Illustration of the steps in the BALBOA ORF classification algorithm. This figure shows the various step of the BALBOA ORF prediction algorithm. In step 1 the expression dataset is divided into its annotated genes and unannotated (unclassified) ORFs. In step 2 biclusters are generated in the annotated gene set only. In step 3 selected biclusters (where E ≥ E
) are used to classify similarly expressed ORFs in the unclassified set. In step 4 ORFs are combined into weighted frequency list for each functional category. Each ORF label weight is derived from the functional enrichments of the classifying biclusters. In step 5 the top ORFs (where F ≥ F
) are selected from this list. In step 6 ORFs consistently classified across independent datasets are returned.