The best part, in my opinion, about being a researcher is being able to share your work with those around you! Sometimes that is in the form of a paper, and other times that is through presentations. I typically find the latter more enjoyable to put together, as you really have to consider how the audience will interpret what you put together in a short amount of time. Below are some of the talks I have given so far and I hope you enjoy them.

European Meeting of Statisticians, Warsaw 2023

• Title: Mendelian randomisation: why do we need to talk about pleiotropy?
• Date: Thursday 6th July 2023, 17:20-17:40
• Location: The University of Warsaw
• Room: 1.40

The days of classical association are coming to the near end in the world of genomics and biology, with many turning to a Causal Inference framework in an attempt to pursue true causal relationships. One such example is the well utilised framework of Mendelian randomisation [2]. Through the use of very large data sets, referred to casually as `omics’ data, we can look to form a causal directed acyclic graph (DAG) which determines the true causal relationship between an exposure (typically a transcriptomic or proteomic fea- ture) and outcome (such as some form of cancer) by introducing the genomic data as an instrumental variable.

Sadly however as time has gone on, it has been discovered that the conditions of the Mendelian framework are often broken by complex biological situations, such as the situation of pleiotropy. As such we now require some extended form of the Mendelian randomisation framework to ensure we are discovering the true causal relationship and not simply introducing the genomic data as an additional confounder of the exposure and outcome. For this, the work of Sun et al [1] proves to be a good, yet mathematically complex, solution.

This talk will briefly cover the concept of the Mendelian randomisation framework and the way in which the situation of pleiotropy breaches the exclusion criteria before discussing a solution to the problem we face. This will be done through the introduction of a newly introduced estimator which is asymptotically efficient, proven through simulations. Following this, the talk will mainly focus on exploring a couple of new and novel appli- cations and results will be presented through the use of large biobank data from the UK. These results will show that we must address the issue of pleiotropy in a real world setting. The results will be compared to traditional Mendelian randomisation methods so the need for a pleiotropic robust estimator can be transparent.

[1] Sun B., Liu Z., Tchetgen Tchetgen E., Semiparametric Efficient G-estimation with In- valid Instrumental Variables, Online, 2021 [https://arxiv.org/abs/2110.10615, accessed: 13-03-23].

[2] Sanderson, E et. al, Mendelian randomization, Nature Reviews Methods Primers, 2022, Volume 2.

Download talk slides here Recommended citation: Oldnall, Christopher A. (2023). “Mendelian Randomisation: Why Do We Need to Talk About Pleiotropy, Talk.” Unpublished.