Published April 4, 2017 | Version v1
Journal article Open

Influence of Varying Training Set Composition and Size on Support Vector Machine-Based Prediction of Active Compounds

  • 1. Department of Life Science Informatics, B-IT, LIMES Program Unit Chemical Biology and Medicinal Chemistry, Rheinische Friedrich-Wilhelms-Universitaẗ,Dahlmannstrasse2,D-53113Bonn,Germany


Support vector machine (SVM) modeling is one of the most popular machine learning approaches in chemoinformatics and drug design. The influence of training set composition and size on predictions currently is an underinvestigated issue in SVM modeling. In this study, we have derived SVM classification and ranking models for a variety of compound activity classes under systematic variation of the number of positive and negative training examples. With increasing numbers of negative training compounds, SVM
classification calculations became increasingly accurate and stable. However, this was only the case if a required threshold of positive training examples was also reached. In addition, consideration of class weights and optimization of cost factors substantially aided in balancing the calculations for increasing numbers of negative training examples. Taken together, the results of our analysis have practical implications for SVM learning and the prediction of active compounds. For all compound classes under study, top recall performance and independence of compound recall of training set composition was achieved when 250− 500 active and 500−1000 randomly selected inactive training instances were used. However, as long as ∼50 known active compounds were available for training, increasing numbers of 500−1000 randomly selected negative training examples significantly improved model performance and gave very similar results for different training sets.



Files (440.2 kB)

Name Size Download all
440.2 kB Preview Download

Additional details


BIGCHEM – Big Data in Chemistry 676434
European Commission