Dima 4d802be7d6 support both af2 and af3 data pipelines (#523) ...

* symmetrical refactoring to support both af2 and af3 data pipelines

* Clean tests

* Keep GPU tests in place

* Reverted accidentally deleted templates

* Add AlphaFold3 feature creation pipeline and per-chain input generation

- Implement `create_pipeline_af3` to construct the AlphaFold3 data pipeline with correct database and binary paths.
- Add `create_af3_individual_features` to generate AlphaFold3 input features for each chain in a FASTA, handling protein, RNA, and DNA sequences.
- Integrate new AF3 logic into the main entry point, dispatching to AF2 or AF3 as appropriate.
- Ensure output directory creation and error handling for missing dependencies or invalid sequences.

* Convert template dates to datetime for af3

* First check for nucleotides, then for amino-acids

* Skip existing features json if --skip_existing=true

* Check if DNA before RNA

* Bump 2.1.0

* Git ignore build/ dir

2025-07-16 12:30:18 +02:00

2.7 MiB

Raw History

View Raw