Answers are grounded in this paper and cite their sources.
Observational associations can reflect confounding or reverse causation, which Mendelian randomisation (MR) avoids by using genetic variation to estimate causal effects. In Diversity Outbred (DO) mice such estimates can be tested experimentally, but applying human summary-data MR, with one scalar association per variant, ignores the differing effects of their eight founder haplotypes. We introduce Founder GLS-IVW, which uses each locus's founder-effect pattern for causal estimation and diagnosis. At Qsox1, it recovered the direction of effect seen in knockout mice, which one scalar coding failed to detect. Across 4,596 trait pairs, scalar and founder analyses disagreed for 29.7%; in seven of eight, only the founder analysis found an effect, using information scalar coding discards. Because exposure and outcome may be measured in different or partly overlapping animals, the method supports two-sample MR across separate experiments. Our R package qtl2mr, built on the widely used R/qtl2, helps prioritise mechanistic experiments.
No takes yet. Share an insight, caveat, or question.
Dai et al. (2026) studied this question.