ContribsBW

Here we discuss the ContribsBW command, which generates sequence contribution scores for either or both of the ChromBPNet heads (the profile head - which predicts the shape of a profile, or the counts head - that predicts the total counts in a profile).

This function integrates the chrombpnet contribs_bw command, that leverages the DeepLIFT algorithm to infer the contribution scores. For a more detailed explanation of the function use, please refer to the ChromBPNet repository.

Usage

cell_type_list=("Lactotrophs" "Gonadotrophs" "Thyrotrophs" "Somatotrophs" "Melanotrophs" "Corticotrophs" "Stem_cells")

for cell in ${cell_type_list[@]}
do
ContribsBW \
        "$model_directory" \
        "$cell" \
        "$genome" \
        "$peaks_file" \
        "$output_directory
done

Input format

Required arguments:
model_directory      # Path to the cohort model directory (e.g., ~/ChromBPNet/Models/mouse)
cell                 # Cell type name identified from cell_type_list - to be left as "$cell"
genome               # Genome name (e.g., "mm10", "rn6", "hg38")
peaks_file           # Path to the consensus peak file - must be in .bed format
output_directory     # Path to output cohort directory (e.g., ~/ChromBPNet/outputs/mouse)

Output format

All output files will be found in ~/ChromBPNet/outputs/$cohort/$cell_type/contribs_bw/

${cell}_contribs.counts_scores.bw           # Contribution count scores stored in BigWig format - can be visualised on IGV
${cell}_contribs.counts_scores.h5           # File containing a dictionary with contribution scores for the count head of the model
${cell}_contribs.profile_scores.bw          # Contribution profile scores stored in BigWig format - can be visualised on IGV
${cell}_contribs.profile_scores.h5          # File containing a dictionary with contribution scores for the profile head of the model
${cell}_contribs.interpreted_regions.bed    # BED file with all regions used for interpretations - same peaks as input peaks_file
${cell}_contribs.interpret.args.json        # JSON file showing paths used within the function

Downstream analyses of the contribution scores and explanation of the file contents can be found in Downstream_analyses/contribution_score_analysis