Background: Information about Pharmacokinetics of compounds is an essential
component of drug design and development. Modeling the pharmacokinetic properties
require identification of the factors effecting absorption, distribution, metabolism and excretion
of compounds. There have been continuous attempts in the prediction of intestinal
absorption of compounds using various Artificial intelligence methods in the effort to reduce
the attrition rate of drug candidates entering to preclinical and clinical trials. Currently,
there are large numbers of individual predictive models available for absorption using
machine learning approaches.
Methods: Six Artificial intelligence methods namely, Support vector machine, k- nearest
neighbor, Probabilistic neural network, Artificial neural network, Partial least square and
Linear discriminant analysis were used for prediction of absorption of compounds.
Results: Prediction accuracy of Support vector machine, k- nearest neighbor, Probabilistic
neural network, Artificial neural network, Partial least square and Linear discriminant
analysis for prediction of intestinal absorption of compounds was found to be 91.54%,
88.33%, 84.30%, 86.51%, 79.07% and 80.08% respectively.
Conclusion: Comparative analysis of all the six prediction models suggested that Support
vector machine with Radial basis function based kernel is comparatively better for binary
classification of compounds using human intestinal absorption and may be useful at
preliminary stages of drug design and development.