EvoR

EvoR is an R interface for the Evo2 Genomic Foundation Model (built by the Arc Institute). It allows researchers to leverage state-of-the-art biological AI for sequence generation, scoring, and embedding extraction directly within the R environment.

Key Features

• Generation: Generate long DNA sequences (up to 1M bp) with biological context.
• Scoring (VEP): Calculate log-likelihood scores to predict the effect of variants (e.g., in Type 2 Diabetes research).
• Embeddings: Extract 4096-dimensional vectors for deep learning and regulatory pattern recognition.

Installation

# Install development version from GitHub
# install.packages("devtools")
devtools::install_github("BDB-Genomics/EvoR")

Quick Start

1. Set your API Key

You need an NVIDIA BioNeMo API key to use EvoR.

Sys.setenv(NVIDIA_API_KEY = "your_key_here")

2. Get Variant Effect Scores

# Query the model
response <- evo2_query("ACTG...")

# Extract scores
scores <- evo2_get_score(response)
print(head(scores))

3. Extract Embeddings

# Get high-dimensional vectors for a sequence
emb_response <- evo2_query_embeddings("ACTG...")
embeddings <- evo2_get_embeddings(emb_response)

# embeddings is now an R matrix [nucleotides x 4096]
dim(embeddings)

Reference

Evo 2 Nature paper: https://www.nature.com/articles/s41586-026-10176-5

---

Developed as part of the BDB-Genomics ecosystem.