Sarah Urbut

@drjohnm Many thanks @drjohnm. Fully agree with you. For that reason, I’m excited by this work led by @tigerstatdoc & with @SashaGusevPosts G. Parmigiani modeling numerous health trajectories simultaneously in interpretable fashion to better predict & prioritize medrxiv.org/content/10.110…

Our most comprehensive global #statistics #epidemiology update with 700 pages on #CV, #metabolic, #kidney, #brain dz @Circ_AHA Most disorders have shared #riskfactors that are preventable and treatable! Want to know the stats on each?? ahajournals.org/doi/epdf/10.11…

@aklfahed @MGHHeartHealth Awesome work with @g_parmigiani, @SashaGusevPosts and @pnatarajanmd, all star team!

When Bayesian statistics meets cardiology @tigerstatdoc @MGHHeartHealth #AHA25

@aklfahed @MGHHeartHealth Thank you @aklfahed !

The Khetarpal lab at UVA is seeking a research associate (postdoc)! Please see the posting and reach out with any questions! jobs.virginia.edu/us/en/job/R007…

@pnatarajanmd @NEJMEvidence @MGBResearchNews @harvardmed @broadinstitute @MGHMedicine @MassGeneralMDs @CGM_MGH @mghcvrc @MGHHeartHealth thank you @pnatarajanmd! such an important question and an honor to be featured in @NEJMEvidence - excited about the use of Bayesian hierarchical models to disentangle sources of variation within and between ancestries, cohorts, and sampling...

Delighted to share our study led by @tigerstatdoc demonstrating the genetic generalizability of LDL-cholesterol's association with coronary artery disease across 1M individuals globally evidence.nejm.org/doi/full/10.10… @NEJMEvidence While clinical trials of LDL-C-lowering medicines have shown improvement in CAD risk, they generally don't reflect the breadth of patients we evaluate for such medicines. Subgroup analyses in trials to assess for differences are generally very underpowered as they are not the main intention of trials. The world's human genetic data is (1) larger and covers many more communities and geographies than clinical trials, and (2) can enable causal inference methodologies across these groups. We assembled data across the US, Europe, Middle East, and South Asia to examine this relationship across individuals of African, admixed American, East Asian, European, Middle Eastern, and South Asian ancestry. Adjusting for variability in allele frequencies and performance, harmonized the same LDL-C effect, we observed (1) a significant positive association with CAD and (2) heterogeneity in this association. While this is not a substitute for clinical trials reflecting the breadth of patients we see, the results (1) provide reassurance regarding current clinical practice extrapolating existing trials of LDL-C-lowering therapies, (2) the relationship between LDL-C and CAD appears generally robust across communities & solidifies this as a core pathway for CAD risk reduction, and (3) indicates that investment in adequately powered LDL-C-lowering therapy trials in other communities is likely to yield favorable returns. Many thanks to the team! @skoyamamd @HornsbyWhitney @gina_peloso @aklfahed @wallacemwang & others

Enjoyed reading this paper on how to jointly model genetics and longitudinal health records medrxiv.org/content/10.110…

@joe_pickrell Thank you so much @joe_pickrell !!

@BellomoTiffany Thank you @BellomoTiffany :)

@tigerstatdoc An absolute math wizard 🧙 congratulations!!!

Announcing our much-awaited updated aladynoulli, a fully interpretable Bayesian model that considers predictive and explanatory disease trajectories across the life course for individuals, anchored on underlying PGS. medrxiv.org/content/10.110…

Amazing work from Sarah on Bayesian inference of longitudinal EHR trajectories and latent disease signatures. Check out the app!

@SashaGusevPosts Thank you @SashaGusevPosts !

@teapoow thanks - as described in the prePrint, the GitHub repo is private to collaborators only now, but please see an example of visualizing model at work at aladynoulli.hms.harvard.edu

@tigerstatdoc Amazing work, congratulations! I can't wait to dive deep into it & use it on my own work! One small thing is that I wasn't able to find the github repo & the one mentioned in the paper (github.com/surbut/aladyno…) doesn't seem to be working. Just a heads up.

@pnatarajanmd @SashaGusevPosts @broadinstitute @MGBResearchNews @harvardmed @MGHHeartHealth @CGM_MGH @mghcvrc an amazing team, thank you @pnatarajanmd !

Excellent work by @tigerstatdoc constructing a generalized lifelong interpretable disease risk prediction model using genetics and clinical factors

@BennyStrobes thank you!

@tigerstatdoc Congrats Sarah! Such cool work.

Follow the math, follow the pt with associated app at aladynoulli.hms.harvard.edu. Our model is truly generative, explicitly modeling the risk captured by EHR histories through Bayesian the *magic* 🪄 of mixture modeling. @SashaGusevPosts @pnatarajanmd @g_parmigiani .

@RonBlankstein @pnatarajanmd @MassGenBrigham @MGHHeartHealth @mghcvrc @CGM_MGH @harvardmed @broadinstitute Thank you @RonBlankstein !

@pnatarajanmd @tigerstatdoc @MassGenBrigham @MGHHeartHealth @mghcvrc @CGM_MGH @harvardmed @broadinstitute Congrats Sarah!

Congrats to @tigerstatdoc for receiving a @MassGenBrigham Heart & Vascular Institute Alan Lerner Junior Faculty Award!