dukegcb/predict-tf-preference repository overview

⁠Predict-TF-Preference

R scripts to predict preferences between two TFs using data from predict-tf-binding

⁠Dependencies

1. R (Included Dockerfile provides r-base, which is adequate)
2. Python 2.7 for filtering NegCtrl (Included Dockerfile installs python 2.7)

⁠Calculating Preference scores

This script is designed to extract binding probabilities out of BED files generated by Predict-TF-Binding⁠.

./predict-tf-preference.R tf1 tf2 tf1.bed tf2.bed prefs.bed

Where:

• tf1 is the name of the first transcription factor (e.g. E2f1)
• tf2 is the name of the second transcription factor (e.g. E2f4)
• tf1.bed is a bed file containing prediction scores for the first TF
• tf2.bed is a bed file containing prediction scores for the second TF
• prefs.bed is the name of the output file you'd like to write.

⁠Filtering for NegCtrl

This script is designed to remove preference scores favoring a TF where the binding prediction was below a NegCtrl threshold

./filter-preference-threshold.py --spaces tf1.bed tf2.bed prefs.bed tf1_threshold tf2_threshold

Where:

• tf1.bed is a bed file containing prediction scores for the first TF
• tf2.bed is a bed file containing prediction scores for the second TF
• tf1_threshold is the NegCtrl threshold value (float) for tf1
• tf2_threshold is the NegCtrl threshold value (float) for tf2
• prefs.bed is the preferences generated by predict-tf-preference.R

⁠Notes

• tf1.bed and tf2.bed must be the same length
• Only certain combinations are valid. See predict-tf-preference.R⁠ for details.
• script assumes that the first 3 columns of tf1.bed and tf2.bed (chrom, start, stop) are identical, and only difference will be the predictions in the 4th column
• Output file is generated by reusing the first 3 columns of tf1.bed and replacing score column with the calculated preference.