Who to read in snp/gwas/personalised medicine?


#1

i’m starting to read the results from these analyses in eg Science magazine. Who are the good statisticians I should read to know the standard methodology or interesting new methodology. Or are they mostly using old methods (bonferroni it seems eg 0.05/20) and there isn’t much development of new methods? I’m seeing ‘manhattan plots’ for the first time and I want to get up to speed. cheers


#2

I hope you get a lot of responses. This is a wide open area but tremendous misunderstandings. I believe that the typically used plots have harmed science by stressing p-values and not magnitudes of effects. I would change the y-axis to use odds ratios and use another coding for p-value in Manhattan plots.


#3

interesting re tweaking manhattan plots … Yes, the first thing I noticed was the use of AUROC which I have seen criticised on your blog (eg A targeted real-time early warning score (TREWScore) for septic shock). I will be fortunate enough to have access to good data and want to use it wisely and not blindly follow the standard approach …


#4

Dr. Stephen Senn, Dr. Michael Joyner and Dr. Cecile Janssens have written great papers on this topic. Eager to read more from others too though so I’d love to hear others input!


#5

Yeah this is a tough subject area. I once asked an adviser Prof vehtari about genomics, and he seemed to roll his eyes, when I mentioned PCA wrt genomics data. He references a paper that uses PCA on genomics data in Europe and each PC is a country. Easy to find with a google search.

But yeah, I worked, as an undergrad “bioinformatics” RA. We had raw next gen sequencing data, in the presence of expression of tfRNA that was known to be expressed in the presence of certain sickness. We “validated” our expression results again “theoretical” sequencing results using bowtie2 and a variety of sequencing and data tools. We used a two sample t test for comparison which, retrospectively, was ridden with false positives.

I think a more important question is what do you want to answer? I think this is more important, then you can subset the literature into what you need to know. It’s a huge field (there’s billions of genes, and many different inference to be made).

Dr. Bhramar Mukherjee is a reliable source for gene environment interaction.

But research quality can vary per subject, you know?