I'm excited to share our new fine-mapping methods: SuSiE-inf and FINEMAP-inf. We improve upon the current state-of-the art methods by modeling infinitesimal effects while fine-mapping larger causal effects. 🧵 1/n

Our SLALOM paper is now out in @CellGenomics!🎉 Huge thanks again to Roy Elzur @weizhouw @dalygene and Hilary Finucane as well as all the analysts & participants in @GlobalBiobanks

Last but not least, huge thanks to co-authors: Roy Elzur, @masakanai @julirsch @oweissb @dalygene @bmneale and co-mentors Zhou Fan and Hilary Finucane!

Our software implementation is available at https://t.co/LcWnVTN75c. Runtime is comparable to existing methods. 5/n

Using the posterior effect sizes of the sparse component generated by our methods to perform polygenic risk score predictions (PRS) significantly improves prediction accuracy. 4/n

We performed simulations with model/prior misspecifications and demonstrated that non-sparse causal effects can be a contributor to miscalibration. Our new methods, where we model infinitesimal effects in addition to sparse causal effects can ameliorate this miscalibration. 3/n

State-of-the art methods may be miscalibrated when applied in real data. We found higher-than-expected Replication Failure Rates for SuSiE, FINEMAP and COJO-ABF when fine-mapping @uk_biobank. In contrast, RFRs are low in simulations without model/prior misspecifications. 2/n

I'm excited to share our new fine-mapping methods: SuSiE-inf and FINEMAP-inf. We improve upon the current state-of-the art methods by modeling infinitesimal effects while fine-mapping larger causal effects. 🧵 1/n

Excited to share our latest manuscript on complex trait fine-mapping across diverse populations! This is a collaborative effort across BioBank Japan @FinnGen_FI and @uk_biobank with mentorship from Hilary Finucane @dalygene @okada_yukinori. Here is 🧵 1/n https://t.co/lnRWRXjrTA

Types of human genetics papers in which I come for all of you and then for myself:

We (@skdeshpande91, @ta_broderick and myself) have a new pre-print out now! It’s called “Confidently comparing estimators with the c-value”. https://t.co/wEU4n7aYBD

I've been incredibly lucky to work on many large, collaborative projects in human genomics during my PhD w/@TalkowskiLab I wrote an informal blog post on three lessons I've learned, with some (surely naive) thoughts on future directions for our field: https://t.co/OXKIXUSXjk

This is the stuff great problems are made of: Simple enough that 10 year olds can understand it, complex enough to remain unsolved for over 50 years now

More than a strict P<5e-8 cut off, biological evidence to support the loci matters more. Here the authors show that ~92% of the 2010 BMI loci at P<5e-5 that overlapped with adipocyte promoters turned out to be genome-wide significant later in 2018. https://t.co/FOnY3d2vZo