Analysis of genetic association using Hierarchical Clustering and Cluster Validation Indices

G. ABRAS; I. PAGNUCO; M.BRUN; J. PASTORE ; V. BALLARIN

GENOMICS

ACADEMIC PRESS INC ELSEVIER SCIENCE

Lugar: Oak Park; Año: 2017 vol. 109 p. 438 - 445

0888-7543

It is usually assumed that co-expressed genes suggest co-regulation in the underlying regulatory network. Determining sets of co-expressed genes is an important task, where significative groups of genes are defined based on some criteria. This task is usually performed by clustering algorithms, where the whole family of genes, or a subset of them, are clustered into meaningful groups based on their expression values in a set of experiment. In this work we used a methodology to obtain sets of co-expressed genes, based on the cluster validation indexes as a measure of cluster quality for individual gene groups, and a combination of several variants of hierarchical clustering to generate the candidate groups. The algorithm provides a balance between search time and detection rate. It avoids the full search, which can be impractical for large number of genes, with the cost of missing some good sets, but it is able to detect most of the top ranked sets, which is not usually possible by using only one clustering algorithm. The effectiveness of this algorithm is related to the ability of the validation clustering index to score properly compact groups, which are at the same time separated from other groups of genes. With this analysis we verified the suitability of the proposed tool for the detection of sets of co-expressed genes, and we considered that it is useful for biologists or researchers in computational biology interested in generating new hypotheses about the co-expression of genes, or genomic markers like QTLs, which are not provided in most standard analysis tools. This algorithm will generate quickly a set of good groups on base of a clustering validation index, and combines the advantages of each variant of hierarchical clustering algorithm.