cdm_mmseqs2

CTS (CDM Task Service) job wrapper for MMseqs2 sequence clustering and search.

Container

Wraps the official soedinglab/mmseqs2 image. Published to ghcr.io/kbaseincubator/cdm_mmseqs2.

Entrypoint: mmseqs (no subcommand) — append the desired subcommand as the first argument (e.g. easy-cluster, easy-search, easy-linclust).

Refdata: Not required for clustering modes (easy-cluster, easy-linclust). Required for search modes (easy-search) which need a pre-built reference database.

Usage via CTS

See the CTS documentation and the demo notebook at global_share/jplfaria/mmseqs2_demo.ipynb on hub.berdl.kbase.us.

Example: easy-cluster (all-vs-all, no refdata)

job = tscli.submit_job(
    "ghcr.io/kbaseincubator/cdm_mmseqs2:0.1.0@sha256:24afa107c0dac1a6f093cc081ec19a577c4d0b156dfa8ee66bb2d3c41b97c082",
    input_files,
    "cts/io/jplfaria/output/mmseqs2/test/v1",
    cluster="kbase",
    declobber=True,
    output_mount_point="/out",
    args=[
        "easy-cluster",
        tscli.insert_files(),
        "/out/cluster_results",
        "/out/tmp",
        "--threads", "8",
    ],
    cpus=8,
    memory="32GB",
    runtime="PT30M"
)

Output

• cluster_results_cluster.tsv — representative → member mapping (imported to Delta Lake)
• cluster_results_rep_seq.fasta — representative sequences
• cluster_results_all_seqs.fasta — all sequences