Mike Gloudemans

It's finally here! Thanks to all my excellent co-authors for making it happen :)

@matias_kaplan variety.com/2021/film/news…

In my "free" time aside from the whole PhD thing, I've been making random (bi-)weekly book review videos for the past year. Here's a selection of my best reads of 2021! youtu.be/mVIC_u_2DtI

Some nice work using multiple related GWAS to improve fine-mapping resolution of causal variants

@JennAsimit @NatureComms Very nice work! How viable would it be to use some kind of computational heuristic to scale this up to 100s of traits or more? It would be awesome at some point to be able to jointly fine-map every GWAS that's ever been done, but I realize there are some obvious obstacles.

Excited to share our work @NatureComms – flashfm uses summary-level data to jointly fine-map signals of up to six traits. #MultiTrait #GeneticStudies #ExeterDiabetes nature.com/articles/s4146…

As demonstrated by this excellent meme, structure determination methods are very different from each other. How does this affect machine learning models trained on protein structures? @awfderry, myself, & @Rbaltman explore this in our new preprint: biorxiv.org/content/10.110… (1/5)

@metapredict That's fair, and we're looking here for tissue-specific QTLs / colocalizations, which adds even more complexity. Especially for lowly expressed genes, we'd see a richer picture at higher sample sizes, so tissue-specific patterns here are more of a clue than definitive evidence.

@mikegloud If you are using RNAseq to inform on tissue specific expression you’ll be grossly under-reporting across-tissue expression. In muscle alone, 80M paired-end RNAseq reads misses >30% of expressed genes.

1/ Excited to share some new work on an integrative colocalization analysis to identify and prioritize new cardiometabolic disease genes! medrxiv.org/content/10.110…

@Eric_Fauman We've also done some additional work evaluating gene prioritization methods on simulated data, which I hope will be out very soon. Of course, simulated data are only as good as the underlying assumptions, so there we've matched attributes of real GWAS as closely as possible.

@Eric_Fauman The downside of using known causal associations as a "gold standard", though, is that known associations with largest effect sizes may be relatively low-hanging fruit, and not representative of the genetic architecture for causal genes in general.

13/ And to learn more about the original perturbation experiments, check out @brunilda_balliu's paper, hot off the press in AJHG: twitter.com/brunilda_balli…

12/ Drop by virtual poster 3660 at ASHG next month to learn more about this analysis, or to brainstorm how to apply it to your own diseases of interest.