Investigating Sensitivity, Specificity and Accuracy of Variant Calling Pipelines for Analyzing SARS-CoV-2 Data

The rapidly increasing popularity of Next Generation Sequencing and analysis methods in clinical and research settings necessitates an understanding of ideal combinations in identifying genomic variants. Especially with the importance of detecting accurate variants for the development of targeted SARS-CoV-2 vaccines. This research compares the results of two ‘Mapping Algorithms ‘, BWA-MEM and Bowtie2, and two ‘Variant Calling Algorithms ‘, LoFreq and FreeBayes, and their combinatory Variant Calling Pipelines on the analyses of Next Generation Sequencing (NGS) data of five SARS-CoV-2 samples collected from patients in the USA, India, Italy, and Malawi and sourced for this research from the publicly available NCBI SRA database. Our analysis of mapping algorithms found that BWA-MEM likely has higher sensitivity and specificity than Bowtie2 for mapping reads, and their specificity and sensitivity vary with read length. Furthermore, the accuracy of variant calling algorithms increases with the number of reads, while higher read length possibly leads to divergence in accuracy and sensitivity. Overall, FreeBayes was found to likely be more sensitive to detecting variants when used with Bowtie2 rather than BWA-MEM for analyzing SARS-CoV-2 data.