Classificação bacteriana baseada em proteínas ribossomais oriundas de dados genômicos

Abstract

In health and agriculture, bacterial identification is essential to understand the composition of the microbial community and its ecology. Microorganism identification techniques seek greater accuracy, speed, and less cost. One technique that has been studied and widely used for the identification of microorganisms is the identification through mass spectra. Through peaks referring to the most abundant molecular masses recorded in the spectrum, it is possible to generate a profile for the recognition of a microorganism. Another way to identify a mass spectrum is through peaks that are expected to appear in the spectrum, the model which this work has made use of. To assume the expected peaks in the spectrum, estimated molecular weights of ribosomal proteins were calculated. These proteins are called housekeeping, that is, they are ubiquitous and responsible for the basic cellular functioning. In addition to their abundant prokaryotic content, ribosomal proteins are highly conserved and do not change their physiology for different cell media or stages. The estimated weights formed a presumed database containing all information obtained from the NCBI repository and only data noted as complete were used, the database created was named Puchuy and has 14689 records. This presumed database was generated for taxonomy at Domain, Phylum, Class, Order, Family, Genus, and Species level, and then subjected to machine learning. Thus, it was possible to obtain classification models of microorganisms based on ribosomal protein values. Models were generated for each taxonomic level, which was used only for those that had better performance for each level. A clustering algorithm was also added to aid classification. With the models generated by the machine learning, the software was developed, able to classify the microorganisms in the Phylum, Class, Order, Family, Genus and Species level. Finally, different classifiers were compared for each taxonomic level, with and without the use of a clustering method.