Full Paper

Visual Interpretation of Kernel-Based Prediction Models

  1. Katja Hansen1,*,
  2. David Baehrens1,
  3. Timon Schroeter2,
  4. Matthias Rupp 1,3,
  5. Klaus-Robert Müller 1,3

Article first published online: 5 SEP 2011

DOI: 10.1002/minf.201100059

Issue

Molecular Informatics

Molecular Informatics

Special Issue: Charting Chemical Space: „Challenges and Opportunities for Artificial Intelligence and Machine Learning”

Volume 30, Issue 9, pages 817–826, September 2011

Additional Information

How to Cite

Hansen, K., Baehrens, D., Schroeter, T., Rupp , M. and Müller , K.-R. (2011), Visual Interpretation of Kernel-Based Prediction Models. Molecular Informatics, 30: 817–826. doi: 10.1002/minf.201100059

Author Information

  1. 1

    Machine Learning Group, Technische Universität Berlin, Franklinstr. 28/29, FR 6–9, 10587 Berlin, Germany phone: 0049 30 31 4 24927

  2. 2

    Market Valid GmbH & Co KG, Große Hamburger Str. 16, 10115 Berlin, Germany

  3. 3

    Institute for Pure and Applied Mathematics, UCLA, Los Angeles, CA 90095-7121, USA

*Machine Learning Group, Technische Universität Berlin, Franklinstr. 28/29, FR 6–9, 10587 Berlin, Germany phone: 0049 30 31 4 24927

Publication History

  1. Issue published online: 14 SEP 2011
  2. Article first published online: 5 SEP 2011
  3. Manuscript Accepted: 21 JUL 2011
  4. Manuscript Received: 13 APR 2011

Funded by

  • FP7-ICT Programme
  • European Community, under the PASCAL2 Network of Excellence
  • ICT-216886
  • DFG. Grant Number: MU 987/4-2

SEARCH

SEARCH BY CITATION

ARTICLE TOOLS

View Full Article with Supporting Information (HTML) Get PDF (1031K)

Keywords:

  • Kernel-based learning;
  • Confidence estimation;
  • Domain of applicability;
  • QSAR;
  • QSPR

Graphical Abstract

Thumbnail image of graphical abstract

Abstract

Statistical models are frequently used to estimate molecular properties, e.g., to establish quantitative structure-activity and structure-property relationships. For such models, interpretability, knowledge of the domain of applicability, and an estimate of confidence in the predictions are essential. We develop and validate a method for the interpretation of kernel-based prediction models. As a consequence of interpretability, the method helps to assess the domain of applicability of a model, to judge the reliability of a prediction, and to determine relevant molecular features. Increased interpretability also facilitates the acceptance of such models. Our method is based on visualization: For each prediction, the most contributing training samples are computed and visualized. We quantitatively show the effectiveness of our approach by conducting a questionnaire study with 71 participants, resulting in significant improvements of the participants’ ability to distinguish between correct and incorrect predictions of a Gaussian process model for Ames mutagenicity.

View Full Article with Supporting Information (HTML) Get PDF (1031K)