Identification of therapeutic targets from genetic association studies using hierarchical component analysis

Hao Chih Lee, Hao Chih Lee, Osamu Ichikawa, Osamu Ichikawa, Osamu Ichikawa, Benjamin S. Glicksberg, Benjamin S. Glicksberg, Benjamin S. Glicksberg, Aparna A. Divaraniya, Aparna A. Divaraniya, Christine E. Becker, Christine E. Becker, Pankaj Agarwal, Joel T. Dudley, Joel T. Dudley

Research output: Contribution to journalArticlepeer-review

1 Scopus citations


Background: Mapping disease-associated genetic variants to complex disease pathophysiology is a major challenge in translating findings from genome-wide association studies into novel therapeutic opportunities. The difficulty lies in our limited understanding of how phenotypic traits arise from non-coding genetic variants in highly organized biological systems with heterogeneous gene expression across cells and tissues. Results: We present a novel strategy, called GWAS component analysis, for transferring disease associations from single-nucleotide polymorphisms to co-expression modules by stacking models trained using reference genome and tissue-specific gene expression data. Application of this method to genome-wide association studies of blood cell counts confirmed that it could detect gene sets enriched in expected cell types. In addition, coupling of our method with Bayesian networks enables GWAS components to be used to discover drug targets. Conclusions: We tested genome-wide associations of four disease phenotypes, including age-related macular degeneration, Crohn's disease, ulcerative colitis and rheumatoid arthritis, and demonstrated the proposed method could select more functional genes than S-PrediXcan, the previous single-step model for predicting gene-level associations from SNP-level associations.

Original languageEnglish
Article number6
JournalBioData Mining
Issue number1
StatePublished - 17 Jun 2020


  • Gene candidate discovery
  • Genome-wide association study
  • Network biology


Dive into the research topics of 'Identification of therapeutic targets from genetic association studies using hierarchical component analysis'. Together they form a unique fingerprint.

Cite this