Using the SPADE Transformation in FCS Express
High-dimensional single-cell technologies, such as Flow, Mass and Image cytometry, can measure dozens of parameters at the single-cell level. FCS Express integrates Spanning-tree Progression Analysis of Density-normalized Events, otherwise known as SPADE, which is a tool that extracts a hierarchy from high-dimensional cytometry data in an unsupervised manner and allows users to visualize multiple cell types in a branched tree structure without requiring the user to define a known cellular ordering.
The final result of the algorithm in FCS Express is a Heat Map plot in which each cell type is depicted as a node of the branched tree. The Heat Map plot can be formatted to color each node based on the expression of a given marker. Moreover, the size of each node can be made proportional to a given statistic (e.g., the number of events within the node).
Gates may be defined on SPADE transformed Heat Map plot to select one or multiple nodes for downstream analysis. Similarly, gates created on classical dot plots can be applied to SPADE plots as well.
Learn more about using SPADE via the FCS Express 6 Manual at:
1. What is SPADE?
- SPADE (Spanning-tree Progression Analysis of Density-normalized Events) is a clustering algorithm. Its goal is to automatically identify cell populations in multidimensional cytometry data files. FCS Express 6 allows users to run SPADE on Flow, Mass and Image cytometry data.
- More details on SPADE can be found in the following publications:
Qiu, P., et al. Extracting a cellular hierarchy from high-dimensional cytometry data with SPADE. Nat. Biotech. 29, 886-891 (2011).
Qiu, P. Inferring phenotypic properties from single-cell characteristics. PLoS One. 5, Web (2012).
2. How does SPADE differ when compared to other transformations, like tSNE, PCA or K-Means?
- SPADE and k-Means are examples of clustering algorithms. The goal of the algorithms is to group data (i.e., cells) together in a way that data in the same group (i.e., the same cluster) are more similar to each other than data from other groups (i.e., other clusters). With respect to k-Means, SPADE includes a series of computational steps that:
- Allow rare events to form their own clusters instead of being outnumbered by abundant cell populations during clustering.
- To generate a minimum spanning tree representation in which clusters are connected with minimum total edge length. t-SNE and PCA are examples of dimensionality reduction algorithms. A dimensionality reduction algorithm attempts to map high-dimensional data into lower dimensions by finding a data set in low-dimensional space that, in some way, represents the original high-dimensional data.
3. Why use SPADE in FCS Express?
- There are many reasons to choose FCS Express to run the SPADE transformation:
- The user interface that allows users to run SPADE in FCS Express is very intuitive and easy to use.
- SPADE analysis in FCS Express is completely integrated with the layout, meaning that the user can draw gates on SPADE plots to select one or multiple nodes for downstream analysis. Similarly, gates drawn on classical dot plots can be applied to SPADE plots as well.
- SPADE is extremely fast in FCS Express, as the team at De Novo Software made great efforts to enhance running of the algorithm and the software in general.