Descriptor-based computational analysis reveals a new classification scheme for feruloyl esterase

Poster (konferens), 2010

One of the most important groups of enzymes, the feruloyl esterases (FAEs), is ubiquitous in both simple and complex organisms. FAEs have gained importance in biofuel, medicine, food and nutrition industries due to their capability of acting on a large range of substrates for cleaving or/and synthesizing ester bonds. Despite the extensive studies in the past two decades on the production and partial characterization of FAEs there is poor knowledge on the control mechanisms of substrate recognition. The present work centers on the descriptor-based characterization and structural analysis of FAEs to propose a complete classification system of this industrially important enzyme. We collected 365 sequences from plants, fungi and bacteria and clustered them based on 91 descriptors derived from the amino acid sequence. A Support Vector Machine (SVM) model was subsequently trained for the classification of the sequences, which led to the identification of ten different sub families. The SVM model successfully recognized 98.30% of the sequences correctly that belong to respective clusters and all the sequences of the blind test set. Another important aspect of the present work involved the structural analysis of previously characterized FAEs based on the pharmacophoric features of the known substrates. In this work we apply an array of computational tools and we succeed to develop a new classification scheme for FAEs which opens new vistas in the application of this intriguing group of enzymes in biocatalytic transformations. The present work is not restricted to FAEs but represents a framework for the functional characterization and identification of substrate specificity for any poorly characterized enzyme group. In addition we demonstrated that sequence information can be used for constructing models that reveal the underlying structural reasons determining substrate specificities.