A nice overview of the [https://doi.org/10.2217/pgs-2016-0023 Requirements for comprehensive pharmacogenetic genotyping platforms] was published by ''Volker Lauschke et al.'' They claim that rare variants account for 30-40% of functional variability in PGx. However they argue that pre-emptive PGx should only include validated variants, and rare variants should be investigated only when the patient experience unexpected drug response.

* Short read NGS requires a priori knowledge Bad coverage of likelihood pharmacogenes. Notably, the lack of particular haplotypesintronic variants in WES.* Haplotype calling is challenging due to short reads. NGS requires ''In in silico'' haplotype estimation can e.gSince GATK 3.3, the [https://software.broadinstitute.org/gatk/documentation/tooldocs/3.8-0/org_broadinstitute_gatk_tools_walkers_haplotypecaller_HaplotypeCaller.php Haplotype Caller] assemble contigs of 300 base pairs, but it does not consider the phasing of the reads. Haplotype phasing can be performed obtained by Trio analysis, or estimated by imputation methods [http://faculty.washington.edu/browning/beagle/beagle.html Beagle]or [http://dx.doi.org/10.1038/ng.3679 Eagle2] or [https://doi.org/10.1371/journal.pgen.1004234 SHAPEIT]. Other possible options at Oslo University Hospital are long read technology from [https://www.sequencing.uio.no/services/pacbio.html PacBio] or synthetic long reads from [https://www.sequencing.uio.no/news/2019/10x%20Genomics%20is%20here%21 10x Genomics]* Variants Difficult variant calling in homologous regions are hard , such as CYP2D6. I.e. regions with copy number variations (CNV) or pseudogenes. The CYP2D6 genotyping tool that was used by [https://github.com/PharmGKB/PharmCAT/wiki PharmCAT] is [https://www.nature.com/articles/npjgenmed201639 Astrolabe]. We intend to captureuse the more recent program [https://github.com/inumanag/aldy Aldy].* HLA-typing require special software, e.gThere are many options. [https://doi.org/10.1002/cpt.411 Yang et al. ] proposed to use [https://software.broadinstitute.org/cancer/cga/polysolver Polysolver] for whole exome sequencing (WES) or [https://github.com/FRED-2/OptiType OptiType] for whole genome sequencing(WGS). [https://doi.org/10.1101/356204 Reisberg et al.] proposed [https://doi.org/10.1371/journal.pone.0064683 SNP2HLA] for WGS. The candidate that we are investigating closer is [https://github.com/humanlongevity/HLA xHLA]. * New variants are discovered, and needs to be [[gene function|functionally assessed]]. ==Solutions==[[PGx in Estonia|A solution for genotyping biobank data]] has been investigated by ''Reisberg et al.'' in their article [https://doi.org/10.1101/356204 Translating genotype data of 44,000 biobank participants into clinical pharmacogenetic recommendations: challenges and solutions]. An excellent solution for genotyping a collection of ADME (absorption, distribution, metabolism and elimination) genes for individual patients is Aldy, which is fast, accurate and relatively simple to use. ==Interesting publications=={| class="wikitable sortable"|-! Institution !! Article !! Comments|-| St. Jude Children’s Research Hospital || [https://www.ncbi.nlm.nih.gov/pubmed/27311679 Comparison of Genome Sequencing and Clinical Genotyping for Pharmacogenes] || WES and WGS can be used for PGx with ''in silico'' CNV calling and HLA calling|-| National Human Genome Research Institute || [https://www.ncbi.nlm.nih.gov/pubmed/27537706 Assessing the capability of massively parallel sequencing for opportunistic pharmacogenetic screening] || Suggests developing tools for PGx based on WES and WGS|-| University of Washington || [https://doi.org/10.1097/FPC.0000000000000202 PGRNseq: A Targeted Capture Sequencing Panel for Pharmacogenetic Research and Implementation] || Targeted PGx panel with 84 genes|-| University of Tartu || [https://doi.org/10.1101/356204 Translating genotype data of 44,000 biobank participants into clinical pharmacogenetic recommendations: challenges and solutions] || Well-described pipeline for PGx on biobank data. WES cannot be used for PGx (important variants missing, imputation and CNV calling difficult)|}