Aims/hypothesis: There is an extensive body of literature suggesting the involvement of multiple loci in regulating the action of metformin; most findings lack replication, without which distinguishing true-positive from false-positive findings is difficult. To address this, we undertook evidence-based, multiple data integration to determine the validity of published evidence.
Methods: We (1) built a database of published data on gene-metformin interactions using an automated text-mining approach (n = 5963 publications), (2) generated evidence scores for each reported locus, (3) from which a rank-ordered gene set was generated, and (4) determined the extent to which this gene set was enriched for glycaemic response through replication analyses in a well-powered independent genome-wide association study (GWAS) dataset from the Genetics of Diabetes and Audit Research Tayside Study (GoDARTS).
Results: From the literature search, seven genes were identified that are related to the clinical outcomes of metformin. Fifteen genes were linked with either metformin pharmacokinetics or pharmacodynamics, and the expression profiles of a further 51 genes were found to be responsive to metformin. Gene-set enrichment analysis consisting of the three sets and two more composite sets derived from the above three showed no significant enrichment in four of the gene sets. However, we detected significant enrichment of genes in the least prioritised category (a gene set in which their expression is affected by metformin) with glycaemic response to metformin (p = 0.03). This gene set includes novel candidate genes such as SLC2A4 (p = 3.24 × 10-04) and G6PC (p = 4.77 × 10-04).
Conclusions/interpretation: We have described a semi-automated text-mining and evidence-scoring algorithm that facilitates the organisation and extraction of useful information about gene-drug interactions. We further validated the output of this algorithm in a drug-response GWAS dataset, providing novel candidate loci for gene-metformin interactions.
Keywords: G6PC; Gene-set enrichment; Metformin; SLC2A4; Text-mining; Type 2 diabetes.