Scholarlink

Identification of toxic substances is a critical element in drug discovery process and in the efforts towards a healthier environment. Predicting toxicity from the structural and compositional features of substances is therefore significant in devising effective strategies to nullify or mitigate the effects of toxins on living or environmental systems. Machine learning systems trained to perform discrimination of toxic substrates from the non-toxic can also potentially provide insights into the characterizing aspects of the such substances. The most common and simplest representation of substances is in the form of SMILESSimplified Molecular Input Line Entry System (SMILES) (Simplified Molecular Input Line Entry System), which is a simplified depiction of molecular structure as a string of characters. This study aims to look at CNN as a possible deep learning model capable of recognizing substantive patterns in the recognition of toxicity. A two-dimensional molecular representation is generated using RDKitRDKit and subsequently fed to CNN as input. The results indicate CNN as a potential tool to be further explored in cheminformatics.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Toxicity Prediction Using Convolutional Neural Networks: A Study of Deep Learning Approach

  • 章节

  • 发表日期：

    2025年

查看全文
  • Manjumol Mathew,
  • Neelesh Ashok,
  • S. Sachin Kumar

摘要

Identification of toxic substances is a critical element in drug discovery process and in the efforts towards a healthier environment. Predicting toxicity from the structural and compositional features of substances is therefore significant in devising effective strategies to nullify or mitigate the effects of toxins on living or environmental systems. Machine learning systems trained to perform discrimination of toxic substrates from the non-toxic can also potentially provide insights into the characterizing aspects of the such substances. The most common and simplest representation of substances is in the form of SMILESSimplified Molecular Input Line Entry System (SMILES) (Simplified Molecular Input Line Entry System), which is a simplified depiction of molecular structure as a string of characters. This study aims to look at CNN as a possible deep learning model capable of recognizing substantive patterns in the recognition of toxicity. A two-dimensional molecular representation is generated using RDKitRDKit and subsequently fed to CNN as input. The results indicate CNN as a potential tool to be further explored in cheminformatics.