scE2G

A computational pipeline for predicting genome-wide enhancer-gene regulatory links from single-cell ATAC-seq or paired ATAC and RNA-seq (multiome) data.

Overview

Input: Single-cell ATAC-seq or paired ATAC and RNA-seq (multiome) data per cell cluster

Output: Genome-wide enhancer-gene regulatory link predictions per cell cluster

Pipeline components

1. ABC model predictions - Compute ABC model predictions for each cell cluster
2. E2G feature generation - Generate element-gene features from ABC predictions
3. Correlation analysis (multiome only) - Compute Kendall correlation and/or ARC-E2G score for each cell cluster
4. Feature integration - Combine components 2 & 3 to construct feature file for predictive model
5. Model training (optional) - Train predictive model using CRISPR-validated E-G pairs from K562 dataset
6. Prediction - Apply trained model to assign scores to each element-gene pair

System requirements

Hardware requirements

The scE2G pipeline necessitates a standard computer furnished with ample RAM to facilitate the operations as defined by a user. For optimal performance, we suggest using a computer…