AlphaGenomeR

High-resolution functional genomic predictions from AlphaGenome, directly in R

_{Bridging the official AlphaGenome Python SDK into Bioconductor-friendly R workflows}

Overview

AlphaGenomeR provides an R interface to Google DeepMind's AlphaGenome API. It uses reticulate to call the official Python client and returns R-friendly objects for downstream analysis.

Typical uses:

• query a 1 Mb genomic interval
• retrieve RNA-seq, ATAC-seq, DNase-seq, CAGE, TF/histone, splicing, and contact predictions
• convert results into matrices and metadata tables in R
• integrate predictions into Bioconductor workflows

Project Status

• Bioconductor submission: 0.99.0
• Validated on real AlphaGenome API outputs
• Actively developed

Installation

1. Create a dedicated Python environment

AlphaGenomeR depends on the Python package alphagenome. Use the included Conda environment file to avoid dependency conflicts with your base environment.

conda env create -f dev/alphagenome.yaml
conda activate alphagenomer

2. Install the R package

For the development version from GitHub:

if (!requireNamespace("BiocManager", quietly = TRUE)) {
    install.packages("BiocManager")
}

BiocManager::install("BDB-Genomics/AlphaGenomeR")

Once accepted into Bioconductor:

if (!requireNamespace("BiocManager", quietly = TRUE)) {
    install.packages("BiocManager")
}

BiocManager::install("AlphaGenomeR")

3. Point reticulate to the environment

Set RETICULATE_PYTHON before loading the package.

Sys.setenv(
    RETICULATE_PYTHON = "/path/to/miniconda3/envs/alphagenomer/bin/python"
)

library(AlphaGenomeR)

Quick Start

Sys.setenv(
    RETICULATE_PYTHON = "/path/to/miniconda3/envs/alphagenomer/bin/python"
)

library(AlphaGenomeR)

results <- alphagenome_query(
    access_token = "YOUR_API_KEY",
    genomic_region = "chr17:42560601-43609177",
    ontology_terms = "UBERON:0002048",
    requested_outputs = c("RNA_SEQ", "ATAC")
)

rna <- alphagenome_get_rna_seq(results)
atac <- alphagenome_get_atac(results)

Requirements

• Python >= 3.10
• alphagenome Python package >= 0.6.1
• valid AlphaGenome API key
• internet access for live API queries

Supported Outputs

• alphagenome_get_rna_seq()
• alphagenome_get_atac()
• alphagenome_get_cage()
• alphagenome_get_dnase()
• alphagenome_get_chip_tf()
• alphagenome_get_chip_histone()
• alphagenome_get_splice_sites()
• alphagenome_get_splice_junctions()
• alphagenome_get_splice_usage()
• alphagenome_get_procap()
• alphagenome_get_contact_maps()

Typical Workflow

1. Create or activate the dedicated Python environment.
2. Set RETICULATE_PYTHON.
3. Query a genomic region with alphagenome_query().
4. Extract modality-specific data with the helper functions.
5. Use the returned matrices and metadata in downstream R analyses.

Citation

If you use AlphaGenomeR, please cite:

• AlphaGenomeR (R package): https://doi.org/10.5281/zenodo.19774275
• AlphaGenome model publication

You can also run:

citation("AlphaGenomeR")

---

Developed by Himanshu Bhandary