Majority Score Clustering Algorithms to Identify the Chemical Compounds Having Alike Antibacterial Activity

Abstract

This work presents a detailed study of majority-based clustering algorithms decision on three different data sets of anti-microbial evaluation, the minimum inhibitory concentration of antibacterial, antibacterial, and anti-fungal activity of chemical compounds against 04 bacteria (E. Coli, P. Aeruginosa, S. Aureus, S.Pyogenes) and 02 Fungus (C. Albicans, As. Fumigatus). Clustering is an unsupervised machine learning method used to divide the chemical compounds on the bases of their similarity. In this thesis we applied the K-means clustering, Gaussian mixture model (GMM), and Mixtures of multivariate t distribution on antibacterial activity data sets. For an optimal number of clusters and to determine which clustering algorithm performs best we used a variety of clustering validation indices (CVI) which are within sum square (to be minimized), connectivity (to me minimized), silhouette width (to be maximized), the Dunn index (to be maximized). On the bases of the majority score clustering algorithm, we conclude that K-means and the mixture of multivariate t distribution satisfy the maximum and the Gaussian mixture model satisfies the minimum cluster validation indices. The K-means algorithm and mixture of multivariate t distribution give 3 optimal number of clusters in an anti-microbial evaluation of antibacterial activity data set and 5 number of optimal clusters in minimum inhibitory concentration (MIC) of anti bacteria’s data set. K-means, Mixtures of multivariate t distribution and Gaussian mixture model give 3 optimal number of clusters in the antibacterial and anti-fungal activity data set. The K-means clustering algorithm gives the best performance on the bases of a majority-based decision. This study may help the pharmaceutical industry, alchemists as well as doctors in the future.