ShRangeSim: Simulation of Single Nucleotide Polymorphism Clusters in Next-Generation Sequencing Data

Abstract

Genomic variations are in the focus of research to uncover mechanisms of host-pathogen interactions and diseases such as cancer. Nowadays, next-generation sequencing (NGS) data are analyzed through dedicated pipelines to detect them. Surrogate NGS data in conjunction with genomic variations help to evaluate pipelines and validate their outcomes, fostering selection of proper tools for a given scientific question. I describe how existing approaches for simulating NGS data in conjunction with genomic variations fail to model local enrichments of single nucleotide polymorphisms (SNPs), so called SNP clusters. Two distributions for count data are applied to publicly available collections of genomic variations. The results suggest modeling of SNP cluster sizes by overdispersion-aware distributions.

GeDi: applying suffix arrays to increase the repertoire of detectable SNVs in tumour genomes

“…Since GeDi is a suffix array-based SNV caller, it does not rely on mapping coordinates to detect SNVs, and therefore, should retain sensitivity for SNV detection at complex variant loci. Short Range SNP clusters (SRSC) are variant genomic features formally defined as a genomic loci containing k SNPs within window length W = 100 [1]. SRSC have been shown to induce incorrect mapping coordinates for reads covering them.…”

“…SRSC have been shown to induce incorrect mapping coordinates for reads covering them. Hence, reference-based variant callers are likely to show insensitivity for SRSC detection [1]. We define a somatic version of SRSC, the somatic Short Range SNV Cluster (sSRSC), as genomic loci containing k ≥ 2 SNVs within window length W = 100 occurring in tumour genomes; we define, k-sSRSC as a sSRSC containing k SNVs.…”

“…We generated five 30x simulated datasets containing 500 sSRSC of size 2 ≤ k ≤ 20 of hg19 chromosomes 1, 8, 15, 17 and 22 (randomly chosen) and calculated GeDi's and MuTect's precision and recall when analysing these datasets. Our chosen size distribution 2 ≤ k ≤ 20 reflects the known size distribution of SRSC within the human genome [1]. As were not testing the effect of allele frequency, average allele frequency kept at 50% for all datasets.…”

“…A considerable drawback of reference-based SNV callers is the reduced sensitivity they exhibit in the presence of incorrect mapping coordinates. This limits their ability to characterise complex variant loci (variant loci other than sparsely distributed SNVs) where incorrect mapping coordinates arise frequently [1,2].…”