FlowSOM
A self-organizing map based approach for the automated analysis of cytometry data.
Description:
FlowSOM is a clustering and cluster visualisation technique for analysing cytometry data. For clustering, FlowSOM uses a Self-Organising Map (SOM), an artificial neural network consists of interconnecting nodes forming a grid, and consensus hierarchical clustering. FlowSOM’s clustering process is divided into two steps. In the first step, cells are assigned to the nodes in the SOM grid, culminating in the SOM grid shaped to resemble the distribution of the data. In the second step, a consensus hierarchical clustering is used to cluster the SOM nodes by repeatedly subsampling and hierarchically merging the SOM nodes. The second step culminates in the construction of meta-clusters, derived based on the number of times the SOM nodes are clustered together by consensus hierarchical clustering. To aid cluster validation, annotation, and visualisation, the SOM nodes can be visualised as a collection of pie charts organised into either a Minimum Spanning Tree (MST) or a table (star chart).
More Information:
Originates from the AI-Driven Data Science research challenge (FAIR 2019–2023).
Responsible Research Lead: Yvan Saeys
Authors: Sofie Van Gassen, Katrien Quintelier, Artuur Couckuyt, Annelies Emmaneel, Yvan Saeys
Programming Language: R
Categories: Single cell and DNA sequence analysis, networks and graphs, un/semi-supervised learning
Publications:
K. Quintelier, A. Couckuyt, A. Emmaneel, J. Aerts, Y. Saeys and S. Van Gassen "Analyzing high-dimensional cytometry data using FlowSOM", Nat Protoc 16, 3775–3801 (2021).