Workflow from Scientific Research

We build the count matrix using different bin sizes as well as a GeneTSS annotation and peaks called on the pseudo bulk with both MACS2 and SICER (only for the human PBMC dataset), for the MDA-MB468 dataset we also used author provided annotations. We then simulate in silico different experimental conditions for studying the role of the number of cells in a dataset, and the effect of the coverage per cell, as well as different feature selection strategies. Afterwards, we run 9 different dimension reduction methods to obtain the cell representations.

DOI