alphafold2

1# AlphaFold2 (ColabFold runner)
2 
3This skill wraps AlphaFold2 and AlphaFold2-Multimer through `colabfold_batch`,
4which replaces DeepMind's local-database MSA pipeline with a call to the public
5MMseqs2 server — so a prediction is one command and one FASTA, not a 2 TB
6database mount. AF2 remains the reference monomer predictor and the multimer
7model is still a strong protein–protein validator, but it does not handle
8ligands or nucleic acids; for those, route to `boltz`, `chai1`, or `openfold3`.
9The ColabFold code is MIT (github.com/sokrypton/ColabFold) and the AlphaFold2
10code is Apache-2.0 (github.com/google-deepmind/alphafold); the AF2 model
11parameters are CC-BY-4.0 with DeepMind's terms of use.
12 
13## Running it
14 
15```bash
16colabfold_batch input.fasta out \
17  --num-recycle 3 \
18  --model-type alphafold2_multimer_v3
19```
20 
21The input is a plain FASTA. For a complex, put every chain on one sequence
22line separated by `:` — `colabfold_batch` builds a paired MSA per segment and
23runs the multimer model when it sees the colon (so the explicit `--model-type
24alphafold2_multimer_v3` above is belt-and-braces). For monomers omit
25`--model-type` and the colon. `--templates` and `--amber` add PDB templates
26and OpenMM relaxation respectively; both are off by default and both add
27minutes per model.
28 
29ColabFold runs all five AF2 model weights by default and ranks them by pLDDT
30(pTM/ipTM for multimer), so output per query lands in `out/` as five ranked
31…