Análise comparativa de estratégias de identificação de polimorfismos de nucleotídeo único em gossypium hirsutum

Abstract

High-throughput sequencing platforms make it possible to generate huge amounts of data extremely quickly. However, sequencing methodologies are highly sensitive to errors, making the process of obtaining data highly dependent on bioinformatics tools. The challenge of identifying allelic variants in the processing of sequencing data encompasses the alignment of sequences with the reference genome of the target species. Variations in genetic variants, acquired through diverse approaches to polymorphism identification, may impart direct impacts upon the utilization of such data in genetic studies, including associative mapping and genomic selection.The present study aimed to compare two variant calling approaches (Fast-GBS and BWA/BFCTools) to determine their impact on the identification of single polymorphism nucleotides (SNPs) in a panel of 250 cotton (Gossypium hirsutum) genotypes of which 72 correspond to genotypes from TMG's germplasm bank (single-end sequences obtained by GBS with Ion Torrent sequencing) and 178 come from a study in the literature (paired-end sequences obtained by Illumina sequencing). The results were compared by surveying the total number of SNPs recovered, as well as the number of SNPs recovered per chromosome. Other metrics used were the SNP-Score (capable of weighting the number of SNP occurrences per allele calling pipeline), computational time and principal component analysis. The Fast-GBS pipeline retrieved a total of 417,975 SNPs for the TMG raw data subset and 38,685,370 SNPs for the literature raw data subset. We can identify that there are 24,402 SNPs in common across all datasets when the BWA/BCFTools pipeline was used and 15,348 SNPs in common across all datasets when the Fast-GBS pipeline used. At the end of these analyses, it was possible to conclude that the Fast-GBS pipeline has a better computational performance and that paired-end sequences suffer little influence from the software used to call alleles, due to its high precision. For single-end sequences, Fast-GBS pipeline performed better for raw sequencing data and BWA/BCFTools performed better with filtered data. The results obtained reinforce the need to consider several aspects when choosing methods for analysis.