Análise estrutural de ortólogos da proteína Glicolato Oxidase (GOX) em plantas utilizando Inteligência Artificial

Abstract

Beans have undeniable socioeconomic, cultural and nutritional importance for Brazil. Cowpea (Vigna unguiculata [L.] Walp) is among the most consumed bean species in Brazil and in the world, being one of the most widely adapted, versatile and nutritious grain legumes. Even showing resilience to various environmental stress conditions, cowpea is susceptible to pathogens such as cowpea severe mosaic virus (CPSMV). Recent studies have shown that several proteins are differentially expressed during the interaction between CPSMV and cowpea. Among these proteins is the enzyme Glycolate Oxidase (GOX), whose activity is the oxidation of glycolate to glyoxylate and H2O2. For a better understanding of the biological role of this enzyme in V. unguiculata and phylogenetically close species, the protein modeling tool that uses algorithms based on deep learning, RosettaFold, through the Robetta server, was used to predict structural models of GOX in species of beans (Vigna unguiculata, Vigna radiata, Vigna angularis, Phaseolus vulgaris), barrelclover (Medicago truncatula), soybean (Glycine Max), sorghum (Sorghum bicolor) and maize (Zea mays). With the obtaining of the structural models, 3 fundamental parameters were used to qualify the models obtained, being the RMSD (square root of the mean square deviation), TM-Score (Template Modeling Score) and the confidence parameter obtained in the Robetta server. The mean values of RMSD for each species ranged from 0.865 to 0.906, the confidence of the Robetta Server ranged from 0.85 to 0.88, and the TM-Score ranged from 0.7059 to 0.9101 with the exception of S. bicolor which had a mean of 0.1777. The values obtained from the parameters prove the quality of the models produced and that also proves that RosettaFold can predict protein structures from amino acid sequences accurately. In additionto providing structural models of the GOX protein from different plant species for application in functional studies, it is expected that this work will serve to encourage further studies of predictive models of protein structures related to stress resistance.