EM Clustering with Single-Cell Transcriptomic Data

Expectation-Maximization (EM) -based clustering algorithm to identify groups defined by biological variates as clusters in single-cell transcriptomic data.

Quick links

• Project summary report [PDF].
• Data subsets [CSVs].

Clustering results with Poisson and Gaussian Priors

• Data with only sex as biological variate - neither captured by Poisson nor Gaussian.
  

• Data with sex and cell ontology as biological variates - cell ontology captured by Gaussian prior.
  

Key references

• Tabula Muris Consortium, et al. "Single-cell transcriptomics of 20 mouse organs creates a Tabula Muris." Nature 562.7727 (2018): 367-372.
• Moon, Todd K. "The expectation-maximization algorithm." IEEE Signal processing magazine 13.6 (1996): 47-60.
• Dempster, Arthur P., Nan M. Laird, and Donald B. Rubin. "Maximum likelihood from incomplete data via the EM algorithm." Journal of the royal statistical society: series B (methodological) 39.1 (1977): 1-22.