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

BMC Bioinformatics

Table 3 Table of F-measure statistics as given by (7), where we use the manual gating as the ground truth

	flowMeans	DeepCyTOF	optimalFlowTemplates + DeepCyTOF	optimalFlowTemplates + optimalFlowClassification
\({\mathcal {C}}^2\)	0.8988	0.9546	0.9736	0.9610
\({\mathcal {C}}^5\)	0.8977	0.9161	0.9196	0.9587
\({\mathcal {C}}^7\)	0.9508	0.7514	0.9769	0.9768
\({\mathcal {C}}^9\)	0.8936			0.9172
\({\mathcal {C}}^{14}\)	0.9004	0.9838	0.9530	0.9066
\({\mathcal {C}}^{15}\)	0.8974	0.9408	0.9352	0.9556
\({\mathcal {C}}^{17}\)	0.9405	0.7847	0.9810	0.9848
\({\mathcal {C}}^{18}\)	0.9004	0.7837	0.9796	0.9849
\({\mathcal {C}}^{26}\)	0.9024			0.9313
\({\mathcal {C}}^{27}\)	0.8645			0.9306
\({\mathcal {C}}^{29}\)	0.9475			0.9744
\({\mathcal {C}}^{31}\)	0.9290			0.9656
\({\mathcal {C}}^{40}\)	0.9330			0.9538

First column: results of the unsupervised gating procedure flowMeans on \(\mathcal {TS}\). Second column: results of the supervised procedure DeepCyTOF on \(\mathcal {TS}'\). Third column, results of DeepCyTOF on the clusters \(\mathcal {TS}'_1\), \(\mathcal {TS}'_2\) and \(\mathcal {TS}'_3\) produced by optimalFlowTemplates. Forth column: results of our supervised procedure optimalFlowTemplates + optimalFlowClassification on \(\mathcal {TS}\). In underline we have the best performance according to the F-measure

ISSN: 1471-2105