Structuring dependencies in homogeneous groups can be possibly useful to gain insight into the behavior of complex systems. This is the case, for example, of biological and regulatory processes. In this contribution we approach the problem by applying Kernel Methods in order to cluster homogeneous pairs of gene-to-gene dependencies deriving from activation or inhibition relationships. Specifically, we apply Support Vector Clustering (SVC), which is a novelty detection algorithm, to provide groups of similarly interacting pairs of genes in respect to some measure, i.e. kernel function, of their regulatory activity. In our application we take advantage of the adjacency graph obtained from the approximation of a combinatorial optimization problem i.e. the Maximum Gene Regulatory Network (MGRN). The effectiveness of the proposed approach is given by numerical evaluation by comparing the modularity results of the obtained clusters with other standard techniques using a biological data set of microarray experiments.
Zoppis, I., Mauri, G. (2008). Clustering dependencies with support vectors. In Trends in Intelligent Systems and Computer Engineering (pp. 155-165). Springer [10.1007/978-0-387-74935-8_11].
Clustering dependencies with support vectors
ZOPPIS, ITALO FRANCESCOPrimo
;MAURI, GIANCARLOUltimo
2008
Abstract
Structuring dependencies in homogeneous groups can be possibly useful to gain insight into the behavior of complex systems. This is the case, for example, of biological and regulatory processes. In this contribution we approach the problem by applying Kernel Methods in order to cluster homogeneous pairs of gene-to-gene dependencies deriving from activation or inhibition relationships. Specifically, we apply Support Vector Clustering (SVC), which is a novelty detection algorithm, to provide groups of similarly interacting pairs of genes in respect to some measure, i.e. kernel function, of their regulatory activity. In our application we take advantage of the adjacency graph obtained from the approximation of a combinatorial optimization problem i.e. the Maximum Gene Regulatory Network (MGRN). The effectiveness of the proposed approach is given by numerical evaluation by comparing the modularity results of the obtained clusters with other standard techniques using a biological data set of microarray experiments.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.