# ToxAssay: a hierarchical model-driven tool for advanced toxicogenomics biomarker discovery

## TL;DR

ToxAssay is a new tool that improves the discovery of toxicity biomarkers by modeling complex genetic relationships in drug toxicity data.

## Contribution

ToxAssay introduces a hierarchical linear model for more accurate and efficient biomarker identification in toxicogenomics.

## Key findings

- ToxAssay identified 71 key genes and 26 core genes with high accuracy in glutathione depletion-induced toxicity.
- The tool showed strong cross-correlation (Pearson’s r = 0.88) with external datasets.
- An AOP analysis algorithm linked glutathione depletion to disease outcomes.

## Abstract

Understanding the genetic basis of drug-induced toxicity is crucial for drug development. In-silico analysis of toxicogenomics datasets facilitates early detection of toxicity biomarkers. However, existing tools struggle with the complex interdependencies among hierarchically structured variables, leading to inaccurate biomarker identification. To address this limitation, we developed a Hierarchical Linear Model (HLM) and implemented it in the R package ToxAssay, offering extensive functionality for comprehensive toxicity assessment.

ToxAssay outperforms existing methods by improving biomarker detection and computational efficiency. Applied to glutathione depletion-induced toxicity, it prioritized 71 key genes and identified 26 core genes with high discriminative accuracy (AUC = 0.97) and strong cross-correlation (Pearson’s r = 0.88) with external datasets. Additionally, our advance outcome pathway (AOP) analysis algorithm uncovered disease outcomes linked to glutathione depletion. These findings provide precise insights into the molecular mechanisms driving drug-induced toxicity.

ToxAssay is available as an open-source R package at https://github.com/Fun-Gene/toxassay.