DIALECT

Driver Interactions and Latent Exclusivity or Co-occurrence in Tumors

DIALECT identifies mutually exclusive (ME) and co-occurring (CO) driver mutation pairs by modeling each somatic count as passenger background + latent driver, conditioned on a background mutation rate (BMR).

The commands below require a Git checkout installed in editable mode. This is necessary for the default CBaSE provider: it resolves DIALECT's tracked CBaSE fork under external/CBaSE/. The configured wheel contains the installable dialect package, CLI metadata, and license. The configured source distribution contains the source package, tests, and selected README, license, and build metadata. Neither contains the external/CBaSE runtime scripts or auxiliary data.

The checkout includes the CBaSE fork and its NOTICE, but its large external/CBaSE/auxiliary/ directory is intentionally not tracked. Before running dialect generate --bmr cbase, provision a compatible auxiliary data set there and review the external/CBaSE/NOTICE, which records the upstream source and lineage caveat. DIALECT does not automate that acquisition. If you already have count_matrix.csv and bmr_pmfs.csv, an installed wheel can run dialect identify without CBaSE.

pip install -e .
dialect generate -m cohort.maf -o out/cohort
dialect identify -c out/cohort/count_matrix.csv -b out/cohort/bmr_pmfs.csv -o out/cohort -k 100

For development, install the test and lint dependencies with pip install -e ".[dev]".

For a zero-complete CBaSE cohort, provide the exact ordered sample axis as one unique, nonempty identifier per line. The optional scalar is an equality assertion; CBaSE's per-sample denominator is derived from the axis.

dialect generate -m cohort.maf -o out/cohort --bmr cbase \
  --cbase-sample-axis cohort_samples.txt --cbase-samples 137

from dialect import estimate_bmr, identify_interactions
estimate_bmr(
    "cohort.maf",
    "out/cohort",
    provider="cbase",
    sample_ids="cohort_samples.txt",
    n_samples=137,
)
result = identify_interactions(
    "out/cohort/count_matrix.csv", "out/cohort/bmr_pmfs.csv", "out/cohort", top_k=100
)

Agent / contributor context: see AGENTS.md.

Cite

@article{shuaibi2024dialect,
  author  = {Ahmed Shuaibi and Uthsav Chitra and Benjamin J. Raphael},
  title   = {A latent variable model for evaluating mutual exclusivity and
             co-occurrence between driver mutations in cancer},
  journal = {bioRxiv},
  year    = {2024},
  doi     = {10.1101/2024.04.24.590995}
}

BSD-3-Clause — see LICENSE.