MAGPIE framework

Basic requirements

1. Python 3.9.12.

2. Python packages listed in requirements.

Use conda env create -f magpie.yml to create the conda environment is recommanded.

Additional requirements for training from denovo

3. SpliceAI (Use conda env create -f spliceai.yml to create the conda environment is recommanded).

4. MATLAB CLI.

5. AnnoVar (register required).

6. OMIM database(application required).

Usage

Reminder: running MAGPIE on single CPU may take some time because single process of autoFE.

Input format

MAGPIE supports variants in CSV format as input. The input file should contain at least 5 columns in the header as follows. Sample file

Chr	Start	End	Ref	Alt	...

Use pretrained MAGPIE model to predict variants

1. Install packages listed in requirements or use magpie conda environment.

Annotated variants

2. Run source magpie.sh --mode pred --test_file [filepath] --file_state annotated --visualization e.g. source magpie.sh --mode pred --test_file data/datasets/test.csv --file_state annotated --visualization

Unannotated variants

2. Run source magpie.sh --mode pred --test_file [filepath] --file_state unannotated --visualization e.g. source magpie.sh --mode pred --test_file data/datasets/test.csv --file_state unannotated --visualization

Results would be saved in data/result.

Train MAGPIE model from denovo

1. Download and decompress required database for annotating using bash download.sh.

2. Apply for AnnoVar access, and place all execuatable annotation tools in ./annovar.

3. Apply for OMIM database access, and place 'genemap2.txt' in data/annotation_database.

4. Install packages and pieces of software manually or use Dockerfile to create and run a docker image.

   docker build -t magpie .
   docker run -it magpie
   

5. Run source magpie.sh --mode train --input_file [filepath] e.g. source magpie.sh --mode train --input_file data/datasets/denovo.csv. Model would be saved in output/result/.

Technical support

Please feel free to contact me(yichengliu at zju.edu.cn) for technical support.