Prediction of Pioneer Transcription Factors Using Chromatin Accessibility and Gene Expression

Nae-Chyun Chen, Arun Das, Xinyu Feng

Project Overview

Our project has been coded and run using Python 3.X, on both our personal laptops (OSX) and on MARCC. All code has been commented to indicate its function. Required packages can be found in requirements.txt.

Step-by-step walkthrough of our project

Download the data:

sh experiments/download_data.sh

Convert Gene IDs

Convert JASPAR Gene IDs into Ensembl IDs:

python3 convert_gene_id.py human_pwm_ids_sorted.txt converted_id.csv

Get expression levels

Get expression levels from the replicates:

cd experiments
python3 ../scripts/get_expression_levels.py -c ../converted_id.csv -l list.csv -o expr.tsv
cd ..

Generating expression weighted training data:

build_matrix.py K562_rep1_all_matrix.tsv expression_levels.csv .
build_matrix.py ENCFF342EGB_all_matrix.tsv expression_levels.csv .

(K562_rep1_all_matrix.tsv and ENCFF342EGB_all_matrix.tsv are generated in an outside project and are available upon request)

Train a logistic classifier

./11-classifier2.py -r ENCFF342EGB_ENCFF305QBE_TPM_matrix.tsv 56659 ../human_pwm_ids_sorted.txt lr ENCFF342EGB_ENCFF305QBE_TPM_lr_scaled

./11-classifier2.py -r ENCFF342EGB_ENCFF853TRI_TPM_matrix.tsv 56659 ../human_pwm_ids_sorted.txt lr ENCFF342EGB_ENCFF853TRI_TPM_lr_scaled

./11-classifier2.py -r K562_rep1_ENCFF285HUZ_TPM_matrix.tsv 263576 ../human_pwm_ids_sorted.txt lr K562_rep1_ENCFF285HUZ_TPM_lr_scaled

./11-classifier2.py -r K562_rep1_ENCFF297CNO_TPM_matrix.tsv 263576 ../human_pwm_ids_sorted.txt lr K562_rep1_ENCFF297CNO_TPM_lr_scaled

Significance Analysis

Get the corrected p-values of each of the features:

python3 GWAS_selection.py K562_rep1_labels.csv K562_rep1_ENCFF285HUZ_TPM_matrix.tsv K562_rep1_ENCFF297CNO_TPM_matrix.tsv

python3 GWAS_selection.py ENCFF342EGB_labels.csv ENCFF342EGB_ENCFF305QBE_TPM_matrix.tsv ENCFF342EGB_ENCFF853TRI_TPM_matrix.tsv

Additional Files

All *TPM_matrix.tsv files are availabe here: https://livejohnshopkins-my.sharepoint.com/:f:/g/personal/xfeng17_jh_edu/EhCj3Tdk4NBEhquIVJYfaioB9pPZcaZQU76YgosuKh5sDg?e=onIubH

Plotting

Plotting the expression levels (Figure 2a):

python3 plot_expression.py expression_levels.csv

Plotting the expression heat map:

python3 plot_expression_heatmap.py expression_levels.csv

Plotting the correlation heatmap (Figure 2b):

python3 plot_correlation_heatmap.py ENCFF342EGB_ENCFF853TRI_TPM_matrix.tsv

Plotting p-values (Figure 3a):

python3 plot_pvalues.py GWAS_pvalues_ENCFF342EGB_ENCFF853TRI.csv

Plotting coefficients:

python3 plot_coeff.py ENCFF342EGB_ENCFF853TRI_TPM_lr_ranks.tsv

Plotting coefficients across cell lines:

python3 plot_coeff_cell_lines.py ENCFF342EGB_ENCFF305QBE_TPM_matrix_weighted_counts_only_lr_ranks.tsv

Please look at plot_roc.py to see the code used to plot the ROC curve shown in the final report.

milkschen/openchromatinprediction