Back

MRSL: A phenome-wide causal discovery algorithm based on GWAS summary data

Hou, L.; Geng, Z.; Shi, X.; Wang, C.; Li, H.; Xue, F.

2022-06-30 genetic and genomic medicine
10.1101/2022.06.29.22277051 medRxiv
Show abstract

Causal discovery is a powerful tool to disclose underlying structures by analyzing purely observational data. Genetic variants can provide useful complementary information for structure learning. Here, we propose a novel algorithm MRSL (Mendelian Randomization (MR)-based Structure Learning algorithm), which combines the graph theory with univariable and multivariable MR to learn the true structure using only GWAS summary statistics. Specifically, MRSL also utilizes topological sorting to improve the precision of structure learning and provides three adjusting categories for multivariable MR. Results of simulation reveal that MRSL has up to two-fold higher F1 score than other eight competitive methods. Additionally, the computing time of MRSL is 100 times faster than other methods. Furthermore, we apply MRSL to 26 biomarkers and 44 ICD10-defined diseases from UK Biobank. The results cover most of expected causal links which have biological interpretations and several new links supported by clinical case reports or previous observational literatures.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.