Published August 11, 2025 | Version v4
Software Open

**DEPRECATED** Borzoi Human Model Predictor (Linder et al. 2025) using the Genomic API for Model Evaluation (GAME) Framework

Authors/Creators

  • 1. School of Biomedical Engineering, University of British Columbia

Description

This record provides a Predictor container for the Borzoi human model (Linder et al. 2025, Nature Genetics). Borzoi is a deep learning model that predicts cell-type and tissue-specific RNA-seq coverage directly from DNA sequence, enabling interpretation of genetic variants across multiple layers of gene regulation, including transcription, splicing, and polyadenylation. It is trained on human RNA-seq data from ENCODE (with 866 datasets across diverse biosamples, including cell lines and adult tissues) and Genotype-Tissue Expression (GTEx) data (with 2-3 replicates for each tissue, processed by the recount3 project). The training dataset also includes epigenomic datasets from the Enformer model, such as CAGE, DNase-seq, ATAC-seq, and ChIP-seq tracks.

The Predictor container (borzoi_human_predictor.sif) includes:

  1. Predictor script for sequence processing and error handling.
  2. Integrated Borzoi model with its dependencies and borzoi-gpu conda environment created using borzoi_gpu_environment.yml.
  3. Baskerville and Borzoi packages, installed from local source copies of repositories (as of March 2025), which also contain helper scripts and 4 replicate model weights.
  4. Support scripts like:
    • api_preprocessing_utils.py,
    • error_message_functions_updated.py,
    • predictor_help_message.json.

NOTE: This container requires a GPU for execution because the Borzoi model relies on TensorFlow's GPU-accelerated operations.

Running the container:

Ensure Apptainer is intalled in the system the container is intended to run. Always run the Predictor first, so it can listen for incoming connections from Evaluators:

apptainer run --containall --nv borzoi_human_predictor.sif HOST_IP HOST_PORT MATCHER_IP MATCHER_PORT
 
Additional information about the GAME framework can be found on GitHub: Genomic API for Model Evaluation
Borzoi Human Predictor-specific information, along with a sample Evaluator, can be found within the same repository: Borzoi Predictor
Code repository for Borzoi models: Borzoi GitHub
Paper can be found here: Linder, J., Srivastava, D., Yuan, H. et al. Predicting RNA-seq coverage from DNA sequence as a unifying model of gene regulation. Nat Genet 57, 949–961 (2025). https://doi.org/10.1038/s41588-024-02053-6

Files

Files (11.1 GB)

Name Size
md5:8016534560cf7c124892c872a6aa5f35
11.1 GB Download

Additional details

Software

Repository URL
https://github.com/de-Boer-Lab/Genomic-Model-Evaluation-API
Programming language
Python