Biological and Medicinal Chemistry

Don’t overweight weights: Evaluation of weighting strategies for multi-task bioactivity classification models

Authors

  • Lina Humbeck Medicinal Chemistry Department, Boehringer Ingelheim Pharma GmbH & Co. KG, Birkendorfer Str. 65, 88397, Biberach an der Riss, Germany ,
  • Tobias Morawietz Bayer AG, Pharmaceuticals, R&D, Digital Technologies, Computational Molecular Design, 42096 Wuppertal, Germany ,
  • Noe Sturm BioMedical Research, Novartis Campus, CH-4002 Basel, Switzerland ,
  • Adam Zalewski Amgen Research (Munich) GmbH, Staffelseestraße 2, 81477 Munich, Germany ,
  • Simon Harnqvist Computational Sciences, GlaxoSmithKline, Gunnels Wood Road, Stevenage SG1 2NY, United Kingdom ,
  • Wouter Heyndrickx Janssen Pharmaceutica N.V., Turnhoutseweg 30, 2340 Beerse, Belgium ,
  • Matthew Holmes Computational Sciences, GlaxoSmithKline, Gunnels Wood Road, Stevenage SG1 2NY, United Kingdom ,
  • Bernd Beck Medicinal Chemistry Department, Boehringer Ingelheim Pharma GmbH & Co. KG, Birkendorfer Str. 65, 88397, Biberach an der Riss, Germany

Abstract

Machine learning models predicting the bioactivity of chemical compounds belong nowadays to the standard tools of cheminformaticians and computational medicinal chemists. Multi-task and federated learning are promising machine learning approaches that allow priva-cy-preserving usage of large amount of data from diverse sources, which is crucial for achieving good generalization and high-performance results. Using large, real world data sets from six pharmaceutical companies, here we investigate different strategies for averaging weighted task loss functions to train multi-task bioactivity classification models. The weighting strategies shall be suitable for federated learning and ensure that learning efforts are well distributed even if data are diverse. Comparing several approaches using weights that depend on the number of sub-tasks per assay, task size, and class balance, respectively, we find that a simple sub-task weighting approach leads to robust model performance for all investigated data sets and is especially suited for federated learning.

Content

Thumbnail image of manuscript_weighting_melloddy.pdf

Supplementary material

Thumbnail image of SI_weighting_melloddy.pdf
Don’t overweight weights: Evaluation of weighting strategies for multi-task bioactivity classification models
The following are available Figure S1: Full performance plots phase II AUROC, Figure S2: Full performance plots phase II AUPR, Figure S3: Correlation analysis of AUPR and AUROC, Table S1: Tested weighting schemes during a pretest for phase II, Figure S4: Full performance plots phase III AUROC, Figure S5: Full performance plots phase III AUPR.

Supplementary weblinks

MELLODDY-TUNER
pipeline for data preparation developed as part of the MELLODDY project
Sparsechem
machine learning package for biochemical applications