MuAt

Mutation-Attention Model

Quick Start (Tested on Linux)

• Clone muat repository
• Go to muat repository
• Create conda environment

conda env create -f muat-conda.yml

• activate muat environment

conda activate muat

• Download genome reference to ./ref:

wget http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/technical/reference/phase2_reference_assembly_sequence/hs37d5.fa.gz -O ./ref/ref.gz
gunzip ./ref/ref.gz

• Make sure that preprocessing/dmm/annotate_mutations_with_bed.sh is permitted to be executed.

chmod 755 preprocessing/dmm/annotate_mutations_with_bed.sh

Predicting .vcf file (GRCh37) with MuAt pretrained model :

• Look at the example file in ./extfile/example_for_alldata_prediction_gz.tsv file. This lists all files which will be predicted by the model
• Run this code

python3 main.py --dataloader 'pcawg' --predict-filepath '/path/to/muat/extfile/example_for_alldata_prediction_gz.tsv' --reference-h19 '/path/to/muat/ref/ref' --load-ckpt-file '/path/to/muat/bestckpt/wgs/ensemble/finalpcawgFeaturefold1_11110_wpos_TripletPositionF_bs5000_nl2_nh2_ne256_cl3/new_weight.pthx' --output-pred-dir '/path/to/muat/data/raw/output/' --predict-all

• If it succeed, you can see all preprocessed files in /path/to/muat/data/raw/temp/
• and all prediction outputs in --output-pred-dir

Predicting .vcf file (GRCh38) : *add --convert-hg38-hg19 and --reference-h38

• Download genome reference GRCh38

wget http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/technical/reference/GRCh38_reference_genome/GRCh38_full_analysis_set_plus_decoy_hla.fa -O ./ref/ref_hg38

• add --convert-hg38-hg19 and --reference-h38 parser

python3 main.py --dataloader 'pcawg' --predict-filepath '/path/to/muat/extfile/example_for_alldata_prediction_gz.tsv' --reference-h19 '/path/to/muat/ref/ref' --load-ckpt-file '/path/to/muat/bestckpt/wgs/ensemble/finalpcawgFeaturefold1_11110_wpos_TripletPositionF_bs5000_nl2_nh2_ne256_cl3/new_weight.pthx' --output-pred-dir '/path/to/muat/data/raw/output/' --predict-all --convert-hg38-hg19 --reference-h38 '/path/to/muat/ref/ref_hg38'

DOWNLOAD PRETRAINED MODELS (CHECKPOINTS)

Whole Genome Sequence (trained on PCAWG) : for the best results, input type should be the same as pretrained models type

Model	Description	Link
SNV	Model trained on SNV input	Download
SNV+MNV	Model trained on SNV+MNV input	Download
SNV+MNV+indels	Model trained on SNV+MNV+indels input	Download
SNV+MNV+indels+SV/MEI	Model trained on SNV+MNV+indels+SV/MEI input	Download
SNV+pos	Model trained on SNV+pos input	Download
SNV+indel+pos	Model trained on SNV+indel+pos input	Download
SNV+MNV+indel+SV/MEI+pos	Model trained on SNV+MNV+indel+SV/MEI+pos input	Download

Whole Exome Sequence (trained on TCGA)

Model	Description	Link
SNV+MNV	Model trained on SNV+MNV input	Download

Training MuAt PCAWG

Read README_training.md