Deep Mendelian Randomization: explaining causality between different hereditary traits at genome-wide scale

2024, diiP, Masters Internship, Mathematics/Statistics

2024

Master’s Projects

@Mathematics/Statistics

#Mendelian Randomization

#Deep Learning

#Double Machine Learning

#Genomics

#Pleiotropy

 

Project Summary

Mendelian Randomization is a method that infers the causality between risk factors and diseases using genetic variants as instrumental variables. However this method suffers from major biases such as pleiotropy where a single variant influences multiple traits. To address these limitations we propose a novel approach called Triple Machine Learning-MR which leverages the inherent capabilities of AI by utilizing extensive genome-wide and multi-omic data. This strategy aims to 1) predict causal effects 2) provide a robust estimator and 3) refine models by selecting the most relevant genetic variants. Ultimately this comprehensive approach will offer greater precision in understanding the causality between traits and will be used to uncover the complex relationship between the immune system and cancer.

 

Marie Verbanck
marie.verbanck@u-paris.fr

Professor (Junior Professor Chair)
– Inserm U900, Institut Curie, PSL Research University.
– BioSTM, University Paris Cité.

Projects in the same discipline