Welcome to ChromBERT’s documentation!

ChromBERT is a pre-trained deep learning model designed to capture the genome-wide co-association patterns of approximately one thousand transcription regulators, thereby enabling accurate representations of context-specific transcriptional regulatory networks (TRNs). As a foundational model, ChromBERT can be fine-tuned to adapt to various biological contexts through transfer learning and provide insights into the roles of transcription regulators in the specific biological contexts without the need of additional genomic data for each regulator.

Note

This project is under active development.

Getting started:

• Installation
  • Installation From Source
  • Installation Using Singularity
• Quick Tour
  • 1 Customize the input
  • 2 Customize the model
  • 3 Customize the training process
  • 4 Start training !
  • 5 Task templates

Tutorials:

• Tutorial for fine-tuning ChromBERT
• Compiled Scripts for fine-tuning of ChromBERT
• Tutorial for get context specific TRN embedding
• CLI Reference

Examples:

• Example for cistrome imputation using prompt-enhanced ChromBERT
• Example for causal eQTL identification using prompt-enhanced ChromBERT
• Example for context-specific TRN: functional collaborations with EZH2 on funtional distinct loci
• Example for context-specific TRN: perturbation of STARR-seq
• Example for key regulators inference during cell state transition