Statistical Causal Inference and its Applications to Genetics

We announce an exciting interdisciplinary workshop as part of a month long research program on causal inference in genetics. Many of the recent breakthroughs in high-dimensional statistics have been driven by problems in genetics, especially by the difficulties associated with inference where the number of covariates (such as SNPs) massively exceeds the number of individual samples.

Significant progress towards attacking such problems has been achieved in recent years through methods based on regularization, sparsity, and control of false discovery rates. These methods are mainly intended for observational data such as case-control studies or genome wide association studies. However, modern sequencing methods and gene knockout techniques are leading to radically different datasets, and require a new generation of statistical methodology to match them. Specifically, we need new statistical methods that:

• can take huge quantities of data from multiple experimental settings, including time course data, and provide a coherent picture of the mechanistic interactions at work,

• generate and rank causal models based on observational and limited experimental data so as to guide future hypotheses and experiments,

• advance efficient experimental design,

• incorporate prior information in a computationally tractable way,

• provide causal methods for time series data,

• increase the power of Mendelian randomization, i.e. to use genetic information as instrumental variables,

• efficiently incorporation of prior structure or information from multiple experimental settings.

NOTICE: We are open to submissions for research presentations and posters, and particularly encourage the participation of junior researchers. To register your interest in attending and/or presenting work, please contact Robin Evans.