agentsclimarketplace

Alterlab alphafold

Skill AlterLab-IEU/AlterLab-Academic-Skills/skills/bioinformatics/alterlab-alphafold

239 evaluated academic Claude/agent skills across 17 research domains (bioinformatics, data science, clinical, social-science methods, Turkish academia & more). Executable eval per skill, deterministic citation verifier, research→write→review→publish pipeline, and a skill-finder front door. Claude Code, Cursor, Codex, Gemini CLI & Copilot.

Install
npx -y skills add AlterLab-IEU/AlterLab-Academic-Skills --skill alterlab-alphafold

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Predict protein 3D structures with AlphaFold2 via ColabFold — MMseqs2-accelerated MSAs, monomer and AlphaFold2-Multimer complex folding, and confidence-based validation (pLDDT, pTM/ipTM, PAE). Use when folding a protein sequence or complex from FASTA, generating a predicted structure with confidence metrics, ranking models, or checking self-consistency of a design. For co-folding a protein WITH a small-molecule ligand or predicting binding affinity prefer alterlab-boltz; for antibody–antigen or one-FASTA multi-entity complexes prefer alterlab-chai; to LOOK UP an already-computed structure prefer alterlab-alphafold-db; for ESM embeddings or inverse folding prefer alterlab-esm. Part of the AlterLab Academic Skills suite.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

4.8 KB, as published. Nobody here has run it

AlphaFold (via ColabFold)

Overview

Predict a protein's 3D structure from its amino-acid sequence with AlphaFold2, run through ColabFold (Mirdita et al., Nature Methods 2022) — which replaces AlphaFold's slow genetic-database MSA search with the fast MMseqs2 API, making folding practical on a single GPU. Handles single chains (monomer) and complexes via AlphaFold2-Multimer (Evans et al. 2021), and reports per-residue and per-interface confidence metrics so you know which parts of a prediction to trust.

This skill runs folding and returns structures + confidence. To retrieve an already-computed AlphaFold prediction for a known UniProt entry without running anything, use alterlab-alphafold-db instead.

When to Use This Skill

Use this skill when the user wants to:

  • Fold a protein sequence (FASTA) into a predicted 3D structure (PDB/mmCIF).
  • Predict a protein complex (AF2-Multimer) and score the interface (ipTM).
  • Rank multiple models and read confidence (pLDDT, pTM, PAE) to judge reliability.
  • Validate a designed sequence by refolding it and checking self-consistency vs. a target.

Does NOT Trigger

ScenarioUse instead
Co-fold a protein with a ligand (SMILES/CCD) or predict binding affinityalterlab-boltz
Antibody–antigen / arbitrary multi-entity complex from one FASTAalterlab-chai
Look up a precomputed AlphaFold model by UniProt idalterlab-alphafold-db
ESM embeddings, inverse folding, generative designalterlab-esm
Dock a ligand into an existing structurealterlab-diffdock
De-novo backbone generationalterlab-rfdiffusion

Core Capabilities

1. Monomer folding

# One sequence per FASTA record; MSAs via the hosted MMseqs2 API (--msa-mode)
colabfold_batch input.fasta out/ --num-models 5 --num-recycle 3

Outputs per record: ranked *_relaxed_rank_001_*.pdb, a JSON with plddt/pae, and coverage/pLDDT plots. TODO(verify) exact flag names against your installed ColabFold.

2. Complex folding (AF2-Multimer)

Join chains with a colon in one FASTA record to fold a complex:

>my_complex
MKT...AAA:MSE...GGG
colabfold_batch complex.fasta out/ --model-type alphafold2_multimer_v3

Read ipTM (interface confidence) and the inter-chain PAE block to judge whether the predicted interface is meaningful, not just the intra-chain pLDDT.

3. Confidence and validation

MetricReads
pLDDT (0–100, per residue)local confidence; <50 = likely disordered/unreliable
pTMglobal fold confidence
ipTMinterface confidence (complexes) — the number that matters for binding
PAEexpected positional error between residue pairs; low off-diagonal = confident relative orientation

Self-consistency check (validating a design): fold the candidate, then compare to the intended backbone (e.g. TM-score / RMSD). A design that folds back to its target with high pLDDT and low PAE is self-consistent — the standard acceptance gate in a design→fold→score loop (see alterlab-proteinmpnn, alterlab-rfdiffusion).

4. Running on a GPU

Folding needs a CUDA GPU. For anything beyond a quick monomer, dispatch through alterlab-remote-compute (SLURM or a managed GPU provider): submit colabfold_batch, poll to completion, and harvest out/.

Resources

  • references/colabfold_usage.md — install/pinning, MSA modes (API vs. local DB), templates, relaxation, batch/array runs, and full metric interpretation. Loaded on demand.

Part of the AlterLab Academic Skills suite.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.