Integração de bases de dados de genes homólogos e aplicação em análises de sequências

Abstract

Biological databases are very useful sources for scientific research. Some secondary databases organize their data in orthologous groups and functional categories, such as COG (Cluster of Ortholog Groups) and KO (KEGG Orthology). The KO database was used for an automatic annotation test with C. elegans' ESTs. We performed a control experiment on which an EST is designated to its cognate protein in C. elegans. To the annotation stage we simulated a new transcriptome by removing the worms sequences from the database. We obtained three annotation classes: correct or changed (when the annotated KO was equal or different from the designated, respectively) and speculated (when the EST is annotated, but not designated). We obtained 68%, 4% and 28% correct, changed and speculated annotations, respectively. However, the speculation decreases to 4,4% when we designate those EST using proteins that are not included in KO database. Trying to increase the amount of information in databases like COG and KO, we developed a methodology based on recruiting sequences that share the UniRef50 cluster as a recruiter protein that is already present on the original database. A size selection filter removed recruited proteins with a difference higher than 10% the recruiter protein length. Using this methodology we increased the amount of proteins in the COG database from 124.369, from 63 genomes, to 961.725, representing 3.477 genomes. The database was denominated UniRef Enriched COG (UECOG). Recently a new enrichment was performed using a filter which we required that the alignment between the recruited and recruiter proteins showed an valor-e lower than 1x10-10 and cover at least 50% of the recruiter protein. We obtained 2.450.485 entries, from 5.748 distinct genomes (UECOG 2.0). The previous procedure was used to enrich the KO database, increasing the amount of data from 1.940.617 proteins to 4.447.538, and the amount of organisms from 1.315 to 32.213. The usage of alignment significance filter and recruiter sequence coverage showed high accuracy in separating similar proteins, but with different orthologous groups. The enriched database UEKO (UniRef Enriched KO) was used to test the automated annotation of ESTs, as described previously. The proportion of changed annotation decreased to 1% and the correct increased to 74%. However, the speculation remained frequent, showing that we still have a lot of information to be added. The amount of correct annotation increased in 12%. We also performed studies of the human gut microbial metagenome. One of them, using 13 public samples, compared the annotation provided by KO and UEKO. This comparison showed that the UEKO database annotates more sequences than KO, once that more than 100 groups have exclusive alignment with the enriched database. However, the major difference is in qualitative aspect, once that we have an improvement in BLAST scores and proteins from closer clades annotate the sequences, which was demonstrated by phylogenetic analysis. The other study aimed in analyzing, phylogenetic and functionally, the microbiota structure and we identified certain phylogenetic and functional patterns. Those groups, known as enterotypes, have some features that differentiate them from the others, such as the over-representation of enzymes related to vitamin biosynthesis in some enterotype when compared to the others.