Casual analysis of cell-cell communication with single-cell and spatial omics data

Deciphering cell state dynamics from single-cell gene expression data using applied topology and optimal transport

Mentors:
Lead: Zixuan Cang (Mathematics, NC State)
Collaborator: Nicholas Buchler (Molecular & Biomed Sciences, NC State)

Outline:
Cells grow and replicate through the cell cycle process, which is crucial to maintaining homeostasis, repair, and regeneration. Recent single-cell gene expression data enables the investigation of cellular processes with great resolution. Yet, extracting cell cycle mechanisms remains challenging due to a lack of prior knowledge and high dimensionality [17]. DRUMS students will explore the innovative integration of applied topology[18] and optimal transport [19] with deep learning to reveal the cell cycle structures from single-cell data and reconstruct the spatiotemporal dynamics of cell cycles. They will also use data-driven modeling to identify key factors controlling cell cycles and predict impacts of perturbations. We will generalize this approach to study general non-cycle cell state dynamics as well.

Objectives:
(i) Learning cyclic embeddings of data representing cell cycles using topologically guided deep learning. (ii) Model cell cycle dynamics on the learned manifold using optimal transport.  (iii) Identify cell cycle driving genes through the reconstructed spatiotemporal dynamics. (iv) Generalizing the approach to general cell state dynamics studies.

Outcomes:
Computational and deep learning methods and software package for the new analysis of cell cycle and cell state dynamics from single-cell gene expression data.

References:
1. Cang Z, Zhao Y, Almet AA, Stabell A, Ramos R, Plikus M, Nie Q. Screening cell-cell communication in spatial transcriptomics via collective optimal transport. Nature Meth. 2023;20:218-228

2. Jin S, Guerrero-Juarez C, Zhnag L, Chnag I, Ramos R, Kuan C, Nie Q. Inference and analysis of cell-cell communication using CellChat. Nature communications. 2021;12:1088.

3. Tan X, Yang S, Ye W, Faries D, Lipovich I, Kadziola Z. When doubly robust methods meet maching learning for estimating treatment effects from real-world date: A comparative study. asXiv. 2022;arXiv:2204.10969.