A great thread from @uthsavc describing some of the math behind GASTON, our new method for modeling spatial variation in gene expression, now out in @naturemethods!.

Congratulations to all the winners and finalists of the 2024 @Regeneron Science Talent Search! It was inspiring to meet the next generation of leaders in science and technology!.

Thank you @rohitsingh8080 for hosting me at @DukeCellBiology! A great visit with a packed auditorium and many stimulating conversations! This photo was me being delightfully surprised that I was speaking in a lecture series next to these truly distinguished cell biologists.

RT @uthsavc: Also, a quick observation: our 1-D coordinate is a spatial *potential function* - like the height in the famous Waddington's l….

GASTON software available at Tutorials are here with more coming soon (7/7).

In the olfactory bulb, the GASTON isodepth indicates the “differentiation potential” of neurons – with isodepth gradients showing trajectories of neural maturation. Moreover, GASTON reveals the distinct layers and cell type gradients within the layers. (6/7)

GASTON also identifies gradients in the tumor microenvironment; e.g. gradients of metabolic activity in the tumor interior, and gradients of EMT-related genes in the adjacent stroma – indicating possible metastatic migrations. (5/7)

By aggregating sparse gene expression measurements across contours of equal isodepth, GASTON identifies expression gradients in complex tissue geometries, e.g. the curved layers of the cerebellum. (4/7)

To derive such a topographic map, we introduce the *isodepth*: a 1-D coordinate in a tissue slice analogous to elevation in a topographic map. GASTON learns the isodepth using an interpretable deep learning model. (3/7).

From a topographic map of a landscape, a hiker can identify mountains and valleys or determine the steepness of a path. Similarly, from a topographic map of a tissue slice, one can find spatial domains in a tissue and examine gradients of gene expression within domains (2/7)

Our new method GASTON learns a “topographic map” of a tissue slice, enabling simultaneous identification of spatial domains and gene expression gradients in spatial transcriptomics Led by @UthsavC w/ @BrianJohnArnold @hrksrkr @Cong992 Sereno + Kohei (1/7)

RT @ZaccaSimo: Consider this great opportunity to submit algorithmic or analysis cancer comp bio contributions to present to the phenomenal….

Belayer first appeared @Recomb2022 and in preprint form (. Excited to be one of several exciting papers in the “Focus on #recomb2022” special section in @CellSystemsCP 6/6.

Belayer software and tutorial available on Github: 5/6.

By using a global model of tissue organization, Belayer accurately identifies tissue layers and spatially varying genes in the brain and skin. 4/6

What about tissues with curved layers? Belayer “straightens” the curves using the theory of conformal maps from complex analysis, or in the simpler case by modeling heat diffusion through the layers with the isotherms defining the relative depth within a layer. 3/6

Belayer models gene expression as a piecewise linear function of layer depth – accounting for both continuous gradients of expression within a layer and discontinuities in expression at layer boundaries – and combining information across spots at same depth 2/6

Spatial transcriptomics meets complex analysis – our new method Belayer models spatial gene expression in layered tissues, now published in @CellSystemsCP. Work led by @Cong992 and @uthsavc with Shirley Zhang. 1/6

RT @PrincetonCS: Five computer science graduate students have been named 2023 @SiebelScholars. The award recognizes academic achievement an….

RT @brangwynnelab: Princeton Bioengineering is hiring! (please ReTweet!). The search is open rank, but we're particularly interested in tal….