Predicting metabolic pathways of plant enzymes without using sequence similarity: Models from machine learning

Author(s)

Almeida, Rodrigo de Oliveira ^{[1, 2]} ; Valente, Guilherme Targino ^{[2, 3]}

Affiliation

^[1] Inst Fed Educ Ciencia & Tecnol Sudeste Minas Gera, Muriae - Brazil

^[2] Sao Paulo State Univ, Sch Agr, Dept Bioproc & Biotechnol, UNESP, Botucatu, SP - Brazil

^[3] Max Planck Inst Heart & Lung Res, Dept Dev Genet, Bad Nauheim - Germany

• Document type: Journal article
• Source: PLANT GENOME; v. 13, n. 3 AUG 2020.
• Web of Science Citations: 0

Abstract

Most of the bioinformatics tools for enzyme annotation focus on enzymatic function assignments. Sequence similarity to well-characterized enzymes is often used for functional annotation and to assign metabolic pathways. However, these approaches are not feasible for all sequences leading to inaccurate annotations or lack of metabolic pathway information. Here we present the mApLe (metabolic pathway predictor of plant enzymes), a high-performance machine learning-based tool with models to label the metabolic pathway of enzymes rather than specifying enzymes' reactions. The mApLe uses molecular descriptors of the enzyme sequences to perform predictions without considering sequence similarities with reference sequences. Hence, mApLe can classify a diversity of enzymes, even the ones without any homolog or with incomplete EC numbers. This tool can be used to improve the quality of genomic annotation of plants or to narrow down the number of candidate genes for metabolic engineering researches. The mApLe tool is available online, and the GUI can be locally installed. (AU)

• FAPESP's process: 17/08463-0 - Systems biology and synthetic biology to re-engineering metabolic pathways in Saccharomyces cerevisiae: dealing with ethanol tolerance
• Grantee: Guilherme Targino Valente
• Support type: Program for Research on Bioenergy (BIOEN) - Regular Program Grants
• FAPESP's process: 15/12093-9 - Integrative analysis applied to ethanol tolerance in Saccharomyces cerevisiae strains: an approach using transcriptomes, proteomes, system biology and machine learning
• Grantee: Guilherme Targino Valente
• Support type: Program for Research on Bioenergy (BIOEN) - Regular Program Grants