Fig. 4 |. Use of cluster and combine methodology to define cell types.

Gene expression profiles for individual cells are noisy and incomplete, with the holes in each shape representing the missing information. In this example, there are three different cell types, as indicated by different colours and shapes. Although the data sets for each individual cell are incomplete, they are sufficient to allow clustering of similar cells into groups. The data for all cells within a group are then combined to give a robust view of that cell type. The gene expression information missing in one cell can be provided by other cells in that group. Thus, through complementation, each cell contributes something to the total picture, resulting in a combined profile of the cell type that is very complete, enabling detection of very low gene expression levels.