New life update! 🎆. 🎓 This Fall, I will be joining the Department of Computer Science at Johns Hopkins University (@JHUCompSci) as an Assistant Professor, with an affiliation at the new Data Science and AI Institute (@HopkinsDSAI).

Congrats to @gillianychu et al!!.

I am so deeply grateful for the support of many mentors and role models throughout the application process — there are too many to tag here, but thank you all 🙏 . (Also grateful for the @HDSIUCSD Rising Stars in Data Science workshop which pushed me to apply this cycle!).

The Chitra Lab will build the next generation of machine learning and AI algorithms for addressing fundamental problems in biology. I am actively recruiting students — please visit our website for more details!

RT @joshuasweitz: The White House Vision for Dismantling Science in One Simple Plot .

RT @Schmidt_Center: 🎉Congrats!.🥇@uthsavc: Mapping the topography of spatial gene expression with interpretable deep learning.🥈Anurendra Kum….

RT @TravisEGibson: Congrats Youn! and team, . @lab_berger @HultgrenLab @broadinstitute @mit @GeorgKGerber1 @BrighamWomens @MGBResearchNews….

RT @sebacultrera: Excited to share our latest preprint, introducing the hierarchical cross-entropy (HCE) loss — a simple change that consis….

RT @rohitsingh8080: I met @uthsavc at RECOMB '22. We got lunch and talked science. His GASTON work (RECOMB '24), with its isodepth, was th….

RT @rohitsingh8080: At RECOMB-Seq, I will be presenting GLACIER, with @uthsavc and my trainees Prannav and @hliang74 , where we introduce S….

RT @fabian_theis: 2/ Here, wanted to highlight interesting approach from Bonnie Berger’s lab @MIT, velorama @CellSystemsCP. It goes the oth….

RT @Schmidt_Center: Congratulations to @Schmidt_Center postdoc @uthsavc and team on their preprint, which introduces GASTON-Mix, an unsuper….

RT @henrismitch: 🚀 introducing fastppm: turbocharging tumor phylogeny reconstruction from bulk DNA-seq! we develop a new technique, tree-st….

@benjraphael @henrismitch informed me that the NCI workshop where we developed these ideas is now canceled thanks to the @realDonaldTrump @elonmusk NIH freeze 🙁.

Finally, if you've gotten this far: this work builds on our earlier work with @Cong992 using *complex analysis* (conformal mapping) to model spatial gradients. Check it out!.

Check out the paper for more details and neat biological applications! For example, modeling spatial gradients → much more accurate SVG identification. Thanks to great collaborators @BrianJohnArnold @koheisanno @hrksrkr @Cong992 Sereno @benjraphael 🙂

GASTON algorithm is quite natural: parametrize functions h,d with neural nets and learn from data. Fun back-story: @benjraphael and I derived most of this model at the bar near an NCI workshop 🥂😅

Model implicitly accounts for sparsity by “pooling” measurements across locations (x,y) with equal isodepth d(x,y). These locations look like contours of equal height on an elevation map, hence the “topographic map” analogy.

We prove that f(x,y) = h(d(x,y)), i.e. gene expression f(x,y) is function of a *single* spatial coordinate d(x,y) rather than 2 spatial coordinates x,y. Spatial dimensionality reduction! 🚀.We call d(x,y) the "isodepth"; it characterizes spatial gradients ▽f_g.

We make two assumptions:.(1) genes have *shared* gradient directions, i.e. each gradient ▽f_g(x,y) is proportional to shared vector field v(x,y).(Equivalent to Jacobian of f being rank-1 everywhere). (2) vector field v has no “curl”, so v=▽d is gradient of "spatial potential" d