Topologically Overlapped Fused LASSO Measure for Reconstructing Gene Regulation Networks

Abstract

Identifying target genes in Gene Regulation Network (GRN) models has always been challenging in Systems Biology. In this regard, indirect gene regulatory hierarchical architectures may be promising enough, considering varied topological structures and unknown gene regulation factors. Such causal regulations can be investigated, keeping in force all perturbation experiments of a dataset. Contemporary research primarily highlights direct interaction networks, which mostly forego the inevitable presence of a third entity, if any, towards varied forms of causal regulations. In this article, we have developed a hierarchical algorithm that unveils the genetic wiring through the Fused Least Absolute Shrinkage and Selection Operator (Fused-LASSO) technique with a Topological Overlap (TO) measure as the interaction structure. The potential power of this proposed model is studied over YEAST Cell Cycle, and Human cancer cell line data (HeLa). In this connection, the different statistically significant hierarchical regulation outcomes maintaining parity with the direct interaction structures, if any, to the target genes may throw new light on gene regulation statistics.