Table 3.

Evaluation metrics for marker sets on experimental data. The “Average precision” metric is a weighted average of precision over the clusters. The Matthews correlation coefficient is a summary statistic that incorporates all information from the confusion matrix. See [22] for more information about the classification metrics and [23] for more information about the clustering metrics

General procedure	Methods	Metrics
Supervised classification	Nearest centroid classifier (NCC)Random forests classifier (RFC)	Classification error (1 - accuracy)
		Average precision
		Matthews correlation coefficient
Unsupervised clustering	Louvain clustering	Adjusted rand index (ARI)
		Adjusted mutual information (AMI)
		Fowlkes-Mallows score (FMS)