Approved Research

Using UK Biobank data to support discovery research and drug development at Chugai

Lay summary

Recent  studies have shown that lack of efficacy is the most common cause of failure in drug development. Some companies have increased the success rate of their clinical trials by selecting target molecules with strong evidence ("Right targets") and targeting patient populations with efficacy expected high ("Right patients"). Genetics is a useful resource to discover "Right targets" as it allows us to unravel causal relationship between target and disease. It was reported that drug targets with genetic evidence have higher success rates. Understanding of disease heterogeneity is important to design "Right patients". UK Biobank has diverse baseline/longitudinal data (e.g., EHR (Electronic Health Record)) which help us dissect disease heterogeneity. Furthermore, integration with genomic data might help us understand biological mechanism behind the disease heterogeneity.

In this research project, we will utilize UK Biobank's large cohort data for two main objectives: (Objective 1) exploration of drug targets based on genetic evidence, and (Objective 2) creation of hypothesis on biological mechanisms behind disease heterogeneity.

For the Objective 1 (Identification of right drug targets) we plan the following analyses. (1) Exploration of novel drug targets (e.g., genes, pathways and cell-types) underlying disease risk / disease conditions by using various approaches such as GWAS (Genome-Wide Association Study) and TWAS (Transcriptome-Wide Association Study). (2) Validating our drug targets and hypothetical biological mechanisms in human by methods such as analysis of rare high-impact variants on molecules of our interests and causal inference methods like Mendelian randomization.

For the Objective 2 and 3 (Understanding of disease heterogeneity, stratifying patient populations and finding biomarkers) we plan the following analyses. (1) Exploration of genomic characteristics from patients stratified by traits of our interests (e.g., comorbidity, number of revision surgeries, traits defined from clinical images). (2) Unsupervised clustering based on longitudinal clinical data and genomic data to identify patient populations which are characteristic both clinically and genetically. We will test both publicly available machine learning models and in-house developed model for the unsupervised patient stratification.