Conditioning on Parental Mating Types in Mendelian Randomization to Reduce Bias and Strengthen Assumptions: A Promising Approach for Causal Inference


Observational studies often encounter confounding, which emerges when an unmeasured or uncontrolled factor is connected to both the ventured exposure and the pursued outcome. In order to tackle these confounders, researchers frequently encompass potential variables that could disrupt their findings as covariates in regression models or opt for stratified analyses. Nevertheless, these methods have certain limitations due to their dependence on measured confounders and assumptions about the relationship between confounders, exposure, and outcome.

Mendelian randomization (MR) has emerged as a valuable tool in epidemiological studies for inferring causal relationships. By utilizing genetic variants as instrumental variables, MR aims to overcome the limitations of observational studies and provide unbiased estimates of causal effects. However, MR relies on certain assumptions, and when these assumptions are violated, biased estimates can result. One critical assumption is that there should be no association between the genetic variant used as an instrument and potential confounders. In a new study published in the Journal Frontiers in Genetics, Dr. Keisuke Ejima, Dr. Nianjun Liu, Dr. Luis Miguel Mestre, Dr. Gustavo de Los Campos, and led by Professor David Allison from Indiana University School of Public Health-Bloomington proposed conditioning on parental mating types as a strategy to alleviate bias in MR caused by the correlation between the instrumental variable and confounding variables. Parental mating types are defined as combinations of genotypes of parents at a locus used as an instrumental variable. This approach eliminates the need for assumption in MR, thus reducing the burden of assumptions and potential bias.

The authors conducted simulations to illustrate the bias and inflation of type I error that can occur in MR when the instrumental variable and confounding variables are correlated due to assortative mating or population admixture. The simulations compared three methods: conventional MR, MR conditioning on parental genotypes, and MR conditioning on parental mating types. They showed that conventional MR led to type I error rate inflation and biased estimates in scenarios involving assortative mating or population admixture. Conditioning on parental genotypes partially reduced the bias but did not eliminate the inflation of type I error. However, conditioning on parental mating types in MR effectively eliminated type I error inflation and bias under these circumstances.

The new approach of conditioning on parental mating types in MR has significant implications for causal inference in epidemiological studies. By strengthening assumptions and reducing bias, this approach provides a valuable tool to enhance the validity of MR estimates. However, challenges related to data availability, such as the need for parental genetic data, should be acknowledged. Nevertheless, advancements in DNA sequencing technology and data imputation methods offer potential solutions to overcome these limitations.

By improving the validity of MR estimates, researchers can gain more accurate insights into causal relationships between exposures and outcomes. This has implications for understanding disease etiology, identifying potential therapeutic targets, and informing public health interventions. The authors’ proposed approach has the potential to enhance the reliability of causal inferences derived from genetic data, thereby advancing the field of epidemiology and contributing to evidence-based decision-making. Furthermore, the  new study contributes to methodological advancements in the field of genetics and epidemiology. By introducing and validating the concept of conditioning on parental mating types, the study expands the toolkit available to researchers for causal inference. This novel approach offers a potential alternative to existing methods and opens up avenues for future research and exploration. The study’s findings can inspire further investigations into the application and performance of conditioning on parental mating types in various study designs and scenarios.

Future research should explore the performance of conditioning on parental mating types in more complex scenarios, such as multi-locus models with epistatic interactions. Additionally, the approach should be evaluated in situations involving selection bias on the exposure variable. Further investigations are warranted to assess the superiority of conditioning on parental mating types over conditioning on allele dosages and to examine its applicability in various realistic circumstances.

In conclusion, the study by Professor David Allison and colleagues highlights the potential of conditioning on parental mating types as a strategy to reduce bias and strengthen assumptions in Mendelian randomization. By addressing confounding variables correlated with the instrumental variable, this approach offers a promising avenue for improving causal inference in epidemiological studies. While challenges remain, the proposed method has the potential to enhance the validity of MR estimates and contribute to a better understanding of the complex interplay between genetics, exposures, and outcomes. Continued research and practical implementations will further elucidate the benefits and limitations of this approach, ultimately advancing the field of genetics and epidemiology.

About the author

David B. Allison, Ph.D., is Dean, Distinguished Professor, and Provost Professor at the Indiana University School of Public Health-Bloomington. Prior, he was Distinguished Professor, Quetelet Endowed Professor, and Director of the NIH-funded Nutrition Obesity Research Center at the University of Alabama at Birmingham. Continuously funded by the NIH as a PI for more than 25 years, he has authored more than 600 scientific publications. His awards include the Presidential Award for Excellence in Science, Mathematics, and Engineering Mentoring; the Friends of Albert (Mickey) Stunkard Lifetime Achievement Award from The Obesity Society; the Don Owen Award from the San Antonio Chapter of the American Statistical Association; and the Harry V. Roberts Statistical Advocate of the Year Award from the American Statistical Association. Named a Distinguished Lecturer by Sigma Xi in 2022, Dr. Allison is a staunch advocate for rigor in research methods and the uncompromisingly truthful communication of research findings.


Ejima K, Liu N, Mestre LM, de Los Campos G, Allison DB. Conditioning on parental mating types can reduce necessary assumptions for Mendelian randomization. Front Genet. 2023 ;14:1014014. doi: 10.3389/fgene.2023.1014014.

Go To Front Genet.