How to “appreci8” Variant Calling in NGS Data

Background: Calling SNVs and indels in next-generation sequencing (NGS) data is a long-discussed topic. Numerous variant calling approaches are currently available. Despite considerable advances in recent years, detecting low-frequency variants with equally high sensitivity and positive predictive value (PPV) is still challenging. However, for the application of NGS in clinical routine it is essential to rely on valid variant calling results – for high- as well as low-frequency variants. Methods and results: In close collaboration with medical and biological experts, we developed a variant calling approach called appreci8 - A Pipeline for PREcise variant Calling Integrating 8 tools. Our pipeline combines and filters the output of 8 previously evaluated open-source variant calling tools, using a novel artifact- and polymorphism score. The algorithm aims at reproducing a biologist's manual work when investigating a raw list of calls. Detailed analysis of 7 well-characterized targeted sequencing data sets (678 samples) shows that appreci8 succeeds in calling variants with allelic frequencies at 1% with high sensitivity (0.93 to 1.00) and high PPV (0.65 to 1.00). Appreci8 is able to outperform every individual tool as well as alternative approaches for combining the 8 tools (best alternative: GATK with sensitivity=0.82-0.95 and PPV=0.31-0.94). Extending the idea of appreci8, we also developed the appreci8R, which is an R-version of our algorithm. The appreci8R provides an additional shiny GUI and checkpoints for re-starting the analysis from intermediate results. Furthermore, all analysis parameters, including the artifact- and polymorphism score, are user-definable. Even the combination and number of input tools can be defined without restriction. Comparing appreci8 and the appreci8R we observe highly comparable performance of both approaches. Appreci8 is available via Docker (https://hub.docker.com/r/wwuimi/appreci8/), the appreci8R is available via Bioconductor (http://www.bioconductor.org/packages/release/ bioc/html/appreci8R.html). Discussion: Appreci8 and the appreci8R provide a novel approach for accurate calling of high- as well as low-frequency variants in NGS data.