Dataset Information

A systematic comparison of pharmacogene star allele calling bioinformatics algorithms: a focus on CYP2D6 genotyping.

ABSTRACT: Genetic variation in genes encoding cytochrome P450 enzymes has important clinical implications for drug metabolism. Bioinformatics algorithms for genotyping these highly polymorphic genes using high-throughput sequence data and automating phenotype prediction have recently been developed. The CYP2D6 gene is often used as a model during the validation of these algorithms due to its clinical importance, high polymorphism, and structural variations. However, the validation process is often limited to common star alleles due to scarcity of reference datasets. In addition, there has been no comprehensive benchmark of these algorithms to date. We performed a systematic comparison of three star allele calling algorithms using 4618 simulations as well as 75 whole-genome sequence samples from the GeT-RM project. Overall, we found that Aldy and Astrolabe are better suited to call both common and rare diplotypes compared to Stargazer, which is affected by population structure. Aldy was the best performing algorithm in calling CYP2D6 structural variants followed by Stargazer, whereas Astrolabe had limitations especially in calling hybrid rearrangements. We found that ensemble genotyping, characterised by taking a consensus of genotypes called by all three algorithms, has higher haplotype concordance but it is prone to ambiguities whenever complete discrepancies between the tools arise. Further, we evaluated the effects of sequencing coverage and indel misalignment on genotyping accuracy. Our account of the strengths and limitations of these algorithms is extremely important to clinicians and researchers in the pharmacogenomics and precision medicine communities looking to haplotype CYP2D6 and other pharmacogenes using high-throughput sequencing data.

SUBMITTER: Twesigomwe D

PROVIDER: S-EPMC7398905 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A systematic comparison of pharmacogene star allele calling bioinformatics algorithms: a focus on CYP2D6 genotyping.

Twesigomwe David D Wright Galen E B GEB Drögemöller Britt I BI da Rocha Jorge J Lombard Zané Z Hazelhurst Scott S

NPJ genomic medicine 20200803

Genetic variation in genes encoding cytochrome P450 enzymes has important clinical implications for drug metabolism. Bioinformatics algorithms for genotyping these highly polymorphic genes using high-throughput sequence data and automating phenotype prediction have recently been developed. The CYP2D6 gene is often used as a model during the validation of these algorithms due to its clinical importance, high polymorphism, and structural variations. However, the validation process is ...[more]

PMID: 32789024

Similar Datasets

Project description:BackgroundThe CYP2D6 gene locus has been extensively studied over decades, yet a portion of variability in CYP2D6 activity cannot be explained by known sequence variations within the gene, copy number variation, or structural rearrangements. It was proposed that rs5758550, located 116 kb downstream of the CYP2D6 gene locus, increases gene expression and thus contributes to variability in CYP2D6 activity. This finding has, however, not been validated. The purpose of the study was to address a major technological barrier, i.e., experimentally linking rs5758550, also referred to as the "enhancer" single-nucleotide polymorphism (SNP), to CYP2D6 haplotypes >100 kb away. To overcome this challenge is essential to ultimately determine the contribution of the "enhancer" SNP to interindividual variability in CYP2D6 activity.MethodsA large ethnically mixed population sample (n=3,162) was computationally phased to determine linkage between the "enhancer" SNP and CYP2D6 haplotypes (or star alleles). To experimentally validate predicted linkages, DropPhase2D6, a digital droplet PCR (ddPCR)-based method was developed. 10X Genomics Linked-Reads were utilized as a proof of concept.ResultsPhasing predicted that the "enhancer" SNP can occur on numerous CYP2D6 haplotypes including CYP2D6*1, *2, *5, and *41 and suggested that linkage is incomplete, i.e., a portion of these alleles do not have the "enhancer" SNP. Phasing also revealed differences among the European and African ancestry data sets regarding the proportion of alleles with and without the "enhancer" SNP. DropPhase2D6 was utilized to confirm or refute the predicted "enhancer" SNP location for individual samples, e.g., of n=3 samples genotyped as *1/*41, rs5758550 was on the *41 allele of two samples and on the *1 allele of one sample. Our findings highlight that the location of the "enhancer" SNP must not be assigned by "default." Furthermore, linkage between the "enhancer" SNP and CYP2D6 star allele haplotypes was confirmed with 10X Genomics technology.ConclusionsSince the "enhancer" SNP can be present on a portion of normal, decreased, or no function alleles, the phase of the "enhancer" SNP must be considered when investigating the impact of the "enhancer" SNP on CYP2D6 activity.

Project description:Insertions/deletions (indels) are the second most common type of genomic variant and the most common type of structural variant. Identification of indels in next generation sequencing data is a challenge, and algorithms commonly used for indel detection have not been compared on a research cohort of human subject genomic data. Guidelines for the optimal detection of biologically significant indels are limited. We analyzed three sets of human next generation sequencing data (48 samples of a 200 gene target exon sequencing, 45 samples of whole exome sequencing, and 2 samples of whole genome sequencing) using three algorithms for indel detection (Pindel, Genome Analysis Tool Kit's UnifiedGenotyper and HaplotypeCaller).We observed variation in indel calls across the three algorithms. The intersection of the three tools comprised only 5.70% of targeted exon, 19.52% of whole exome, and 14.25% of whole genome indel calls. The majority of the discordant indels were of lower read depth and likely to be false positives. When software parameters were kept consistent across the three targets, HaplotypeCaller produced the most reliable results. Pindel results did not validate well without adjustments to parameters to account for varied read depth and number of samples per run. Adjustments to Pindel's M (minimum support for event) parameter improved both concordance and validation rates. Pindel was able to identify large deletions that surpassed the length capabilities of the GATK algorithms.Despite the observed variability in indel identification, we discerned strengths among the individual algorithms on specific data sets. This allowed us to suggest best practices for indel calling. Pindel's low validation rate of indel calls made in targeted exon sequencing suggests that HaplotypeCaller is better suited for short indels and multi-sample runs in targets with very high read depth. Pindel allows for optimization of minimum support for events and is best used for detection of larger indels at lower read depths.

Dataset Information

A systematic comparison of pharmacogene star allele calling bioinformatics algorithms: a focus on CYP2D6 genotyping.

Publications

A systematic comparison of pharmacogene star allele calling bioinformatics algorithms: a focus on <i>CYP2D6</i> genotyping.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets