PLOS Genetics paper introduces MR2G framework for inferring causal networks from GWAS summary data

Researchers Zhaotong Lin, Wei Pan, and Haoran Xue have published a paper in *PLOS Genetics* presenting MR2G, a novel computational framework for inferring causal networks among multiple traits using genome-wide association study (GWAS) summary statistics. The study, published on 26 May 2026, addresses a recognised limitation of existing Mendelian randomisation (MR) approaches: standard two-trait MR methods require the analyst to specify a causal direction in advance, which is not always possible when studying complex biological systems.

MR2G uses genetic variants as instrumental variables to estimate causal effects across a set of traits simultaneously, and is specifically designed to handle scenarios where the directions of causal relationships are unknown or where cycles and feedback loops exist in the underlying biology. The authors report that such cyclic structures are common in gene regulatory and metabolic networks, and that ignoring them can bias or distort causal inference.

The paper includes simulation studies and applications to real GWAS datasets, and the framework is positioned as an advance over existing multi-trait MR methods in its ability to recover network topology under these conditions. MR2G adds to a set of recent methodological developments in statistical genetics aimed at moving from pairwise causal estimates to richer network-level causal inference.

This is a methods-focused paper in statistical genetics; the immediate application is to research settings where multi-trait GWAS data are available and causal architecture is of interest.