Challenges and opportunities associated with rare-variant pharmacogenomics

Abstract

Advances in sequencing technologies have facilitated the discovery of hundreds and thousands of human genetic variants, thus providing rich data resources to study the genetic complexity of pharmacogenes.Commonly used computational methods can be used to identify disease-causing variants; however, they underperform for variants that impact on drug response. So far, only a few algorithms for pharmacogenetic variant effect predictions have been developed. However, their predictive power is expected to be further improved by emerging massively parallel experimental assays and artificial intelligence-based methods.Combining pharmacogenomic data with extensive electronic medical records provides exciting opportunities for the identification of rare-variant associations.Significant gaps remain in the clinical implementation of functional rare genetic variants for patient care, even when considering polygenetic associations. Efforts from multiple parties including stakeholders from different healthcare areas are required. Recent advances in next-generation sequencing (NGS) have resulted in the identification of tens of thousands of rare pharmacogenetic variations with unknown functional effects. However, although such pharmacogenetic variations have been estimated to account for a considerable amount of the heritable variability in drug response and toxicity, accurate interpretation at the level of the individual patient remains challenging. We discuss emerging strategies and concepts to close this translational gap. We illustrate how massively parallel experimental assays, artificial intelligence (AI), and machine learning can synergize with population-scale biobank projects to facilitate the interpretation of NGS data to individualize clinical decision-making and personalized medicine. Recent advances in next-generation sequencing (NGS) have resulted in the identification of tens of thousands of rare pharmacogenetic variations with unknown functional effects. However, although such pharmacogenetic variations have been estimated to account for a considerable amount of the heritable variability in drug response and toxicity, accurate interpretation at the level of the individual patient remains challenging. We discuss emerging strategies and concepts to close this translational gap. We illustrate how massively parallel experimental assays, artificial intelligence (AI), and machine learning can synergize with population-scale biobank projects to facilitate the interpretation of NGS data to individualize clinical decision-making and personalized medicine. Pharmacological twin studies in which drug pharmacokinetics are compared between monozygotic and dizygotic twin pairs constitute important trial designs to assess the importance of heritable factors for drug disposition. Seminal studies indicated that 34–98% of interindividual differences in drug exposure are heritable, depending on the probe substrate [1.Lauschke V.M. Ingelman-Sundberg M. Prediction of drug response and adverse drug reactions: from twin studies to next generation sequencing.Eur. J. Pharm. Sci. 2019; 130: 65-77Crossref PubMed Scopus (36) Google Scholar]. However, genetic effects based on common polymorphisms can only explain part of this heritable variation, and a substantial fraction of the genetically encoded variability in drug pharmacokinetics remains to be elucidated [2.Matthaei J. et al.Heritability of metoprolol and torsemide pharmacokinetics.Clin. Pharmacol. Ther. 2015; 98: 611-621Crossref PubMed Scopus (48) Google Scholar]. Over the past few decades major advances in DNA sequencing technologies have revolutionized our understanding of the human genome. Since the sequencing and assembly of the first human genome, population-scale next-generation sequencing (NGS; see Glossary) projects have unveiled millions of genetic variants, including single-nucleotide variants (SNVs), small insertions and deletions (indels), as well as structural variants that can span up to multiple Mb. Aggregation of these sequencing data into consolidated publicly available variant repositories, such as the Genome Aggregation Database (gnomAD) [3.Karczewski K.J. et al.The mutational constraint spectrum quantified from variation in 141,456 humans.Nature. 2020; 581: 434-443Crossref PubMed Scopus (4070) Google Scholar], has provided comprehensive large-scale resources for pharmacogenomic studies. Recent analyses of such population-scale datasets have indicated that pharmacokinetic genes, which are commonly not disease-associated and lack endogenous substrates, contain a plethora of rare genetic variations with unknown functional consequences [4.Ingelman-Sundberg M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google Scholar, 5.Wright G.E.B. et al.The global spectrum of protein-coding pharmacogenomic diversity.Pharmacogenom. J. 2018; 18: 187-195Crossref PubMed Scopus (54) Google Scholar, 6.Ahn E. Park T. Analysis of population-specific pharmacogenomic variants using next-generation sequencing data.Sci. Rep. 2017; 7: 8416Crossref PubMed Scopus (18) Google Scholar]. Similar genetic complexity was identified in pharmacodynamic [7.Schärfe C.P.I. et al.Genetic variation in human drug-related genes.Genome Med. 2017; 9: 117Crossref PubMed Scopus (82) Google Scholar, 8.Nelson M.R. et al.An abundance of rare functional variants in 202 drug target genes sequenced in 14,002 people.Science. 2012; 337: 100-104Crossref PubMed Scopus (507) Google Scholar, 9.Zhou Y. et al.Rare genetic variability in human drug target genes modulates drug response and can guide precision medicine.Sci. Adv. 2021; 7: PubMed Scopus Google and genes M. et al.The variability of genetic factors 2021; PubMed Scopus Google Scholar, et rare variant interpretation in of and Med. 2021; PubMed Scopus Google Scholar, DNA variation human of and Scopus Google Scholar]. multiple important this variability explain part of the pharmacogenetic can from variations in the of experimental how can pharmacogenomic sequencing to guide personalized the between predictions and clinical of the of medical and has been by and et of technologies to precision 2020; PubMed Scopus Google Scholar]. in experimental and computational strategies have been to facilitate the of to clinical the discuss with the identification of genetic variants as well as their functional interpretation and for decision-making emerging and clinical strategies for that to the translational gap. NGS a of technologies in which the sequencing of between and in is massively with millions to of which the generation of of in a on generation sequencing advances and PubMed Scopus Google Scholar]. of sequencing between sequencing sequencing and the human genome, including extensive and of functional on only of the human et of human variation from population-scale PubMed Scopus Google Scholar], sequencing the of important variations which are in However, multiple important in such as and of genetic the of drug sequencing on pharmacogenetic including and a few and of the of the human a on to Med. 2017; Google sequencing the of when considering the of pharmacogenetic variants are when only a few are to be and sequencing are based on individual of that are only in which are a genome. in pharmacogenetic with of these are to the accurate of variants from in these genes can be for multiple including and of commonly used as well as genes in the human which are with adverse adverse exposure to multiple such as V.M. et for improved drug and J. 2017; PubMed Scopus Google Scholar]. computational such as et and accurate from next-generation data with Sci. 2017; PubMed Scopus Google can these they are and have not their into sequencing methods that are of sequencing to these to et human sequencing and 2020; PubMed Scopus Google Scholar]. sequencing in that the of the is thus the of a to be in a sequencing has been used for the accurate identification of of pharmacogenetic such as and et sequencing for of and variants and 2015; PubMed Google Y. et sequencing of the and of 2019; PubMed Scopus Google Scholar]. of sequencing that the of is not to in sequencing in on for and that lack and et and in sequencing data 2020; PubMed Scopus Google Scholar]. However, as the of sequencing and sequenced that to variant identification in importance to accurate variant interpretation and constitute the to clinical of NGS in in the studies and experimental functional assays in have the methods to the of pharmacogenetic variants Y. et methods for the pharmacogenetic interpretation of next generation sequencing Pharmacol. 2018; 9: PubMed Scopus Google Scholar]. However, are not to the of pharmacogenetic variability they are not to rare-variant effects Y. et methods for the pharmacogenetic interpretation of next generation sequencing Pharmacol. 2018; 9: PubMed Scopus Google Y. V.M. to assess the functional consequences of rare and pharmacogenetic Pharmacol. Ther. 2021; PubMed Scopus Google Scholar]. experimental strategies are for the of thousands of variants Y. et methods for the pharmacogenetic interpretation of next generation sequencing Pharmacol. 2018; 9: PubMed Scopus Google Y. V.M. to assess the functional consequences of rare and pharmacogenetic Pharmacol. Ther. 2021; PubMed Scopus Google Scholar]. computational methods constitute the only that variant effect predictions at the this a of algorithms have been to variant based on structural of and functional Y. et methods for the pharmacogenetic interpretation of next generation sequencing Pharmacol. 2018; 9: PubMed Scopus Google Scholar]. these algorithms can from multiple major remain which the of commonly used in for pharmacogenetic well on their when on variant at in part of in et on functional impact of PubMed Scopus Google et al.The of used to the impact of variants is by of 2015; PubMed Scopus Google Scholar]. algorithms are to variant the of the predictions are based on which is based on the that variants in are and are thus to be However, are and the of pharmacogenetic variants is thus only by [4.Ingelman-Sundberg M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google Scholar]. to this variant effect common variants identified in large-scale sequencing projects as data for variants; however, pharmacogenetic such as the and are common in the and are thus in the the that they impact on Y. et of a of population-scale sequencing Pharmacol. Ther. 2017; PubMed Scopus Google Scholar]. a such algorithms on pharmacogenetic methods based on the predictive on pharmacogenomic variant functional methods that such as of and functional Y. et al.An to assess the functional impact of pharmacogenetic J. 2019; PubMed Scopus Google Scholar]. that are for analyses of pharmacogenetic variants are to the functional of and pharmacogenetic the of in assays the functional interpretation of the tens of thousands of rare pharmacogenetic variants identified in population-scale human sequencing this methods have been that the study of thousands of variants parallel assays and 2019; PubMed Scopus Google Scholar]. massively parallel assays J. et identification of human 2019; PubMed Scopus Google and mutational mutational a of PubMed Scopus Google for the of variants and protein-coding mutational has as a to the functional consequences of a of in a of M. et assays to variation in Pharmacol. Ther. 2019; PubMed Scopus Google V.M. Ingelman-Sundberg M. strategies to the between pharmacogenomic and clinical Med. 2020; PubMed Scopus Google Scholar]. mutational is based on of with a and of variant effects using a that abundance mutational a of PubMed Scopus Google Scholar]. has been for variants in the genes the and as well as in which important of response. of the of thousands of variant is provided by a based on a probe et parallel of variant and J. 2021; PubMed Scopus Google Scholar]. variant used variants to that of to the of variant and was by NGS et parallel variant with Sci. 2020; PubMed Scopus Google Scholar]. was using a based on of a which is with a et of variant abundance and and human variant 2020; PubMed Google Scholar]. these studies although mutational functional of variants on the of assays be to remains a major this in assays based on of variants as with a have been which are based on the that of the to of the variant in of et and mutational and functional of Sci. 2020; PubMed Scopus Google et of variant abundance by massively parallel 2018; PubMed Scopus Google Scholar]. of this is that the experimental can be used for the of different However, of variant abundance with functional is and abundance only of the functional variation et parallel of variant and J. 2021; PubMed Scopus Google Scholar]. assays based on functional mutational have been to genes drug genes and genes with drug response have provided into and the the functional variability in these genes these variant to well as resources for computational variant effect that in can to from a of genes to the Similar have to be for the of variant effects in disease-associated genes et with large-scale 2021; 12: PubMed Scopus Google et to from mutational Sci. 2021; Scopus Google Scholar], and emerging strategies for the of machine learning for pharmacogenetic are provided mutational studies for large-scale pharmacogenomic variant of based on the of based on a to and of variants abundance for of the et parallel of variant and J. 2021; PubMed Scopus Google generation of of variants to and sequencing to variant effects on variants et and mutational and functional of Sci. 2020; PubMed Scopus Google variants et and mutational and functional of Sci. 2020; PubMed Scopus Google variants et and of mutational for variant 2021; PubMed Scopus Google based on the of a based on a to and of of variants et of variant abundance and and human variant 2020; PubMed Google based on to and of the in of variants resulted in et of variant abundance by massively parallel 2018; PubMed Scopus Google based on based on to and of variants have et parallel variant with Sci. 2020; PubMed Scopus Google in a provides exciting opportunities to the of the functional consequences of rare learning is a of that and from and to predictions in learning is learning from with a of et learning in J. Med. 2019; PubMed Scopus Google Scholar]. this strategies such as are learning in and 2015; PubMed Scopus Google Scholar]. learning of with a such as drug response for and the of these for et to data personalized Med. 2018; PubMed Scopus Google Scholar]. learning from data commonly from technologies of J. et variant with of 2021; PubMed Scopus Google Scholar]. methods and in a of and of the effects of rare However, based on the power of the a the predictive of So far, different have been in as has been used to variant effect a computational for pharmacogenetic and was based on the of algorithms Y. et al.An to assess the functional impact of pharmacogenetic J. 2019; PubMed Scopus Google Scholar]. for and variants and provided that with Y. et al.An to assess the functional impact of pharmacogenetic J. 2019; PubMed Scopus Google Scholar]. the of rare variants to the genetically encoded functional variability in was estimated to be M. et al.Rare genetic variants in and can be important of interindividual differences in drug Med. 2017; PubMed Scopus Google Scholar]. However, the importance of rare variants between based on the of human that of the and drug on between and variants in a human [4.Ingelman-Sundberg M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google Y. et predictions of and using a J. 2020; PubMed Scopus (18) Google Scholar, V.M. variability and of the human 2019; PubMed Scopus Google Scholar, V.M. genetic of the human 2019; PubMed Scopus Google Scholar, et and variability of human 2020; PubMed Scopus Google Scholar]. are to these functional variants, rare variants account for of this with the of for which rare variants the rare account for the of the variation [4.Ingelman-Sundberg M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google Y. et predictions of and using a J. 2020; PubMed Scopus (18) Google Scholar, V.M. variability and of the human 2019; PubMed Scopus Google Scholar, V.M. genetic of the human 2019; PubMed Scopus Google Scholar, et and variability of human 2020; PubMed Scopus Google Scholar]. to and of a to sequencing data resulted in the identification of variants that are to et machine for the computational functional of pharmacogenomic Genomics. 2021; PubMed Scopus Google of rare variants to genetically encoded functional variation in of variants of functional variants of functional variation to rare variants as variants with a of M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google M. et al.Integrating rare genetic variants into pharmacogenetic drug response predictions.Hum. Genomics. 2018; 12: 26Crossref PubMed Scopus (127) Google Y. et predictions of and using a J. 2020; PubMed Scopus (18) Google Y. et predictions of and using a J. 2020; PubMed Scopus (18) Google V.M. variability and of the human 2019; PubMed Scopus Google V.M. genetic of the human 2019; PubMed Scopus Google et and variability of human 2020; PubMed Scopus Google variants as variants with a of in a multiple effect have been developed. was on variants with functional in data and variants with et variant to identify of Pharmacol. Ther. 2018; PubMed Scopus Google Scholar]. of a based on sequencing data and of pharmacokinetics from a to be that for the of response to the and in patient M. et drug response from sequencing data.Sci. Med. 2021; PubMed Scopus Google Scholar]. that can predictions compared to the predictions based on are in with that learning and functional variant using a can in improved predictions et learning of 2020; PubMed Scopus Google Scholar]. a and response to in the of major using and data of genetic polymorphisms with in studies for response as et of a with Pharmacol. Ther. 2019; PubMed Scopus Google Scholar]. used of multiple variants including and genetic variants, based on J. et on learning in 2019; PubMed Scopus Google Scholar]. the provides that the of individual variants can be based on based on the Google was to the of from their with computational methods and compared to J. et accurate with 2021; PubMed Scopus Google Scholar]. of based on functional analyses of rare genetic variants and for understanding and as well as and compared to methods based on et generation predictions and genetic variant 2021; PubMed Scopus Google Scholar]. However, experimental of for variant effect predictions have been and available data of a with predictions et to the impact of on and 2021; Scholar]. into the of of structural functional effects of genetic variants are of pharmacogenetic variability on the interpretation of individual variants not in in of with functional consequences have been that be by analyses et with of Pharmacol. Ther. 2021; PubMed Scopus Google Scholar], of in has not been of the of sequencing sequencing these and accurate variant advances in computational population-scale analyses et sequencing of the PubMed Scopus Google Scholar]. of this to sequencing data from resulted in the identification of of clinical et sequencing of the PubMed Scopus Google Scholar]. to variability in genes such as factors can and drug of such effects is provided by the identification of between the variant in a that the of and important and of the and that was identified in a with data et of a with in a study for 2020; PubMed Scopus (18) Google et al.The and in Pharmacol. Ther. PubMed Scopus Google Scholar]. and study designs that are to effects to and explain part of the pharmacogenetic which be by individual variants for considering genetic is genetic pharmacogenetic variants are available for the of and et for and and Pharmacol. Ther. 2021; PubMed Scopus Google and and et for and and Pharmacol. Ther. 2021; PubMed Scopus Google such can account for rare variants et and variability of human 2020; PubMed Scopus Google Scholar], such have their into pharmacogenetic the identification and functional interpretation of a pharmacogenetic important to and how this can be used to a pharmacogenetic to be to be as a of drug response a of at the and level be the first to be when considering the implementation of pharmacogenetic in clinical is the level of that a variant can drug response a plethora of pharmacogenomic have been the from effect be and lack experimental the and in the of pharmacogenomic can these to et of studies of response to PubMed Scopus Google et and of response in J. 2017; PubMed Scopus Google however, to substantial in study and of is of the thousands of in the only variants have to and drug and by such as the and the from pharmacogenetic such as the and the et of and of pharmacogenomic drug opportunities for Pharmacol. Ther. 2020; PubMed Scopus Google Scholar]. these to common variants, and variants identified by NGS are not even for variants with and with drug response can be variability although genetic variation in can exposure and of the and at the a substantial fraction of the interindividual variability in the of these remains et of on exposure and of and a 2019; PubMed Scopus Google Scholar]. a the of individual patient and thus clinical of variant as a for the of a clinical the of interindividual differences up different strategies to and adverse is the of for this are the in of the with to as from and the in from with and for which pharmacogenomic have been et of the of Ther. 2018; PubMed Scopus Google Scholar]. the for can be by and and this further personalized drug response predictions et of and is by genetic factors and in human PubMed Scopus Google Scholar, et of in human Pharmacol. PubMed Scopus Google Scholar, et in of a impact of genetic and J. PubMed Scopus Google Scholar]. of the of technologies are the of and data that resulted in the identification of in the variability of the response et in of to Pharmacol. Ther. PubMed Scopus Google Scholar], variants in that impact on of the Y. et and drug and J. Pharmacol. PubMed Scopus Google Scholar], and genetic variants in and that are with et of identify of pharmacogenetic in a 12: PubMed Scopus Google Scholar]. to the implementation of pharmacogenetic in clinical on the that a of healthcare studies in remain the available data that was for of the when the of is the of the are available a Y. et of the on the of for Med. 2020; PubMed Scopus Google Scholar]. of NGS for pharmacogenomic have not been is not to be for the of at this that sequencing and data patient et al.The of a study in and rare from a in the Med. 2020; PubMed Scopus Google Scholar]. are for that are with such as the of and the identification of for such as of NGS to of that sequencing and when is at the of pharmacogenomic that and have available at the of and this of the of a a clinical implementation of clinical are that the and of sequencing to computational variant however, clinical which have been commonly used to consequences of rare genetic variants, are for pharmacogenetic for rare variants, trial designs that a clinical for of pharmacogenomic variants for individual a trial the of sequencing the with sequencing to the of pharmacogenomic methods to the impact of variants of unknown this variant is to for to variations which are available for for and However, in the of such facilitate the clinical of rare pharmacogenetic effects of rare variants are to study of to be in the analyses to of variant identify rare-variant effects the of of patient to electronic records can be contain rich data and the the and provides a for patient to et al.The a of to electronic medical records data for Med. Scholar]. with and the at in the as part of DNA from thus comprehensive of DNA are available and are the of clinical data provides and genetic data from et of clinical the PubMed Scopus Google Scholar]. the DNA and with to for et al.The with and 2018; PubMed Scopus Google Scholar]. this rich studies of rare protein-coding variants with and a of for with a of that for and drug et al.Rare variant to human in 2021; PubMed Scopus Google Scholar]. the of rare variants to human and the of rare variants for the of genetically drug is that the between and rare variants can our understanding of drug response and adverse drug et studies as a to precision PubMed Scopus Google et of genetic variants with in electronic medical PubMed Scopus Google Scholar]. from the study that sequencing of to from at variant in of et al.Genetic variation the data from the Pharmacol. Ther. PubMed Scopus Google Scholar]. sequencing of resulted in the discovery of of rare variants in a of was for variant that to rare variants in genes can be by and functional in et variant and in the sequencing PubMed Scopus Google Scholar]. of population-scale of with drug and adverse data to the identification of response T. et al.Genetic variation in the study of adverse drug effects using electronic J. 2019; PubMed Scopus Google Scholar]. and and identified multiple that population-scale with has power to between drug response and genetic further for clinical of rare variants is the of experimental to genetic and human in and rare that can personalized the of a of this the discovery of in with the identification of with and and genetically et biobank and clinical Med. 2020; PubMed Scopus Google Scholar]. of these trial designs have been for the study of rare variants the of response to in a are in medical for to assess for disease-causing variants trial concepts are et and for in the healthcare Pharmacol. Ther. 2021; PubMed Scopus Google Scholar]. on and multiple the of is a to in genetically based to data from can be to trial has been to to rare variants in the et al.The of a global Med. 2020; PubMed Scopus Google Scholar]. data clinical such as of and can be to the and of in the this can be by assays using resources to and to response for individual which has been used for the of et personalized of Rep. 2019; PubMed Scopus Google Scholar]. on the of further this with important for the of by rare genetic variants and the of NGS technologies have to the that they can in be used for sequencing of the of individual a few the has from of technologies the interpretation and of the variant remains at how the rare and variants that be identified in a be for personalized predictions of drug response and computational algorithms on pharmacogenetic datasets have that methods can the of variant are to further with the of massively parallel experimental assays that the generation of extensive of experimental variant of variant in with improved structural can be expected to further the of for pharmacogenetic these to the functional interpretation of that at understanding pharmacogenetic variation are advances are by the of trial designs to as well as by the of data that data with rich consolidated in population-scale although these for the functional of rare pharmacogenetic variants, remains a for and clinical to the of rare-variant can for and healthcare the of rare pharmacogenomic variants personalized predictions of drug response and pharmacogenomic sequencing a of healthcare assays pharmacogenomic predictions of variants can variant effects be is the of variations to the pharmacogenomic the of rare pharmacogenomic variants personalized predictions of drug response and pharmacogenomic sequencing a of healthcare assays pharmacogenomic predictions of variants can variant effects be is the of variations to the pharmacogenomic was by the and the and and are and of is and of of is of healthcare experimental that can the of thousands of variants in a in a in electronic of a medical including medical a publicly available for genetic variants in different identified by major sequencing and sequencing a study that variants the human for with a of analyses that the between a and a of variants on the a of machine learning that thousands of that data between to between to as NGS a of massively parallel sequencing technologies that sequencing of to into the effects of a drug on a a that is to drug a genetic that the and of a the to the study of between genetic variants and drug to the of a drug which and a genetic variant in a a study that analyses which are with a genetic a clinical trial in which the effect of a are by to a that is used to and a the of of a in genes are to be with data that are used by for and to the can in

Topics & Concepts