Chemical compound classification based on improved Max-Min kernel

Qiangrong Jiang, Can Zhai and

Abstract

In this paper, a new method is defined that encapsulates the diameter information of the fingerprint. Considering fingerprints of the large diameter containing some large and quite precise information, such fingerprints should be given relative large weight. A review of the MIN–MAX kernel is provided followed by a thorough definition of the improved method by which the diameter information of the fingerprint was encoded. The paper concludes with comparative QSAR studies to test the efficacy of this improved MAX-MIN kernel in comparison with a number of published methods and results with two datasets: PTC and NCI for the convenience of comparison.