As stated earlier, a major objective in cancer genome sequencing is to discover somatic mutations. If sequence data from tumor and normal DNA from the same patient is available, candidate somatic mutations can be identified as positions for which p(SNV) is high in the tumor and 1−p(SNV) is high in the normal data. If only tumour data is available, we recommend filtering against dbSNP and performing targeted validation on the remaining positions in both tumor and normal DNA as described previously (Shah et al., 2009b). Moreover, the models we have presented assume identically and independently distributed genotypes. As such, the common prior over genotypes π can be indexed by position (i.e. πi) and thus could encode information about what variants are known for each position i in the genome.
