In this lab, whole mount in situ hybridization has been performed using the cDNAs that have been classified in our EST project (See Thierry-Mieg et al.). Thus far, in situ analysis for some 7600 genes was finished. Minimal annotation has been given to the in situ images; 10 stages for embryogenesis, 4 stages for larval-adult stage, on average 10 patterns (cell(s), tissue, region) per stage, 3 levels of relative intensity of signals per each pattern. Using the information ((10+4)
x10x3 = 420 dimensions), we have done clustering analysis of the in situ patterns. The gene expression patterns were roughly clustered by the factor analysis using the Euclid distance. This treatment classified the embryonic patterns of 5064 genes into 39 clusters that contained 5 to 3355 genes per cluster (12 clusters had more than 50 gene members). Then, we focused on several cell/lineage/stage-specific clusters; gut specific cluster (910 genes), hypodermis specific cluster (421 genes), and body wall specific cluster (72 genes). The three clusters were further classified by the Ward method. This treatment produced 7 sub-clusters along the developmental time course from the gut cluster and 6 sub clusters from the hypodermis and body wall muscle clusters. This means that we know the time of transcription initiation of the member genes in the individual cell lineages and that the members. The same sub-cluster might be regulated by the same mechanisms. Thus, we are now trying to extract common sequence motifs, possible candidates for transcriptional regulatory signals, in the 5'- upstream region of the genes in the same sub-cluster.