A singular multi-omic measurement of gene regulatory mechanisms
Alexander Krohannon, Mansi Srivastava, Neel Sangani, Sarath Chandra Janga
Department of Biomedical Engineering and Informatics, Luddy School of Informatics, Computing, and Engineering, Indiana University Indianapolis
Corresponding Authors: amkrug@iu.edu, scjanga@iu.edu
This repository contains the code and reference data used to compute and analyze the Regulation Ratio (RR) — a normalized, gene-level metric that quantifies the relative contributions of transcriptional and post-transcriptional regulatory mechanisms by integrating ATAC-seq, RNA-seq, and Protein Occupancy Profiling sequencing (POP-seq) data.
The framework was applied across three human cell lines — HEK293, HepG2, and K562 — revealing three distinct regulatory classes: predominantly transcriptionally regulated, predominantly post-transcriptionally regulated, and balanced genes.
Raw and processed sequencing data are accessible at the Sequence Read Archive (SRA) under ID: 16057836.
Gene expression in eukaryotic cells is governed by regulatory mechanisms operating at multiple levels. This project focuses on two major levels:
- Transcriptional regulation — quantified using ATAC-seq peaks mapping chromatin accessibility within an expanded gene boundary (5,000 bp upstream and 500 bp downstream), normalized by gene length in kilobases.
- Post-transcriptional regulation — quantified using POP-seq peaks mapping protein-RNA interaction sites within gene boundaries, normalized by gene length in kilobases.
The Regulation Ratio is defined as:
RR = (# Significant POP-seq peaks) / (# Significant ATAC-seq peaks)
Genes with RR > 1 are classified as predominantly post-transcriptionally regulated, RR < 1 as predominantly transcriptionally regulated, and RR = 1 as balanced.
- Post-transcriptional regulation exhibited a >4-fold higher median density compared to transcriptional regulation in HEK293 cells, while HepG2 and K562 (both cancer-derived) showed the opposite, with ~2.5-fold higher transcriptional regulation.
- A strong correlation between transcriptional and post-transcriptional regulation was observed across all three cell lines (Pearson R² ≈ 0.60), suggesting coordinated rather than independent operation.
- Transcriptionally regulated genes tend to be significantly longer (median ~54,000 bp), while post-transcriptionally regulated genes have more isoforms and higher transcript abundance (median FPKM: 310 vs. 34 vs. 6.75 for post-transcriptional, transcriptional, and balanced, respectively).
- 55.8% of genes maintained identical regulatory classification across all three cell lines, with enriched functional pathways showing consistent directionality — suggesting regulatory strategy is a fundamental property of gene function, not just cellular context.
- Proliferation-associated pathways (Myc Targets, E2F Targets, G2M Checkpoint, mTORC1 Signaling) showed consistent post-transcriptional enrichment across all three cell lines, while developmental and tissue-identity pathways (Hedgehog Signaling, Epithelial Mesenchymal Transition, Notch Signaling) were consistently transcriptionally enriched.
Regulation_Ratio/
├── Regulation_Count.py # Computes the Regulation Ratio profile for a cell line
├── Profile_Analyzer.py # Analyzes and visualizes a computed profile
├── Gene_Information.txt # Ensembl gene annotations (GRCh38)
├── Gene_Type_Data.txt # Ensembl transcript type annotations
├── Protein_Information.txt # UniProt protein annotations for DepMap integration
├── gene_regulation.yml # Conda environment specification
├── HEK293/ # Precomputed profiles and results for HEK293
├── HepG2/ # Precomputed profiles and results for HepG2
└── K562/ # Precomputed profiles and results for K562
It is recommended to use the provided Conda environment:
conda env create -f gene_regulation.yml
conda activate gene_regulationThe environment includes all required dependencies: numpy, pandas, scipy, matplotlib, seaborn, statannot, cliffs_delta, and more.
This script takes ATAC-seq and POP-seq peak files for a given cell line and computes per-gene regulatory metrics, including the Regulation Ratio and regulatory group classification. Optionally, isoform counts, RNA-seq expression, and proteomics data can be incorporated.
Required inputs:
| Argument | Description |
|---|---|
cell_type |
Cell line name (e.g., HEK293) |
atac_file |
Path to ATAC-seq peak file (tab-separated BED-like format) |
pop_file |
Path to POP-seq peak file (tab-separated BED-like format) |
Optional inputs:
| Argument | Description |
|---|---|
--iso_file |
Path to StringTie GTF file for isoform counting |
--rna_file |
Path to Ballgown CSV for transcript abundance |
--prot_file |
Path to DepMap proteomics CSV (default: DepMap_Proteomics.csv) |
--upstream |
Bases upstream of gene boundary for ATAC peak detection (default: 5000) |
--downstream |
Bases downstream of gene boundary for ATAC peak detection (default: 500) |
Example:
python Regulation_Count.py HEK293 HEK293_ATAC_peaks.bed HEK293_POP_peaks.bed \
--iso_file HEK293_isoforms.gtf \
--rna_file HEK293_rna.csv \
--prot_file DepMap_Proteomics.csvOutput: A CSV file named <cell_type>_Regulation_Profile.csv (or <cell_type>_Regulation_<upstream>.csv if --upstream is specified), containing per-gene regulatory metrics and group classifications.
Peak file format: Tab-separated with at least three columns — chromosome, peak start, and peak end (no chr prefix in the chromosome column). This matches standard output from MACS3 (ATAC-seq) and Piranha (POP-seq).
This script takes a profile CSV generated by Regulation_Count.py and produces a suite of statistical comparisons and publication-quality figures.
Required inputs:
| Argument | Description |
|---|---|
profile |
Path to the profile CSV file |
output |
A label used in output file names and plot titles |
Example:
python Profile_Analyzer.py HEK293_Piranha_Profile.csv HEK293Outputs:
<output>_Regulation_Density.png— Box plot comparing transcriptional vs. post-transcriptional regulatory densities across all genes<output>_Regulation_Ratio.png— Scatter plot of regulatory densities colored by regulatory group (Regulation Ratio correlation)<output>_Gene_Length.png— Violin plot of gene lengths across regulatory groups<output>_Isoform_Count.png— Violin plot of isoform counts across regulatory groups<output>_Transcript_Abundance.png— Violin plot of transcript abundance (FPKM) across regulatory groups<output>_Protein_Expression.png— Violin plot of protein expression levels across regulatory groups<output>_Categorical.png— Bar chart of transcript type distribution and enrichment across regulatory groups<output> Transcript Type Enrichment.csv— Fisher's exact test results for transcript type enrichment
Note:
Profile_Analyzer.pyexpectsGene_Type_Data.txtto be present in the working directory for transcript type enrichment analysis.
The peak files used as input for Regulation_Count.py should be generated using the following tools, consistent with the methods in the associated publication:
- ATAC-seq peaks: Called with MACS3 using a shift of 50 bp and extension of 100 bp, with FDR correction (p < 0.05). Biological replicates should be pooled prior to peak calling.
- POP-seq peaks: Called with Piranha (v1.2.1) using a bin size of 20 bp, no normalization, and logarithmic covariates, with FDR correction (p < 0.05). Biological replicates should be pooled prior to peak calling.
- All data should be aligned to GRCh38/hg38 using HISAT2.
If you use this code or the Regulation Ratio framework in your work, please cite:
Krohannon A, Srivastava M, Sangani N, Janga SC. Regulation Ratio: A singular multi-omic measurement of gene regulatory mechanisms. (2025).
This work was supported by the National Institute of General Medical Sciences under Award Number R01GM123314, a Lilly Research Award Program (LRAP) grant, and the IUI Luddy School of Informatics, Computing and Engineering.
This project is licensed under the GPL-3.0 License. See LICENSE for details.