Our paper led by Brianna Flynn on using AI as a phenotyping engine for diseases which are diagnosed largely through imaging is now out in npj Digital Medicine. We double case counts + obtain a quantitive biomarker associated with severity directly from imaging. nature.com/articles/s4174…

In this week's #SciencePodcast 🎙️ with SarCrespi, Sandeep Ravindran, and Eucharist Kun: 

▶️ Adding thousands of languages to the #AI lexicon
▶️ The genes behind our bones

🎧 Listen here: scim.ag/3r4

Genome-wide map reveals regions associated with skeletal changes that enabled humans to walk upright go.nature.com/3K9zb9v

First time on a podcast and it was a great experience, thanks for having me! SarCrespi

nature.com/articles/d4158…

What you can get when you combine deep learning #AI of DXA scans and genomics in >30,000 people!
Science Magazine UK Biobank Science Visuals
doi.org/10.1126/scienc…
science.org/toc/science/cu…

World-changing work from NaturalSciences @ UT! Our evolutionary scholars just landed on the cover of Science Magazine for research that could help doctors better predict patients’ risks! 🤘 #WhatStartsHere

Big congrats to Eucharist Kun Vagheesh Narasimhan and the whole team! Great collaboration using deep learning, GWAS, S-LDSC and enrichment analyses to ask about the genetic basis and evolution of the human skeletal form. Stay tuned for more from us on the evolution of complex traits.

Humans are the only bipedal great apes, due to the evolution of our unique skeletal form. What genomic regions underlie such change? Our cover article in @sciencemagazine led by Eucharist Kun and in collab. with @tarjs is now out! science.org/doi/10.1126/sc…
twitter.com/ScienceMagazin…

There have been several papers looking at selection using time series data that ancient DNA presents. Virtually all use single locus statistics. But how do multi-locus approaches fair? New pre-print examining this on 708 genomes from the past 7k years. biorxiv.org/content/10.110…

Excited to share this amazing work from our group utilizing deep learning methods to phenotype medical images which can complement incomplete electronic health record data and improve the power of gene association studies

A broad-level summary of amazing work from Riley Mangan in the Lowe lab, check out the original paper twitter thread here: twitter.com/riley_mangan/s…