(10/N) If you are interested check out our preprint and code! Thanks for reading!
Arxiv: arxiv.org/abs/2502.07746
Code: github.com/Krishnaswamy...

(9/N) Grateful to lead authors Siddharth Viswanath and Hiren Madhu, and coauthors Dhananjay Bhaskar, Jake Kovalic, Dave Johnson, Chris Tape, Ian Adelstein, Rex Ying and Michael Perlmutter!

(8/N) ✨ In short, HiPoNet is a neural network that takes in a whole point cloud, and utilizes methods from geometry and topology to derive features of the point cloud.
→ Multiple learned views
→ Simplicial wavelets to extract multi-scale structure

(7/N) What’s more, HiPoNet’s learned feature weights are interpretable. For example, immune-related markers like CD11b, CD118, and FOXP3 consistently emerge as important across views, aligning with known tumor-immune biology.

(6/N) We find that HiPONet consistently outperforms other methods on these datasets.

(5/N) We then evaluate HiPOnet on patient level classification tasks using single cell and spatial proteomics data:

(4/N) We first check that HiPoNet preserves topological features of datasets by predicting persistence from the datasets:

(3/N) HiPoNet models each cloud as a simplicial complex, and uses simplicial wavelet transforms to capture point-cloud level embeddings. Not only that---it uses multiview learning, i.e. different projections of the features to capture different underlying factors in the data.

(2/N) Modern technologies such as mass cytometry or scRNA-seq now allow large cohorts of patients to be measured. Each patient is actually a large single cell dataset creating a high-dimensional point cloud. But these point clouds are far more complex than 3D shapes that ML methods have handled..

Great ideas to introduce to the community!! Enjoyed the talk!

Frohe Weinachten!

(10/n) For code see github.com/Krishnaswamy.... Thanks for reading!

(9/n) GSPA-Pt can be used to map patient sample manifolds. We mapped 48 melanoma patient scRNA-seq samples and classified response from the patient embedding using logistic regression. The GSPA-Pt gene embeddings achieved the highest classification performance.

(8/n) GSPA-LR concatenates ligand (L) and receptor (R) for a pair representation. LR pair modules captures a diverse range of LR profiles, finer the cluster-level analysis. For example, in skin cells, Module 5 includes Ccl5–Ccr5 link present in AG and AG CPI, epithelial, myeloid and T cells.

(7/n) With the Nik Joshi Lab at Yale we presented a new dataset of 39K CD8+ cells from LCMV infections. Interestingly only GSPA-based gene localization finds genes associated with type 1 interferon signaling. DE requires clustering, but DE genes in clusters do not reveal this signature!