PersonalBPNet

A PyTorch library of BPNet-family models for predicting base-resolution genomic signal (e.g., PRO-cap read coverage) from DNA sequence, plus CLI tools for prediction and attribution.

• PersonalBPNet — bpnetlite.bpnet.BPNet, with a fit() that validates against a PyTorch DataLoader instead of holding the whole validation set in memory.
• CLIPNET — PersonalBPNet plus batch normalization; PyTorch port of the original TensorFlow CLIPNET.
• ProCapNet — BPNet with a masked profile loss, improving attributions (from the ProCapNet paper).
• PauseNet — transfer-learns a base-resolution model to predict a single scalar phenotype per locus.

Install

With uv:

uv add personalbpnet

Or with pip:

pip install personalbpnet

To load TensorFlow-trained weights (requires h5py), install with the tf extra: uv add "personalbpnet[tf]" or pip install "personalbpnet[tf]".

Installing from source, for local development, is covered in CONTRIBUTING.md.

Documentation

This package is under active development; APIs may change between releases. See CHANGELOG.md.

Citing

If you use PersonalBPNet/CLIPNET/ProCapNet, or the underlying BPNet architecture (via the bpnet-lite dependency), please cite the corresponding paper (see CITATION.cff):

• Avsec et al. (2021). Base-resolution models of transcription-factor binding reveal soft motif syntax. Nature Genetics 53:354–366. — BPNet
• He & Danko (2024). Dissection of core promoter syntax through single nucleotide resolution modeling of transcription initiation. bioRxiv. — CLIPNET
• Cochran et al. (2024). Dissecting the cis-regulatory syntax of transcription initiation with deep learning. bioRxiv. — ProCapNet

License

MIT