Accessible Machine Learning Approaches for Toxicology

Sean Ekins, Alex M. Clark, Alexander L. Perryman, Joel S. Freundlich, Alexandru Korotcov, Valery Tkachenko

Research output: Chapter in Book/Report/Conference proceedingChapter

4 Scopus citations

Abstract

This chapter explores machine learning algorithms and makes them accessible for seeding drug discovery projects. It explores the authors's novel data pruning strategy when constructing Bayesian models to predict other types of properties. The chapter summarizes the application of the machine learning methods to toxicology datasets and transporters. Support vector machine (SVM) is one of the most popular supervised machine learning algorithms used mostly in classification problems and it is quite effective in high-dimensional spaces. The deep learning (DL) model is trained with a dataset by adjusting the weights to give the response expected for a certain input. The chapter compares deep neural networks (DNNs) and classic machine learning (CML) methods with different datasets of toxicological relevance for future embedding into the pen science data repository (OSDR). Collaborative Drug Discovery (CDD) model hosts the software and customers' data vaults on its secure servers.

Original languageEnglish (US)
Title of host publicationComputational Toxicology
Subtitle of host publicationRisk Assessment for Chemicals
Publisherwiley
Pages1-29
Number of pages29
ISBN (Electronic)9781119282594
ISBN (Print)9781119282563
DOIs
StatePublished - Feb 14 2018

All Science Journal Classification (ASJC) codes

  • Chemistry(all)
  • Chemical Engineering(all)
  • Computer Science(all)
  • Pharmacology, Toxicology and Pharmaceutics(all)

Keywords

  • Bayesian models
  • Classic machine learning methods
  • Collaborative drug discovery model
  • Deep learning models
  • Deep neural networks
  • Support vector machine
  • Toxicology

Fingerprint Dive into the research topics of 'Accessible Machine Learning Approaches for Toxicology'. Together they form a unique fingerprint.

Cite this