FirstSV: Fast and Accurate Approach of Structural Variations Detection for Short DNA fragments

Abstract

Structural variations caused by gene fusion represent a major class of somatically acquired variations in human malignancies, and include deletions, inversions, and translocations. Short fragmented reads are the main source of data from 2nd-generation sequencing, and detecting structural variations from this type of data is different from that of 1st-generation sequencing, where the read length is much longer. Current detection methods are low in specificity and are inefficient. We developed a hybrid algorithm, FirstSV, to meet the clinical demand for fast and accurate structural variation detection. Its main features include cluster analysis, realignment, and local assembly. FirstSV was validated with simulated data, with data from real patient samples, with data from standard testing samples, and with downloaded public data sets. FirstSV outperforms public-available methods in terms of sensitivity, precision, and operational efficiency. FirstSV is freely available at https://github.com/shenjia1/FirstSV.