Tags

DMZ does not contain H100 H200 partions, removed.

Refactor workflow: split MSA/inference stages, add dynamic routing by sequence length, add H200 partition support

The workflow `workflow/main.nf` has four processes:

- **parameters**: Parse the input JSON and count total residues
- **alphafold3** / **msa_stage + inference_stage**: Run structure prediction (see below)
- **cif_to_pdb**: Convert AlphaFold3 output from .cif to .pdb for downstream visualization

uncontainerized version of alphafold3/3.0.1

containerized version with module load alphafold3/3.0.1

fix bash file in cif_to_pdb

After migrating from rhel7 to rhel9.4, inside of the process cif_to_pdb, pymol/2.5 is renamed as py-pymol/2.5.0 on rhel9.

Fixed error: cannot run multiple files simultaneously
Vizapp containerized + cif2pdb

test passing multiple files

Fixed error: cannot run multiple files simultaneously
Vizapp containerized + cif2pdb

Vizapp containerized + cif2pdb

The vizapp part is containerized

make cif_to_pdb into folder channel instead of files channel

This version removes cif_to_pdb process, makes vizapp available to read in cif files.

remove cif2pbd

vizapp

results_ch acts as a trigger

When alphafold3 finishes, it emits only a log file to results_ch.

We use this completion event to delay checking for CIF files.

map {} introduces a delay before searching for CIF files

CIF files might take a second to appear after AlphaFold3 finishes.

sleep(5) ensures that the directory is fully written before checking.

Channel.fromPath(..., checkIfExists: true) ensures Nextflow waits

The checkIfExists: true option makes Nextflow wait until files appear.

.flatten() ensures cif_files_ch is a single channel

Since Channel.fromPath() returns a new channel inside map {}, we use .flatten() to merge it into a single stream.

test cif_to_pdb connection with af3

published version doesn't run cif_to_pdb