Abstract
Although species of Brucella genus share more than 90% of their genomic content, their phenotypic characteristics may vary greatly among species and even strains. To explain some of those differences, e.g. zoonotic potential, pathogenicity or immune response specificity, we devised the supervised machine-learning algorithm that compares the 3D protein structure, molecular weight, hydrophobicity, isoelectric point and structure stability. Based on the available whole chromosome sequences of B. abortus, B. melitensis, B. suis, B. ovis, B.canis and B. microti, the database was created, and for the efficacy of the model, only annotated proteins were used to predict the 3D structure based on the Swiss model system. Further, using the Bio.SeqUtils the protein structure characteristics were calculated and used for the comparisons. Our in silico results showed two membrane (cyst, cysW), urease and one heat-chock chaperon proteins characteristic for B. melitensis strains, while in other species, they have 100% homology. In B. ovis strains two cytochrome oxidase subunits II and two sugar metabolic proteins were constantly truncated and their structure stability, iso-electric point and hydrophobicity varied greatly compared to other Brucella species. Also, B. ovis strains in 95% of the cases were missing and the remaining 5% had very shortened transcriptional regulator; iron-sulfur, NAD binding and ATP binding enzymes; dsbA family protein. Carbohydrate hydrolase is characteristically truncated in B. microti, while imidazolonepropionase has B.canis specific properties. All strains of B. ovis and B. microti have similar ATP-dependent helicase HrpB that differs in its’ hydrophilic characteristics and iso-electric point compared to others, while these species completely miss lipids catabolismand cardiolipin sintase enzymes. Two proteins alcohol dehydrogenase AdhP and ornithine/lysine decarboxylase, were found to be present only in B. abortus, B. suis and B.canis, while only Cu(I)-responsive transcriptional regulator is 100% similar between three species. Differences in OMP 2a and OMP 31 among species were observed regarding molecular weights and protein stability. These in silico analyses are in concordance with published results, but further laboratory confirmations are necessary. However, our analysis used intra species differ-ences and, for now, we have not included differences among biovars, nor comparisons of active sites.
