HEAL (Hierarchical Estimate from Agnostic Learning)

• Machine learning-based genome analysis and risk prediction framework.

supported file type

• csv file (Mutation burden matrix)
  • row: sample ID, column: gene name

requirements

• python3
• pandas
• numpy
• scikit-learn
• scipy

How to run

Step0 Prepare mutation burden matrix.

1. Annotate VCF file of whole exome or genome sequencing data with gene name, deleteriousness score, and allele frequency info.
2. Preprosess annotated genotype data to calculate mutation burden. Sample mutation burden file is available in toy_data.

Step 1 Run the HEAL script

1. Input file: Mutation burden matrix.

Step 2 The model outputs

1. Disease gene lists.
2. Genetic risk prediction model.
3. Prediction performance summary.

Usage

Run the HEAL script from the command line with the following arguments:

python HEAL.py --file_path <path_to_input_file> [options]

Command-line Arguments

Argument	Type	Required	Default	Description
--file_path	str	Yes	-	Full path to the input file
--output	str	No	Current working directory	Output path
--splits	int	No	5	Number of splits for cross-validation
--trials	int	No	1	Number of trials to run
--l1	float	No	1.0	Lower bound of lambda candidates
--l2	float	No	40.0	Upper bound of lambda candidates
--lfidelity	int	No	5	Fidelity of linspace of lambda candidates
--scoring	str	No	'roc_auc'	Scoring metric to maximize
--random_state	int	No	42	Random state to start from
--tts	bool	No	False	Use train_test_split instead of StratifiedKFold for outer CV

Citation

Please cite the following paper

• Hirotaka Ieki, Kaoru Ito, Sai Zhang, Satoshi Koyama, Martin Kjellberg, Hiroki Yoshida, Ryo Kurosawa, Hiroshi Matsunaga, Kazuo Miyazawa, Nobuyuki Enzan, Changhoon Kim, Jeong-Sun Seo, Koichiro Higasa, Kouichi Ozaki, Yoshihiro Onouchi, Koichi Matsuda, Yoichiro Kamatani, Chikashi Terao, Fumihiko Matsuda, Michael Snyder, Issei Komuro "Machine Learning Reveals the Contribution of Rare Genetic Variants and Enhances Risk Prediction for Coronary Artery Disease in the Japanese Population" medRxiv 2024 doi.org/10.1101/2024.08.13.24311909