Simulation of Fish Acute Toxicity of Pharmaceuticals Using Simplified Molecular Input Line Entry System (SMILES) Notation as a Representation of Molecular Structure

TL;DR

This study uses molecular structure data to predict how harmful pharmaceuticals are to fish, creating a useful tool for assessing drug toxicity in aquatic life.

Contribution

The novelty lies in building in silico models using SMILES notation to predict fish toxicity of pharmaceuticals with acceptable accuracy.

Findings

A model with a determination coefficient of 0.67 was developed for predicting fish toxicity.

The study demonstrates the feasibility of using SMILES-based models for complex biological endpoints.

The approach is useful for assessing pharmaceutical toxicity in aquatic organisms.

Abstract

The practice of using optimal descriptors has been applied for more than twenty years to develop in silico models. In the present study, a series of in silico models was built to predict the acute fish toxicity of pharmaceuticals using optimal descriptors. The SMILES format was used to represent the chemical structure. The data were split into five training and validation sets. The obtained model for fish toxicity yielded a determination coefficient of 0.67 for the external validation set, representing an acceptable quality, considering the complexity of the pharmaceuticals given their molecular structure and specific biological activity. This study is useful for assessing the acute fish toxicity of pharmaceuticals and, in general terms, as an approach to building models for complex biological endpoints.

Figures