Due to the high dimension of scRNA-seq datasets, most scRNA-seq clustering methods cluster cells after dimension reduction. However, the clustering results may be heavily affected by the method of dimension reduction, and small clusters can be distorted or obscured during dimension reduction. Baker Center personnel (Peng and Dorman) are working on a mixture of factor scores model for joint clustering and dimension reduction on such datasets. A novel computational framework for the mixture factor analysis model is proposed, integrating a recent proposed matrix-free method for factor analysis, FAD (developed by Baker Center member Dutta), in an EM framework.
