PANTHER was initially released in 2003, and was the first database to combine both phylogenetic and functional data to define protein subfamilies of shared function and sequences3. It was also the first database to associate ontology terms describing function to statistical models (HMMs), which can be used to assign genes -- based on sequence information alone -- to subfamilies and functional classes. The novel concept was to annotate not single genes one at a time, but rather subfamilies of related genes that are likely to share function. We demonstrated the accuracy and comprehensiveness of the classifications on the D. melanogaster genome5. In 2005, we began providing annotations of biochemical pathways, viewable with the new PANTHER Pathway Applet6,7. These pathways were curated from the literature by expert biologists and diagrams were drawn with CellDesigner, a pathway-editing tool using controlled graphical notations to represent pathway knowledge8. In 2006, the first version of gene analysis tools was released4. These were primarily for the analysis of gene expression data. The tools were also designed to handle gene list data from any genome wide experiments.
