Thread by @ewanbirney

Great meeting with Dr Morishita and Professor Takeda at the University of Tokyo (other campus though - testing my Tokyo metro skills) on Medaka fish genome and annotation. They have some great HiC data during Medaka development in the works - look out for the publication

Worth stressing again that there are classes of questions we can't answer in human - we can make statistical models, and fit those models, but we can (rarely) challenge or explore the underlying model in human - eg, how many GxE loci are there and do they overlap with GxG?

There is a surprisingly large amount of "this is the simplest model we can think of to describe the data so let's use it" in human genetics, and some real handwaving moments - how much "variance" is due to structured environment and how much due to individual chance events?

We can't get at some of these things easily (some studies try, but it's really hard) and most human genetics sweep all of these terms up together, along with measurement error (confusingly often called "environment" - not a good moniker) and this gets mainly ignored.

This is both because one can't control most variables in human (in particular the environment) and also because the numbers really work against you for some analyses - for example, the product of allele frequencies, in particular >2 make most GxG explorations in human a nightmare

We *need* non human models - both to do the "final mile" set of experiments to really nail down causality (eg, introducing specific changes via CRISPR) but also just simply to understand what is going on - both biologically, physiologically and statistically.

Many models are applicable - Marmoset, Mouse, Rat, Chicken, Xenopus, Zebrafish and ... my favoured Medaka fish ("Japanese rice paddy fish"). Why Medaka fish? The key feature is you can inbreed Medaka fish reliably from the wild (~50% of inbreeding attempts succeed)

We (Felix Loosli, Jochen Wittbrodt, Kiyoshi Naruse and myself) have made the first ever inbred panel from the wild in a vertebrate (old hat for arabidopsis and Drosophila - new for vertebrates!). This panel has 110 lines and alleles and allele frequencies that come from the wild

If you are human geneticist, it is as if Framingham, or part of UK BioBank, or part of Finland, was in fact a set of every lasting identical twins which you can repeat studies on again and again. Doing 10 embryos from a isogenic line is pretty routine in Medaka.

If you come from animal studies, this is like a recombinant inbred line, except the recombination is population scale recombination. This has positives (crazy good mapping properties!) and negatives (too much haplotype diversity, low allele frequencies).

University Tokyo's old campus is lovely - mature Ginko trees with the leaves being swept up; old buildings in the gothic style. Now waiting in a cafe in Tokyo station for the right train to get me to Okazaki where NIBB and @Naruse_kiyoshi is.