Enzyme Similarity Networks

Abstract

There is a crescent use of enzymes in multiple industries and sciences, ranging from materials and fuel synthesis to pharmaceutical and food production. Their applicability in this variety of fields depends not only on their biochemical function but also on their physicochemical properties. In the present work, we describe how the coincidence methodology can be employed to construct similarity networks of seventy well-studied enzymes of the Glycoside Hydrolase Family 13 and to identify communities of physicochemically related enzymes. More specifically, each of the selected enzymes is mapped into a network node, while the links between pairs of enzymes are determined by the coincidence similarity between selected physicochemical features of interest. The obtained networks have modularity and number of isolated nodes optimized respectively to two parameters involved in the coincidence methodology, resulting in highly modular networks. In order to investigate the effect of the considered physicochemical features on the enzymes relationships, the coincidence-based method also is applied to create a meta-network, in which the enzymes similarity networks obtained by the combination of every possible feature becomes nodes of a feature combination network, and the coincidence similarity between those networks defines the respective links. The obtained feature combination network systematically and comprehensively indicates the impact of the selected physicochemical features on enzyme similarity. Several interesting results are reported and discussed, including the identification of subgroups of enzymes with similar physicochemical features within catalytical classes, providing important information for the selection and design of enzymes for targeted biotechnological applications.