Iterative feature removal yields highly discriminative pathways

BMC Genomics

Table 4 Pathway classification accuracy on influenza compared to other published results

Acc	Classifier
100.0	B cell maturation and activation + Cell adhesion molecules
93.0	Antigen recognition genes
93.0	Bayesian Elastic Net (from [10])
91.2	Bayesian Lasso (from [10])
93.0	Elastic Net (from [10])
91.2	Lasso (from [10])
91.2	Relevance Vector Machine (from [10])
93.0	SVM-RFE (from [10])

In this table, the accuracy of our best single-pathway and best pathway-pair classifiers are compared to the best classifiers reported by Chen et al. [10]. The same protocol is followed: classifiers are trained using H3N2 data from time intervals 14-16 and tested on H1N1 from the same time period.

ISSN: 1471-2164