Paul O'Reilly

I am a Reader in Statistical Genetics at the MRC SGDP Centre, IoPPN, King's College London. In the hope of gaining insights into how the human genome evolves and gives rise to disease, I develop statistical genetics methods/software in population genetics (eg. the Ped/Pop method, invertFREGENE) and genetic epidemiology (eg. MultiPhen, PRSice), while my applied work is in psychiatric genetics.

I did my PhD at Imperial College London (2004-08) on detecting selection in the genome, supervised by David Balding and co-supervised by Ewan Birney. After a post-doc working on numerous GWAS projects (w Marjo-Riitta Jarvelin) and methods development, particularly multivariate GWAS and inversion polymorphisms (w Lachlan Coin), I began a lectureship at Imperial in 2011, before moving to King's in 2013. 

Contact: paul.oreilly@kcl.ac.uk
twitter.com/paul_f_oreilly

Methods & Software

• PRSice ('precise') software - for performing Polygenic Risk Score (PRS) analysis. Command-line program that performs the multi-stage PRS process: calculation, application, evaluation and plotting of results. A feature of PRSice is that it can calculate PRS at 'high-resolution' to find the best-fit PRS - see the PRSice webpage for details or our Bioinformatics paper. PRSice code written by Jack Euesden and PRSice-2 by Sam Choi.
• MultiPhen method - for modelling and testing multiple phenotypes jointly in GWAS. This method is fast and can be applied to imputed data, and is available as a CRAN package (see here) that can also perform standard univariate GWAS. Paper describing the method, with results on the NFBC data, here.

• Ped/Pop method - for detecting recent positive selection. Compares population-based recombination rate estimates (underestimated due to recent common ancestor at selected locus) with pedigree-based recombination rate estimates (unaffected by selection). Paper describing the method here.

• invertFREGENE - software for simulating inversion polymorphisms in population genetic data. Paper describing the software and showing the patterns of variation at simulated and real inversions is here. Documentation and code:
  • invertFREGENE.pdf
  • invertFREGENE.tar.gz

• BAYESFST - software/method for detecting selection. Method described in Beaumont and Balding (2004) takes a Bayesian approach to identifying loci with divergent allele frequencies between sub-populations (based on Fst). Contributed to the software code by extending to deal with genome-wide linked markers, primarily by incorporation of a prior term between hyper-parameters to model the correlation between adjacent markers. The program, implemented via MCMC, and coded in C, can be downloaded here.

• BranchSort - algorithm that produces the gene tree at a locus without recombination from a sample of genotype data (thus phasing the data into haplotypes). Planning to publish at some point - but contact me if this would be useful. The future plan is to build a gene tree incorporating recombination - currently works for a small number of recombination events.

Selected Publications

Asherson, P., O'Reilly, PF. 2015. Genetic Effects, Categorical Disorders and Quantitative Traits. J Am Acad Child Adolesc Psychiatry 54:702-3.

Selected Talks

Teaching

Other roles