Mohd Khairul Nizam Mohd Khalid1*, Lua Seok Hian2, Amelia Azman2, Nor Azimah Abdul Azize2, Julaina Abdul Jalil1, Yusnita Yakob2
1. Inborn Errors of Metabolism & Genetics Unit, Nutrition, Metabolism & Cardiovascular Research Centre, Institute for Medical Research, National Institutes of Health, Ministry of Health Malaysia, Setia Alam, 40170 Shah Alam, Selangor, Malaysia
2. Molecular Diagnostics Unit, Specialised Diagnostic Centre, Institute for Medical Research, National Institutes of Health, Ministry of Health Malaysia, Jalan Pahang, 50588 Kuala Lumpur, Malaysia
*Corresponding author: Mohd Khairul Nizam Mohd Khalid.
ABSTRACT
INTRODUCTION: Whole-exome sequencing (WES) remains the most popular platform for many applications, ranging from molecular testing to population genetics. While this works for most clinically relevant genes, the short length of WES becomes a limitation when analysing genes with high sequence similarity to other genomic regions (homologous regions). About 3.5% of exomes are affected by sequence homology, and solving this problem is important to avoid losing important genetic variants in those regions.
OBJECTIVE(S): We aim to compare the performance of two popular exome analysis pipelines, namely DRAGEN and GATK, in two genes containing short homologous regions (GBA1 and GCSH).
MATERIALS & METHODS: FASTQ files for two samples were processed using both DRAGEN version 4.2.7 and GATK version 4.6.1.0, implemented as nf-core/sarek version 3.1.2 using GRCh38 as reference genome. Both Binary Alignment/Map (BAM) and Variant Call Format (VCF) files were inspected manually using Integrative Genomics Viewer (IGV) version 2.18.2.
RESULTS: IGV showed significant homology in exons 10 and 11 of the GBA1 gene and exon 1 of the GCSH gene in both samples. In sample 1, where compound heterozygous variants c.1448T>C and c.1448T>G in the GBA1 gene were expected, DRAGEN detected a homozygous c.1448T>G variant while GATK correctly identified both variants in heterozygous states. When the latest version of DRAGEN (4.3.13) was used, both variants were detected, but the allelic ratio was incorrect (not the expected 50% ratio for heterozygous variants). For sample 2, GATK correctly identified two variants in exon 1 of the GCSH gene, compared to DRAGEN, where only one variant was called despite the region being included in the Browser Extensible Data (BED) file and covered by more than 100 reads.
CONCLUSION: Despite the high accuracy of exome pipelines in variant detection, we observed that their performance was affected by the presence of homologous regions. GATK outperformed DRAGEN in both samples tested, but this is only relevant for homologies spanning a small number of exons.
KEYWORDS: Exome, Homologous Regions, Pipelines, Variant, Performance


