Influence of structural similarities between sequences over ontology annotation proteins using BLASTP

The functional prediction of proteins is one of main purposes of computational biology. Many techniques have been developed to solve this problem. Methods based on alignments of sequences like BLASTP are the most commonly used. However, these techniques have been criticized due to their on failures detecting homologous sequences under some identity thresholds between sequences. Although this is an argument frequently cited, there are no plublished studies truly showing the performance variances regarding the identity percentage between sequences. This study is fundamental to support studies that look for developing alternative methods. The present work contains an analysis of the influence of identity percentage between training sequences for BLASTP in the ontology annotation of proteins belonging to Embryophyta organisms (land plants).