oncologylab/craftgrn

What it solves

CraftGRN addresses the challenge of reconstructing dynamic, condition-specific gene regulatory networks (GRNs). It integrates chromatin accessibility data (ATAC-seq) with gene expression data (RNA-seq) to identify how transcription factors (TFs) bind to specific sites and regulate target genes across different biological conditions.

How it works

The framework operates through a three-module pipeline:

1. Predict TF Binding Sites: It processes multiomic data to collapse TF motif footprints into consensus clusters and predicts condition-specific binding sites by correlating TF expression with chromatin accessibility or footprint scores.
2. Connect TFs to Target Genes: It links the predicted binding sites to target genes using genomic proximity, enhancer-gene maps, or 3D chromatin interaction data, filtering these links based on expression and binding evidence.
3. Learn Regulatory Topics: It uses topic modeling (including VAE-based representations and WarpLDA) to identify regulatory "topics" or programs from RNA and footprint signals, allowing users to compare regulatory links between conditions and visualize differential networks.

Who it’s for

It is designed for computational biologists and researchers studying gene regulation, specifically those working with matched ATAC-seq and RNA-seq datasets to understand how regulatory programs change across different conditions or species.

Highlights

• Multiomic Integration: Combines ATAC-seq and RNA-seq to infer TF binding and gene targets.
• Modular Workflow: Separates the process into binding site prediction, target linking, and topic discovery.
• Advanced Modeling: Employs topic modeling and VAE-based representations to extract high-level regulatory programs.
• Comprehensive QC: Generates detailed HTML quality control reports for each module of the pipeline.
• Cross-Species Support: Includes specialized handling for human and mouse pathway analysis.