Work

Genetically guided drug development

Year: 2023
Type: dissertation
Abstract: Background \n \nAttrition is a major issue in drug development with less than 5% of drug development programmes yielding licensed drugs. Retrospective studies have suggested that human genomic data could be used to help prioritise drug development programmes and reduce the risk of clinical-stage failure. The investment of pharmaceutical companies in healthcare genomic initiatives has been incentivised largely by studies showing that genetically-supported targets would succeed twice as often as those without genetic support, and comparative studies revealing that the effect of licensed drugs on biomarkers and disease endpoints coincide with the observed associations of variants in the genes encoding the corresponding target. However, historically, genome-wide association studies (GWAS) of human diseases and pharmaceutical research and development have largely proceeded independently. Knowledge of the overlap between existing GWAS and current or historical drug development programmes is important to maximise the utility of existing data for repurposing opportunities and mechanism-based adverse effect prediction. Additionally, for novel target identification, questions remain about what type of genomic data is most informative and what methods are most robust. Mendelian randomisation (MR), a genetic epidemiology approach for causal inference, has been used to assess the causal nature of exposures on outcomes. Its application has recently been extended to the evaluation of drug targets against disease (‘drug target MR’). However, very limited validation of the parameters used in drug target MR studies exists across multiple target loci and diseases. \n \nAim \n \nTo investigate the extent to which the spectrum of human diseases has been addressed by genetic analyses, or by drug development, and the degree to which these efforts overlap. To evaluate the genetic support for approved drug target-indication pairs from GWAS and drug target MR applications. \n \nMethods \n \nHuman disease information from the Disease Ontology and drug data from ChEMBL version 25 were used. Genetic associations with diseases and clinical endpoints were sourced from the GWAS Catalog and UK Biobank (through Neale Lab), and genetic associations for circulating protein levels measured by the SomaLogic v4 proteomic platform from the Fenland study and UCLEB Consortium. \n \nI calculated the disease coverage, overlap and divergence of human genetic studies and pharmaceutical research and development. I provided a revised estimate of the value of genetic evidence for drug target-indication pairs in progressing in clinical-phase drug development, and investigated different approaches to assign genetic associations identified by GWAS to causal genes. I evaluated the drug target MR framework with a curated ‘truth’ set of drug target-indication pairs for which genetic associations with the circulating levels of the protein target and the intended indication were available. I applied the drug target MR framework using genetic associations with blood lipids (LDL-cholesterol, HDL-cholesterol and triglycerides) to prioritise drug targets for the treatment and prevention of coronary heart disease. \n \nResults \n \nOnly 9% (953 out of 10,901) of human diseases have been studied by GWAS. Of these, only 369 correspond to diseases with an approved treatment and/or a treatment under clinical or preclinical development, leaving 584 diseases that have been the subject of investigation in GWAS, but which have yet to be investigated in drug development. For those indications that are or have been the subject of clinical phase drug development and have been studied by GWAS, I found that drug target-indication pairings with genetic support are twice more likely to get approved than those without genetic support (2.18; 95%CI: 1.86; 2.51). The evaluation of the drug target MR framework with the subset of target-indication pairings of approved drugs with available genetic associations with the circulating protein levels recapitulated the mechanism of action of up to 13% (16 out of 121) of the drug target gene – indication pairings and returned results in the unanticipated direction of effect for 11% (14 out of 121) of the pairings explored. The systematic application of the biomarker-weighted drug target MR using blood lipid levels robustly identified 30 targets that should be prioritised for the prevention or treatment of coronary heart disease. \n \nConclusion \n \nI identified points of convergence or divergence between genomic research and drug development efforts in the sample space of all the human drug targets and diseases, and demonstrated the utility of GWAS data for drug target identification and validation through the mapping of genetic associations to causal genes and the application of the drug target MR framework. The work of this thesis informs prioritisation strategies in drug development and future research so the investment and impact of human genetic studies can be maximised.
Language: English
Subfield: Genetics
Open Access status: green
OpenAlex ID: W7028734772
Cites:
Cited by:
Related to: 10