ClinPharmSeq: A targeted sequencing panel for clinical pharmacogenetics implementation

Abstract

The accurate identification of genetic variants contributing to therapeutic drug response or adverse effects is the first step in implementation of precision drug therapy. Targeted sequencing has recently become a common methodology for large-scale studies of genetic variation thanks to its favorable balance between low cost, high throughput, and deep coverage. Here, we present ClinPharmSeq, a targeted sequencing panel of 59 genes with associations to pharmacogenetic (PGx) phenotypes, as a platform to explore the relationship between drug response and genetic variation, both common and rare. For validation, we sequenced DNA from 64 ethnically diverse Coriell samples with ClinPharmSeq to call star alleles (haplotype patterns) in 27 genes using the bioinformatics tool PyPGx. These reference samples were extensively characterized by multiple laboratories using PGx testing assays and, more recently, whole genome sequencing. We found that ClinPharmSeq can consistently generate deep-coverage data (mean = 274x) with high uniformity (30x or above = 94.8%). Our genotype analysis identified a total of 185 unique star alleles from sequencing data, and showed that diplotype calls from ClinPharmSeq are highly concordant with that from previous publications (97.6%) and whole genome sequencing (97.9%). Notably, all 19 star alleles with complex structural variation including gene deletions, duplications, and hybrids were recalled with 100% accuracy. Altogether, these results demonstrate that the ClinPharmSeq platform offers a feasible path for broad implementation of PGx testing and optimization of individual drug treatments.

Fig 1. Major design strategies used to construct ClinPharmSeq.

Fig 2. Variant call concordance between WGS and ClinPharmSeq.

Fig 3. Concordance of diplotype calls for 27 pharmacogenes between WGS, ClinPharmSeq, and previous studies.

Fig 4. Examples of SVs detected with WGS and ClinPharmSeq.

WGS data are shown in the left panels (A, C, and E) while ClinPharmSeq data are shown in the right panels (B, C, and D). Each panel contains a copy number profile and an allele fraction profile created by PyPGx. (A-B) CYP2B7/CYP2B6 hybrid in African sample NA19178 with a CYP2B6*6/*29 diplotype. (C-D) Gene duplication in African sample NA18861 with a CYP2A6*1x2/*25 diplotype. (E-F) Complex CYP2D6/CYP2D7 hybrid in East Asian sample NA18526 with a CYP2D6*1/*36x2+*10 diplotype.

Fig 5. Examples of diplotype discrepancy caused by difference in SV interpretation between this study and previous publications.

WGS data are shown in the left panels (A, C, and E) while ClinPharmSeq data are shown in the right panels (B, C, and D). Each panel contains a copy number profile and an allele fraction profile created by PyPGx. (A-B) East Asian sample NA18540 was previously identified to have three CYP2D6 gene copies in total with a CYP2D6(*36+)10/*41 diplotype, but this study identified four gene copies with a CYP2D6*36x2+*10/*41 diplotype. (C-D) African sample NA19908 was previously suggested to have a CYP2E1*7x2/*7x2 diplotype (i.e. allele fraction ratio of 2:2), while this study found evidence of a CYP2E1*7/*7x3 diplotype (i.e. allele fraction ratio of 1:3). (E-F) East Asian sample HG00436 was previously genotyped to have a combination of one known SV (CYP2A6*4) and one novel SV (CYP2A6*1+*S6), while in this study PyPGx produced an ‘Indeterminate’ diplotype call because it could also be just one novel SV, which would be a more parsimonious explanation.

Fig 6. Distribution of predicted phenotypes for nine pharmacogenes with a CPIC genotype-phenotype table.

WGS data (N = 70) is shown as a representative example.