, 2006). To ensure that we compared the same labels by genes, we compiled a table where each record contains the following fields: Official Gene Symbol by Nomenclature The Entrez Gene this website ID was used as a key

field for comparison. The gene information was extracted from weekly updated gene information files on NCBI repository (ftp://ftp.ncbi.nih.gov/gene/DATA/GENE_INFO/Mammalia/). The filter process is conducted in two main steps. First, comparison is done between our query data set and the filter data. We split the query list in two: potential filtered transcripts and potential dendritic/axonal transcripts. The second step is the assessment of false candidate after filtering. False-negative candidates arise in the filtered list due coexpression of those candidates in different cell types. Such records are identified and rescued by comparing the filtered list to transcripts that are present in hippocampus pyramidal signaling pathway neurons (Sugino et al., 2006) or are identified by in situ methods either conducted by us (71 in situ probes) or by previous studies (Table S14). False-positive candidates arise in the cleaned list due to genes that were detected by 454 but were not present on the microarray chip from the reference studies. Those genes were checked in the Allen Brain Atlas for pyramidal neuron expression in area CAI of

the hippocampus. The genes that were de-enriched in the investigated area were pulled out of the result list and a false-positive rate was determined. The Gene Ontology analysis was conducted using the Bingo Plug-In (v 2.44) for Cytoscape (Maere et al., 2005). The Cytoscape output is a text file with the

following parameters: term id, term name, p value, x (number of genes from the query list annotated to a certain term), X (number of genes from the query list that are annotated to a specific ontology), n (total and number of genes annotated to a certain term by the rat genome database), N (total number of genes annotated to an ontology by the rat genome database). One file was generated per ontology (biological process, molecular function, and cellular component). We calculated cluster frequency, total frequency, fold change, for each term graph level, where: ClusterFrequency=100∗xX TotalFrequency=100∗nN FoldChange=Cluster FrequencyTotal Frequency. Three biographs with the ancestors of all overrepresented terms in the corresponding ontology were built. An application was developed in order to search for the shortest path from each overrepresented term to the root of its graph and assign the distance as the depth level for the term (Dijkstra, 1959). We used an additional custom application to combine the results from the three ontologies in one file. The file was imported to Microsoft Excel in order to obtain one table per query list. The table was sorted by depth level and fold change. All terms in the table are overrepresented with p value less than 0.