An empirical evaluation of error correction methods and tools for Next Generation Sequencing data

Abstract

Next Generation Sequencing (NGS) technologies produce massive amount of low cost data that is very much useful in genomic study and research. However, data produced by NGS is affected by different errors such as substitutions, deletions or insertion. It is essential to differentiate between true biological variants and alterations occurred due to errors for accurate downstream analysis. Many types of methods and tools have been developed for NGS error correction. Some of these methods only correct substitutions errors whereas others correct multi types of data errors. In this article, a comprehensive evaluation of three types of methods (k-spectrum based, Multi- sequencing alignment and Hybrid based) is presented which are implemented and adopted by different tools. Experiments have been conducted to compare the performance based on runtime and error correction rate. Two different computing platforms have been used for the experiments to evaluate effectiveness of runtime and error correction rate. The mission and aim of this comparative evaluation is to provide recommendations for selection of suitable tools to cope with the specific needs of users and practitioners. It has been noticed that k-mer spectrum based methodology generated superior results as compared to other methods. Amongst all the tools being utilized, Racer has shown eminent performance in terms of error correction rate and execution time for both small as well as large data sets. In multisequence alignment based tools, Karect depicts excellent error correction rate whereas Coral shows better execution time for all data sets. In hybrid based tools, Jabba shows better error correction rate and execution time as compared to brownie. Computing platforms mostly affect execution time but have no general effect on error correction rate.