OpenDDE Co-folding Screen

Co-fold every ligand in a SMILES library with a protein sequence using OpenDDE and rank the predicted complexes by interface confidence.

by Temple Compute

Run OpenDDE Co-folding Screen on Horus

W-36 · OpenDDE Co-folding Screen

Co-fold every ligand in a SMILES library with a protein sequence using OpenDDE and rank the predicted complexes by interface confidence.

Domain: Drug Discovery

Overview

Co-folds each ligand in a SMILES library with a protein sequence using OpenDDE (Aureka Research, Apache-2.0) and ranks the predicted complexes by interface confidence. OpenDDE has no affinity head. The workflow has 3 tasks.

Pipeline

prep    target.fasta + ligands.smi ──► results/opendde_inputs.tar.gz (jobs.json)
predict opendde pred (+ MSA search) ──► results/predictions.tar.gz
rank    parse summary confidence    ──► results/top_hits.csv

Inputs / Outputs

fasta: target protein; the first record is used.
ligands: one SMILES [name] per line; the name becomes the job and output id.
results/predictions.tar.gz: CIF structures and confidence JSONs.
results/top_hits.csv: rank, ligand, iptm, ranking_score, ptm, plddt, has_clash, cif. iptm is interface confidence, not binding affinity.

Parameters

| id | Default | Meaning | |---|---|---| | fasta | examples/target.fasta | Target protein | | ligands | examples/ligands.smi | Ligand library | | use_msa | on | Fetch a protein MSA via the ColabFold API before co-folding | | n_samples | 1 | Diffusion samples per job (1-10); the best per ligand is kept |