What problem does it solve?
Enable automated prediction and scoring of protein structures from amino acid sequences or FASTA inputs to support model selection, downstream structural analysis, and reproducible pipeline integration.
Core Features & Use Cases
- Sequence and FASTA Input: Accepts comma-separated sequences or a FASTA file to generate predictions in sequence or fasta modes.
- Per-Model Scoring Summaries: Produces per-model summaries including model index, CIF path, per-model scores, and an aggregated best_model selection.
- Operational Controls: Supports sample count, dry-run mode for input preparation without inference, and an info mode for metadata inspection.
- Quality & Download Guidance: Includes confidence checkpoints (pLDDT, pTM, ipTM), a strict output-download policy using server_file_to_base64, and a recommended preprocessing step for PDB inputs.
- Use Case Example: Run a five-sample dry-run prediction from a short membrane peptide sequence to inspect input.fasta and verify scoring summaries before full inference.
Quick Start
Use molclaw-chai1-predict to run a dry-run Chai-1 prediction on a provided sequence, generate input.fasta, and return per-model scoring summaries.
