Home

RiSPICE

RiSPICE (Rice SNP Prioritization Integrating Chromatin Effects) is a method for prioritizing non-coding rice variants by predicting their effects on chromatin features using a fine-tuned DNA language model.

This repository accompanies the manuscript:

Integrating Chromatin Features with a DNA Language Model Enables Prioritization of Non-coding Variants in Rice Association Studies

Overview

RiSPICE predicts the effects of non-coding variants on 12 chromatin features and prioritizes variants based on their predicted chromatin impact. Using the provided RiSPICE models, the workflow enables users to:

• Predict chromatin feature probabilities for reference and alternate alleles.
• Compute Per-Feature Scores and Overall Scores for each variant.
• Prioritize variants by ranking them according to their Overall Scores.
• Analyze predicted effects on individual chromatin features.
• Perform in silico saturation mutagenesis to identify impactful mutations within genomic regions.

In addition to the prediction workflow, the repository also includes the scripts used to reproduce the study, including data preprocessing, dataset generation, model fine-tuning, and evaluation.

Getting Started

If you are new to RiSPICE, begin with the Getting Started guide, which walks through the complete workflow using the OsHAK1 promoter case study from the manuscript.

The tutorial covers:

1. Running the prediction pipeline.
2. Analyzing and prioritizing variants.
3. Performing in silico saturation mutagenesis.

Wiki Contents

• Installation – Set up the RiSPICE environment, download the models, and prepare the empirical background distribution.
• Prediction Pipeline – Learn how to run the prediction pipeline on your own variants.
• Analysis – Perform variant prioritization, visualize chromatin effects, and conduct in silico saturation mutagenesis.