Benchmarking of Structural Variant Detection Tools for Microorganisms

A variety of structural variant (SV) detection tools exist and they are generally tested on the human genome to assess performance. However, these tests may be inapplicable to the genomes of microorganisms. The focus of this paper is a discussion of benchmarks that were run on microbial genomes. Representative tools from a variety of methods for SV detection, such as: read-pair, read-depth or read-count, split-read, and de novo assembly were selected to determine the best approach for detecting SVs in microbial genomes. The tools are tested on simulated data that includes deletion, duplication, insertion, and inversion SVs. Experimental results show that when execution time, precision, recall, and coverage of the reads are considered, the most accurate results come from tools using the read-pair method mixed with either the split-read or read-count method, specifically, LUMPY [1] and SVDetect [2] performed best.