Analysis of mass spectrometry data from the secretome of an explant model of articular cartilage exposed to pro-inflammatory and anti-inflammatory stimuli using machine learning

Table 1 Comparison of performance accuracies, as percentage of samples correctly classified, for classification of canine articular cartilage data for seven different machine-learning methods, using leave-one-out cross-validation

Dataseter	Naive Bayes	SVM	k-nearest neighbour	JRip (rule based)	Random forest	C4.5	BioHEL
ProteinProphet	39.1	52.2	34.8	43.5	34.8	52.2	73.9
emPAI	52.2	56.5	52.2	78.3	39.1	73.9	56.5
ProteinProphet and emPAI combined	52.2	52.2	43.5	78.3	26.1	47.8	78.3

For the ‘ProteinProphet and emPAI combined’ the two scores were combined into one dataset. The highest accuracies achieved in each dataset are shown in bold.

ISSN: 1471-2474