Fig. 6

The revised scRNA-seq workflow implemented in Piccolo. We contrast the conventional scRNA-seq workflow (left) with the revised scRNA-seq workflow (right) proposed by us and implemented in Piccolo. The starting point for all analyses are the counts matrices. In the conventional workflow, the counts are first normalized and variance stabilized followed by feature selection to identify HVGs. Dimensionality reduction is then performed on the normalized counts. In contrast, in the revised workflow proposed by us, we first perform feature selection to identify both the HVGs as well as stable genes. We rely on the stable genes to estimate cell-specific size factors for performing normalization. During normalization, we also ensure variance stabilization which in turn leads to significant improvement while performing dimensionality reduction with PCA