Skip to main content

Table 4 Parameters and performance (as measured by (10) and manual gating as ground truth) of the best results obtained by optimalFlowTemplates + optimalFlowClassification on \(\mathcal {TS}\)

From: optimalFlow: optimal transport approach to flow cytometry gating and population matching

  \({\mathcal {C}}^2\) \({\mathcal {C}}^{5}\) \({\mathcal {C}}^{7}\)
Median F-measure 0.9441931 0.8530806 0.957045
Database Clustering Complete linkage HDBSCAN Complete linkage
Template Formation Pooling Pooling HDBSCAN
Assigned Cluster 1 6 2
Sample Clustering tclust tclust tclust
Supervised Classification QDA QDA from template Random forest
Assigned Cytometry \({\mathcal {C}}^{1}\)   \({\mathcal {C}}^{8}\)
  \({\mathcal {C}}^{9}\) \({\mathcal {C}}^{14}\) \({\mathcal {C}}^{15}\)
Median F-measure 0.9458429 0.9254252 0.8807339
Database Clustering HDBSCAN HDBSCAN HDBSCAN
Template Formation Pooling k-barycenter k-barycenter
Assigned Cluster 9 1 1
Sample Clustering tclust tclust tclust
Supervised Classification QDA Label transfer with (6) Random forest
Assigned Cytometry \({\mathcal {C}}^{13}\)   \({\mathcal {C}}^{1}\)
  \({\mathcal {C}}^{17}\) \({\mathcal {C}}^{18}\) \({\mathcal {C}}^{26}\)
Median F-measure 0.9679446 0.9575489 0.8316279
Database Clustering HDBSCAN HDBSCAN Complete linkage
Template Formation HDBSCAN HDBSCAN HDBSCAN
Assigned Cluster 7 7 4
Sample Clustering tclust flowMeans tclust
Supervised Classification Random forest Random forest Random forest
Assigned Cytometry \({\mathcal {C}}^{20}\) \({\mathcal {C}}^{20}\) \({\mathcal {C}}^{24}\)
  \({\mathcal {C}}^{27}\) \({\mathcal {C}}^{29}\) \({\mathcal {C}}^{31}\)
Median F-measure 0.9312977 0.9259644 0.931515
Database Clustering Complete linkage Complete linkage HDBSCAN
Template Formation Pooling k-barycenter Pooling
Assigned Cluster 5 6 4
Sample Clustering tclust flowMeans tclust
Supervised Classification Random forest Random forest QDA from template
Assigned Cytometry \({\mathcal {C}}^{28}\) \({\mathcal {C}}^{33}\)  
  \({\mathcal {C}}^{40}\)
Median F-measure 0.8240522
Database Clustering Complete linkage
Template Formation Pooling
Assigned Cluster 6
Sample Clustering tclust
Supervised Classification Random forest
Assigned Cytometry \({\mathcal {C}}^{30}\)
  1. Database Clustering refers to the clustering method used in line 17 in Algorithm 1. Template Formation refers to the method used in line 19 in Algorithm 1. Assigned Cluster refers to the label of the cluster as given in Table 2 to which the new cytometry is assigned. Sample Clustering refers to how we obtain \({\mathcal {C}}^u\) in Algorithm 5. Supervised Classification refers to the method used in line 13 in Algorithm 5. Assigned Cytometry refers to the optimal cytometry in the respective cluster that is used for learning (when applicable)