MarginalFootprints

This describes the MarginalFootprints command used to generate bias-corrected TF-binding footprints.

The function includes the chrombpnet footprints algorithm, which conducts the following steps:

• Insert a motif sequences center in the center of background regions to make synthetic sequences
• Find the profile probability predictions for the synthetic sequences of this given motif
• Find the predictions for the reverse complement of the given sequences and then reverse the predictions
• Average the footprints generated to get the marginal footprint of one motif.

For all marginal footprint analysis, the chrombpnet_nobias.h5 model is used as it generates the bias-corrected footprints.

Usage

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

for cell in "${cell_types_list[@]}"
do
     MarginalFootprints \
           "$genome" \
           "$fold_id" \
           "$motif_sequences"
           "$output_directory" \
           "$cell" \
           "$batch_id"
done

Other functions written for this research include:

Marginal_footprints_young            # Function for the mm10 neonatal cohort
Marginal_footprints_experiment       # Function for any bulk ATAC-seq models 
Marginal_footprints_clusters         # Function for the mm10 adult cohort - downstream PCA analysis specific
Marginal_footprints_young_clusters   # Function for the mm10 neonatal cohort - downstream PCA analysis specific

These all require the same inputs, but differ within the function as they required different peak file paths or output directory names, which are pre-determined in the function.

Input format

genome             # Genome name (e.g., "mm10")   
fold_id            # Fold used (e.g., "fold_0")
motif_sequences    # Path to a tsv file containing motif names in the first column and motif sequence string in the second column
output_directory   # Path to the output directory of the selected cohort for analysis (e.g., ~/ChromBPNet/outputs/mouse)
cell               # Cell name from cell_types_list - leave as "$cell"
batch_id           # Name for the group of motifs within the motif_sequences file (e.g., "SOX_motifs" or "PCA_analysis")

Motifs used throughout the analysis are available in the metadata directory of the repository.

• motif_sequences.tsv is a list of all 2059 motifs from the JASPAR_CORE_2026_non-redundant.meme database. This list was generated using a pipeline to convert the JASPAR PFMs to motif sequences.

This function can be found in Downstream_Analyses/marginal_footprint_analysis/1_MF_file_interrogation.ipynb under the section 'Convert JASPAR PFMs to motif sequences'.

• PCA_motif_sequences.tsv is a filtered list of 733 for downstream PCA analysis.

The filtering is explained and evidenced in the Downstream_Analyses/other/TF_expression_PCA.qmd notebook.

Output format

All output files will be found in ~/ChromBPNet/outputs/$cohort/$cell_type/marginal_footprints/ or /marginal_footprints_clusters/ depending on the above function used.

{cell}_{batch_id}_footprints.h5   # File format including all motif sequence predicted footprint profiles
*.png                             # PNG images of the marginal footprints generated

Downstream analyses of the marginal footprints and explanation of the file contents can be found in Downstream_analyses/marginal_footprint_analysis