How should I choose the number of clusters?

All of the clustering methods in OmicsAnalyst will automatically detect a reasonable number of clusters, however this can be updated on the Parameter Tuning page for SNF and Spectrum. Several diagnostic plots are provided to help users decide whether they want to keep, increase, or decrease the number of clusters to detect for these options. For example, the image below shows the results from using SNF on the pregnancy example data.

The algorithm determined that two clusters optimize some measure of variance on the integrated proteomics and metabolomics sample-sample similarity matrix. PCA plots of the individual 'omics data sets show sample-cluster membership.