Menu

Javascript is not activated in your browser. This website needs javascript activated to work properly.
You are here

Applicability Domain Dependent Predictive Uncertainty in QSAR Regressions

Author:
  • Ullrika Sahlin
  • N. Jeliazkova
  • T. Oberg
Publishing year: 2014
Language: English
Pages: 26-35
Publication/Series: Molecular Informatics
Volume: 33
Issue: 1
Document type: Journal article
Publisher: John Wiley & Sons

Abstract english

Predictive models used in decision making, such as QSARs in chemical regulation or drug discovery, call for evaluated approaches to quantitatively assess associated uncertainty in predictions. Uncertainty in less reliable predictions may be captured by locally varying predictive errors. In the current study, model-based bootstrapping was combined with analogy reasoning to generate predictive distributions varying in magnitude over a model's domain of applicability. A resampling experiment based on PLS regressions on four QSAR data sets demonstrated that predictive errors assessed by k nearest neighbour or weighted PRedicted Error Sum of Squares (PRESS) on samples of external test data or by internal cross-validation improved the performance of the uncertainty assessment. Analogy using similarity defined by Euclidean distances, or differences in standard deviation in perturbed predictions, resulted in better performances than similarity defined by distance to, or density of, the training data. Locally assessed predictive distributions had on average at least as good coverage as Gaussian distribution with variance assessed from the PRESS. An R-code is provided that evaluates performances of the suggested algorithms to assess predictive error based on log likelihood scores and empirical coverage graphs, and which applies these to derive confidence intervals or samples from the predictive distributions of query compounds.

Keywords

  • Earth and Related Environmental Sciences
  • Predictive error
  • Variance
  • Reliability
  • Bootstrap
  • Risk assessment

Other

Published
  • ISSN: 1868-1751
me in Lundagård
E-mail: ullrika.sahlin [at] cec.lu.se

Researcher

Centre for Environmental and Climate Research (CEC)

+46 46 222 68 31

+46 73 827 44 32

E-D340

50

Read more on the blog

Bayes@Lund2017 19-12 April

2017-03-23
The program for Bayes@Lund2017 is now ready Don’t forget to sign up – and be fast to not miss out! We start in room MA7, Maths b…

Workshop on Bayesian Networks for risk assessment and decision making

2017-02-08
You most welcome to attend a workshop on Bayesian Networks in risk assessment and decision making in Lund March 28 and 29th, 2017. This work…

Studentprojekt: Tuggummin som en indikator för nedskräpning i urbana miljöer

2017-02-06
Studenter sökes för ett projekt under våren 2017 Detta projekt går ut på att ta reda på hur förekomst av tuggummin förhåller sig till föreko…

Centre for Environmental and Climate Research, CEC

Sölvegatan 37
223 62 Lund, Sweden

Visiting address
The Ecology building, Sölvegatan 37, Lund