Skip to content

Latest commit

 

History

History
153 lines (99 loc) · 3.59 KB

README.md

File metadata and controls

153 lines (99 loc) · 3.59 KB
Raw

Workflow

Calculate single-nt resolution gwRVIS

config_file="conf/topmed/config.single_nt_gwrvis.yaml"
./submit_single_nt_gwrvis.sh $config_file

sbatch ./compile_gwrvis_single_nt_scores.sh $config_file

sbatch ./create_tabix_indexes.sh $config_file

Plot distributions of single-nt gwRVIS across CCDS or all genomic classes

cd modules/ad-hoc/

Create single-nt feature table (for calculating the genome-wide single-nt jarvis):

sbatch ./gwrvis_core/for_prediction_annotate_feat_table_w_mut_exl_genomic_class.sh conf/topmed/config.NEW_genome_wide_scores.yaml static_files/input_classes.txt

Calculate single-nt JARVIS

for i in `seq 1 22`; do sbatch -o chr${i}.jarvis ./jarvis/deep_learn_raw_seq/submit_prediction_per_chr.sh conf/topmed/config.NEW_genome_wide_scores.yaml both $i; done

JARVIS Training

# example config file: config.NEW_ClinVar_pathogenic.yaml

python jarvis/deep_learn_raw_seq/prepare_data.py conf/topmed/config.NEW_ClinVar_pathogenic.yaml

./submit_all_jarvis_jobs.sh conf/topmed/config.NEW_ClinVar_pathogenic.yaml 1

JARVIS prediction on test set

  • Main trained model:
# example config file: config.NEW_ncER-GWAS-testing.yaml
python jarvis/deep_learn_raw_seq/prepare_data.py conf/topmed/config.NEW_ncER-GWAS-testing.yaml

python -u jarvis/deep_learn_raw_seq/test_jarvis_model.py conf/topmed/config.NEW_ncER-GWAS-testing.yaml both intergenic,utr,lincrna,ucne,vista 1 1

Feature table for Machine Learning / Deep Learning

This is compiled in gwrvis_core/full_table_intersect_regulatory_features.sh

The final feature table (with genomic classes, clinvar annotation and all associated features) is saved into: clinvar_feature_tables/full_feature_table.[pathogenic_set]_[benign_set].bed

  • phastCons_primate needs to be imputed with the median in:
    • jarvis/variant_classification/run_variant_classification.py
    • jarvis/deep_learn_raw_seq/prepare_data.py

Master script:

Main config file:

conf/topmed/config.NEW_ClinVar_pathogenic.yaml

# all
[sbatch] ./wgs.sh config.yaml input_classes.txt


# coding
[sbatch] ./wgs.sh config.coding.yaml input_classes.coding.txt

Run wgs.sh for multiple values of MAF, win_len and variant type (SNVs, INDELs or both):

./submit_all_jobs.sh

More analytically:

  • Read input parameters:
config_log=config.yaml;
input_classes=input_classes.txt;

  • Record features across fixed and tiled genomic windows (e.g. common/all variants, mut. rate, CpG islands, GC content, etc.):
./parse_all_chromosomes.sh $config_log;

  • Perform logistic regression (common ~ all variants) to get gwRVIS scores:
python run_full_regression.py $config_log;

  • Convert window indexes (0-based) to real genomic coordinates:
python convert_window_indexes_to_genomic_coords.py $config_log;

  • Get gwRVIS distribution by genomic class:
./run_gwrvis_extraction_by_genomic_class.sh $config_log $input_classes;

  • Aggregate gwRVIS scores from all chromosomes:
python aggregate_gwrvis_scores.py $config_log $input_classes;

  • Get gwRVIS distribution per genomic class across the entire genome:
python get_whole_genome_rvis_distr.py $config_log $input_classes;

  • Make refined density plots and boxplots for gwRVIS distribution per genomic class with ggrdiges:
python make_ggridges_plots.py -c $config_log;