Matthew Stephens
@mstephens999
Professor in Statistics and Human Genetics at University of Chicago
PLZ RT: I’m happy to present @nmancuso_ Lab’s new software SuShiE (Sum of Single Shared Effects) that performs variational SNP fine-mapping on multi-ancestry molecular QTL data. Stay tuned for the future manuscript! (1/8) github.com/mancusolab/sus…
Pleased to share our latest preprint. We extended the SuSiE model originally developed by the @mstephens999 Lab to a cross-population fine-mapping method. Led by the talented @KaiYuan1990 with @hailianghuang @RyanLongchamps @AFPopgen + many others. medrxiv.org/content/10.110…
Thanks so much @JovMaksimovic for sharing with us your love for #workflowr and huge thanks to the #workfowr developer @jdblischak for this incredible & easy to use tool that is helping the community towards better analyses & science! Find talk slides github.com/R-LadiesMelbou…
Glad to see our paper, with @xinhe2 @mstephens999 @wescrouse Sheng and Kevin, is out today at @NatureGenet! nature.com/articles/s4158…. We developed a new statistical method, causal-TWAS or cTWAS, to reliably identify causal genes in genome-wide association studies (GWAS). 1/n
The Department of Human Genetics at University of Chicago is searching for a faculty candidate in Computational genetic and genomics. Please share this job ad with your colleagues: apply.interfolio.com/127605
My PhD work "On the Genes, Genealogies, and Geographies of Quebec" is now published in SCIENCE!!🧑🔬 🌎We explore the emergence of genetic variation in Quebec's settler population with a detailed model using a genealogy of 5M records. 🔗science.org/doi/10.1126/sc… 🧵for details : 1/7
recent work from my lab on applying and interpreting topic models for single cell data: biorxiv.org/content/10.110…
I missed tweeting this when posted in Dec. Our attempt to ameliorate some problems with confounding in TWAS. Key idea is to compete (imputed) gene expression and SNPs against one another as explanations for an asssocation. biorxiv.org/content/10.110…
Glad to see this implementation out! We extended the SuSiE model to latent factors with SuSiE-PCA! Using JAX, SuSiE-PCA can run on CPU, GPU, or TPU. Preprint to come shortly, but in meantime, please try it out for your high-dim data settings and share your thoughts!
I am excited to announce the sum of single effects principle component analysis (SuSiE PCA), a scalable Bayesian variable selection technique for PCA and more than 300 times faster compared to existing methods. It is implemented in python. Try it out here: github.com/mancusolab/sus…
"Poisson-mash", or how to detect gene expression differences in multi-condition experiments (from the group of @mstephens999, together with Nicolas Chevrier and colleagues). arxiv.org/pdf/2210.00697…
So, this came out ..
Meta-analysis of (single-cell method) benchmarks reveals the need for extensibility and interoperability biorxiv.org/cgi/content/sh… #biorxiv_bioinfo
We have a new open-rank faculty search in the department of Medicine at the University of Chicago! We are looking for PhD scientists in a wide range of research areas. Please share widely. academicjobs.uchicago.edu/positions/1029…
The University of Chicago, Department of Human Genetics, is searching for a new computational faculty. apply.interfolio.com/112673 We are a small, friendly department doing top-quality science. Come join us! (Please RT)
NIH peer review system is far from perfect, but it's the best I've seen. I just "retired" from the GCAT study section after 4 years' of service, which prompted me to share a few thoughts re computational proposals. These are the points that I also remind myself. Other thoughts?🧵
Very happy to share that my lab was awarded an @NIH @NIGMS #MIRA #R35 grant to expand our research on human complex trait genetics! As a result, we are hiring at all levels (e.g., graduate students, postdocs, research scientists). Get in touch if you are interested!
Gil Mcvean on the interpretation of PCA applied to population genetics data journals.plos.org/plosgenetics/a…
FWIW this paper wildly overstates the problem: PCA is an exploratory data analysis tool with well-known theoretical properties journals.plos.org/plosgenetics/a…
📢Big news: my lab is moving to @UChicago! I'm thrilled to join the superb genomics and microbiome research community at UChicago, and beyond excited about new research directions and collaborations! And we're recruiting at all levels! 👇 blekhmanlab.org
Ask yourself which 2-D representation better represents the structure of the hypercube. An orthogonal projection or t-SNE?
using t-SNE to embed the 1024 vertices of a 10-dimensional hypercube in two dimensions
Our bioinformatics team at Center for Research Informatics at U Chicago is ever-growing. cri.uchicago.edu/bioinformatics/ Now multiple scientist positions are available. Please check us out if you are on the job market. See below post for details.