Fig. 2: Flow diagram representing how the PWC-MR approach aims to disentangle causal effect between trait pairs from confounding or pleiotropy, as well as systematically search for confounders of the trait pair. | Nature Communications

Fig. 2: Flow diagram representing how the PWC-MR approach aims to disentangle causal effect between trait pairs from confounding or pleiotropy, as well as systematically search for confounders of the trait pair.

From: PheWAS-based clustering of Mendelian Randomisation instruments reveals distinct mechanism-specific causal effects between obesity and educational attainment

Fig. 2

Panel a represents the main steps of the PWC-MR method: (i) Instrument selection and PheWAS; (ii) Informative IV clustering using K-means; and (iii) Enrichment analysis and cluster-specific MR. Panel b represents a complimentary approach to PWC-MR where a systematic candidate confounder trait search is performed. These candidate confounder traits are defined as having an effect on both the exposure and the outcome. A stepwise multivariable MR (MVMR) of the candidate confounder traits is performed to select those with a strong effect on the outcome. These are then added with the primary exposure (BMI) to a standard MVMR and the multivariable causal effect on the outcome (EDU) is estimated. Acronyms: N: sample size, rG: genetic correlation, T: trait, p: t-test p-value; MR p: MR p-value.

Back to article page