Using the devtools package, you can install directly from GitHub:
devtools::install_github("blueprint-genetics/amiss")
The framework can be configured using JSON files that determine which preprocessing and imputation steps are taken.
The file combination_orig.json in the repository contains parameters that largely match the original code of the manuscript. combination_minimal.json produces the smallest dataset for processing and uses only zero imputation, and thus should be quickest to run.
For different options available for parameter values, see parameter_grid.json.
To run the framework with a single set of parameters up to computation of the classification statistics (note: this may take a long time):
library(amiss)
# Parse
S01_parse_vcf("clinvar_20190624.vep.vcf", cadd_snv_filename = "CADD_clingen.tsv", cadd_indel_filename = "CADD_clingen_indel.tsv", output_root_dir = "output/data", parameters_path = "combination_minimal.json")
# Preprocess
S02_preprocess_data("output/data/quality-clingen.restriction-missense.transcript-canonical", parameters_path = "combination_minimal.json", output_path = "output/data", seed=10)
# Imputation and classifier training
impute_and_train(training_path = "output/data/preprocessed_training_data.csv",
outcome_path = "output/data/training_outcomes.csv",
output_path = "output/trained",
mice_hyperparameter_grids = mice_hyperparameter_grids,
other_hyperparameter_grids = other_hyperparameter_grids,
single_value_imputation_hyperparameter_grids = single_value_imputation_hyperparameter_grids,
parameter_list=rjson::fromJSON(file = "combination_minimal.json"),
cores = 1, seed = 10, lean = TRUE)
# Prediction
predict_on_test_set(test_path = "output/data/preprocessed_test_data.csv",
outcome_path = "output/data/test_outcomes.csv",
tr_output_path = "output/trained",
results_dir_path = "output/results",
seed = 10)
# Result CSVs now in output/results
This project compares methods for handling missing data in variant annotations for the purpose of building variant pathogenicity predictors.
See the paper for a more detailed description.
The result files are available in Zenodo with DOI 10.5281/zenodo.6656616.
This project conforms to the principles of open science:
- Open data:
- We use and reference publically available datasets and will citably archive any data and code we produce using Zenodo with a DOI
- Open source:
- The source code is freely available under the MIT license at GitHub
- Open notebook:
- You can follow development from the start on GitHub at https://github.com/blueprint-genetics/amiss
- The research plan is available in the GitHub repository (see section above for producing a PDF version)
- Open access:
- We will upload a preprint of the resulting paper(s) on biorXiv
- We will submit the results for publication in a peer-reviewed open access journal
- Open communication:
- We intend to present the results in public scientific conferences
- Open collaboration:
We license our code with the MIT license (see the LICENSE file), but note that the license of the entire system may depend on the licenses of individual libraries.