PEAgent Tool

Sequence-to-single-cell chromatin accessibility for plant genomes.

PEAgent Tool is the inference package for the PEAgent models: per-species neural networks that predict, from 1,344 bp of DNA sequence alone, whether a region is accessible in each cell type of a single-cell ATAC atlas — and which bases in that sequence drive the prediction.

from peagent_tool import predict_celltypes

predict_celltypes(my_sequence, species="maize", top_k=5)

What you can do with it

• Predict accessibility

  ---

  Score any sequence — genomic, edited or fully synthetic — across every tissue_celltype group of an atlas.

  Tutorial

• Contribution maps

  ---

  Per-base attribution by bottleneck in-silico mutagenesis, in both the raw and the tissue-normalized convention used by the PEAgent browser.

  Tutorial

• Variant effects

  ---

  REF/ALT scoring of biallelic SNVs from a genome FASTA or from sequences you built yourself.

  Tutorial

• Motif synergy

  ---

  In-silico marginalization: test whether two motifs act more than additively, across a spacing and orientation grid.

  Tutorial

The three atlases

Species	Assembly	Nuclei	tissue_celltype groups	Tissues
Soybean (Glycine max)	Wm82.a4.v1	200,732	98	10
Maize (Zea mays)	B73 NAM-5.0	50,639	94	6
Rice (Oryza sativa)	Nipponbare MSU7	104,029	128	9

Every model shares the same architecture: a convolutional trunk, a 64-dimensional bottleneck embedding shared by all cells, and a per-cell output head. Because the bottleneck is shared, one in-silico mutagenesis pass yields contribution scores for every cell type at once.

Getting started

1. Install the package.
2. Get the model weights and cell-group metadata.
3. Work through Predicting accessibility.

Model weights

The trained weights are released once the accompanying manuscript completes peer review. The package, the tutorials and their outputs are complete and public now; only the download step is pending. See Models and data.

Every table and figure in the tutorials is the real, executed output of the code shown above it — the notebooks are run against the published models on each documentation build, not written by hand.

Reproducibility

The tissue-normalized contribution scores this package computes reproduce the attribution tracks served by the PEAgent genome browser to within float32 noise (median Pearson r = 0.999995 over 115 cell-type tracks). See the consistency check.

Related

• PEAgent analysis code — the full analysis pipeline accompanying the manuscript (training, motif discovery, cross-species transfer, ACR evolution).
• PEAgent portal and browser — interactive prediction and per-cell-type attribution tracks.