored too.
  • Adjustable thresholds. For one sequence or a pair, drag the threshold to see recall, precision and false-positive rate on our evaluation data (PDB entries released after the training cutoff, through 1 October 2026).
  • Optional template search against PDB biological assemblies (snapshot 2026-05-05), with an optional latest release date.
  • Model selection chooses a release: current release (training cutoff 2026-01-01, default) or manuscript release (training cutoff 2023-10-01).
  • Input limits: at most six distinct sequences per job; a sequence repeated under two headers is not accepted.
  • New result page with a summary at the top and detailed views in tabs.
  • Jobs submitted earlier keep their original results.

New model collection and download page

30 May 2026
  • Models trained on PDB entries up to 2026-01-01 deployed: default complex model default_20260101 and single-sequence unknown model unk_single_20260101; the manuscript complex model default_20231001 kept.
  • Model selection on the submission form; dates in model names are training cutoffs.
  • Separate download page for the container image, model releases and model naming documentation.

FreddolinoLabumich.edu | 1150 W. Medical Center Dr, Ann Arbor, MI 48109-0600