Weighted Network Analysis of Biologically Relevant Chemical Spaces

TL;DR

This paper introduces a weighted network approach to analyze biologically relevant chemical spaces, revealing that compounds with extreme bioactivity levels are more strongly interconnected than others.

Contribution

It proposes a novel weighted network method that avoids thresholding, providing new insights into the connectivity related to bioactivity in chemical spaces.

Findings

High or low bioactivity compounds have stronger network connections.

Weighted networks reveal bioactivity relationships without threshold bias.

The approach improves understanding of chemical space topology.

Abstract

In cheminformatics, network representations of the space of compounds have been suggested extensively. Among these, the threshold-network consists of nodes representing molecules. In this network representation, two molecules are connected by a link if the Tanimoto coefficient, a similarity measure, between them exceeds a preset threshold. However, the topology of the threshold-network is affected significantly by the preset threshold. In this study, we collected the data of biologically relevant compounds and bioactivities. We defined the weighted network where the weight of each link between the nodes equals the Tanimoto coefficient between the bioactive compounds (nodes) without using the threshold. We investigated the relationship between the strength of the link connection and the bioactivity closeness in the weighted networks. We found that compounds with significantly high or low bioactivity have a stronger connection than those in the overall network.