CADA: The Next Generation

This is a re-implementation of the CADA method for phenotype-similarity prioritization.

• Free software: MIT license
• Documentation: https://cada-prio.readthedocs.io/en/latest/
• Discussion Forum: https://github.com/bihealth/cada-prio/discussions
• Bug Reports: https://github.com/bihealth/cada-prio/issues

Running Hyperparameter Tuning

Install with tune feature enabled:

pip install cada-prio[tune]

Run tuning, e.g., on the "classic" model. Thanks to optuna, you can run this in parallel as long as the database is shared. Each run will use 4 CPUs in the example below and perform 1 trial.

cada-prio tune run-optuna \
    sqlite:///local_data/cada-tune.sqlite \
    --path-hgnc-json data/classic/hgnc_complete_set.json \
    --path-hpo-genes-to-phenotype data/classic/genes_to_phenotype.all_source_all_freqs_etc.txt \
    --path-hpo-obo data/classic/hp.obo \
    --path-clinvar-phenotype-links data/classic/cases_train.jsonl \
    --path-validation-links data/classic/cases_validate.jsonl \
    --n-trials 1 \
    --cpus=4

Managing GitHub Project with Terraform

# export GITHUB_OWNER=bihealth
# export GITHUB_TOKEN=ghp_<thetoken>

# cd utils/terraform

# terraform init
# terraform import github_repository.cada-prio cada-prio
# terraform validate
# terraform fmt
# terraform plan
# terraform apply

Changelog

0.7.0 (2024-08-29)

Features

• command to dump OpenAPI schema (#58) (#59) (317a767)

0.6.1 (2023-11-16)

Bug Fixes

• pinning python to 3.11 for build so we have setuptools (#36) (54d4e8c)

0.6.0 (2023-11-16)

Features

• adding API prefix, OpenAPI and docs to REST server (#35) (1a2f605)
• adding classic and current model (#25) (44ddf24)

0.5.0 (2023-09-18)

Features

• adding "tune run-optuna" command (#23) (6cc753b)
• re-useable implementation of "tune train-eval" (#21) (c80c4bf)

0.4.0 (2023-09-14)

Features

• adding dump-graph to cli (#18) (3aace31)
• adding param-opt command with single parameter evaluation (#20) (83141c6)
• allow running with legacy model/graph data (#16) (9d3cc7c)
• embedding parameters can be provided via CLI and contains seeds (#19) (bbd5d86)

0.3.1 (2023-09-13)

Bug Fixes

• add missing line endings to hgnc_info.jsonl (#13) (aa14b9b)
• properly parsing comma-separated list on REST API (#14) (97fdfee)

0.3.0 (2023-09-11)

Features

• also adding gene-to-phen edges from HPO (#9) (d5a8337)

0.2.1 (2023-09-08)

Bug Fixes

• removing spurious debug print statement (#7) (98e7443)

0.2.0 (2023-09-08)

Features

• gene to phenotype links file can be gziped (#5) (66c48bf)

0.1.0 (2023-09-07)

Features

• adding REST API server for prediction (#4) (8bb7516)
• initial training implementation (#1) (10d3a7c)
• prioritization prediction with model (#3) (48d504c)