# Toxicity Prediction by Multimodal Deep Learning

## TL;DR

This paper introduces a multimodal deep learning approach that combines various neural network types and data representations to improve toxicity prediction accuracy for chemical compounds.

## Contribution

It presents a novel ensemble method integrating multiple neural networks and data modalities, achieving superior accuracy over existing toxicity prediction techniques.

## Key findings

- Significantly improved accuracy on standard toxicity benchmarks.
- Effective combination of string, image, and numerical data representations.
- Ensemble of diverse neural networks enhances predictive performance.

## Abstract

Prediction of toxicity levels of chemical compounds is an important issue in Quantitative Structure-Activity Relationship (QSAR) modeling. Although toxicity prediction has achieved significant progress in recent times through deep learning, prediction accuracy levels obtained by even very recent methods are not yet very high. We propose a multimodal deep learning method using multiple heterogeneous neural network types and data representations. We represent chemical compounds by strings, images, and numerical features. We train fully connected, convolutional, and recurrent neural networks and their ensembles. Each data representation or neural network type has its own strengths and weaknesses. Our motivation is to obtain a collective performance that could go beyond individual performance of each data representation or each neural network type. On a standard toxicity benchmark, our proposed method obtains significantly better accuracy levels than that by the state-of-the-art toxicity prediction methods.