# REDAC: RNA-seq expression data analysis chatbot

## TL;DR

REDAC is a chatbot tool that simplifies RNA-seq data analysis using natural language and provides reproducible results with visualizations and biological interpretations.

## Contribution

REDAC introduces a novel chatbot interface for RNA-seq analysis using LLMs and a retrieval-augmented generation module for pathway interpretation.

## Key findings

- REDAC enables rapid and transparent RNA-seq analysis through natural language queries.
- The tool integrates Gemma and LLaMA models with a PubMed-based module for pathway enrichment interpretation.
- REDAC supports reproducibility by generating automated analysis reports.

## Abstract

To date, due to the complexity of both the analytical processes and the result interpretation of RNA-seq expression data analyses, researchers often require the support of bioinformaticians expertise. Selecting appropriate statistical tests and performing essential data manipulations, such as normalization and filtering, in a rigorous and reproducible manner remains a significant challenge for many users.

We developed REDAC, a web-based R application that offers an interactive platform designed to simplify and enhance RNA-seq expression data exploration and analysis. REDAC provides a straightforward approach to perform differentially RNA-seq analysis rapidly, easily, and transparently through natural language queries from users. Moreover, it allows to run complete analyses, generate comprehensive visualizations, and obtain biological interpretation of pathway enrichment results via two popular Large Language Models: Gemma and LLaMA guided by a PubMed based Retrieval-Augmented Generation module. Finally, REDAC promotes reproducibility through the automated generation of analysis reports.

REDAC is available for local (https://github.com/franruss/REDAC) and online use (https://frusso.shinyapps.io/REDAC). User manual: https://github.com/franruss/REDAC/blob/main/docs/REDAC_user_manual.pdf