The SJD values were then used to globally assess tissue-level differences in alternative splicing. A set of 25 human tissues for which at least 20,000 genomically aligned ESTs were available was compiled for this comparison (see Materials and methods) and the SJD values were then computed between all pairs of tissues in this set (Figure 5a). A clustering of human tissues on the basis of their inter-tissue SJD values (Figure 5b) identified groups of tissues that cluster together very closely (for example, the ovary/thyroid/breast cluster, the heart/lymph cluster and the bone/B-cell cluster), while other tissues including the brain, pancreas, liver, peripheral nervous system (PNS) and placenta occur as outgroups. These results complement a previous clustering analysis based on data from microarrays designed to detect exon skipping [24]. Calculating the mean SJD value for a given tissue when compared to the remaining 24 tissues (Figure 5c) identified a set of human tissues including the ovary, thyroid, breast, heart, bone, B-cell, uterus, lymph and colon that have 'generic' splicing patterns which are more similar to most other tissues. As expected, many of
