Although each of these software packages is relatively straightforward, important advantages of PAC are that it allows detection of unique outlier exons without any prior knowledge of the encoding gene or its transcript Oligomycin A structure and that it does not require predefined subgroups of samples with differential expression of the outlier exons. As with any global screening strategy, PAC has its preconditions for detecting outlier exons. First and foremost, identification of outlier exons requires their transcript expression level to be within the linear detection range of the exon array, which is determined by their transcript expression level as well as the hybridization efficiency and specificity of the probe sets involved. The constituency of the test samples is another consideration, particularly when both mutant and wild-type CUDC-907 transcripts may be expressed. For example, the breast cancer cell line cohort included two splice site mutants that escaped detection by PAC because each had a second transcript length of major intensity that resulted from cryptic splicing. Furthermore, PAC detection of the EGFRvIII transcript isoform in clinical glioblastomas was determined by the overall expression level of EGFR transcripts, that was near the limits of linear detection in all five EGFRvIII glioblastomas, but also by the ratio of the EGFRvIII isoform versus wild-type EGFR transcripts. A corollary is that PAC performance may be compromised in detecting an outlier exon when wild-type transcripts represent more than one-fourth of all transcripts of that particular gene, which could be the case in tumor samples with less than 75% neoplastic cells. However, expression levels of mutant and wild-type alleles typically are disproportional to their allele frequency and detection by PAC thus again is determined by the expression level of the outlier transcript. PAC therefore performs best in the absence of wild-type transcript expression. Homozygous transcripts are predominantly found among tumor suppressor genes, where often one allele is mutated accompanied by loss of the other allele. The influence of allele ratios was further stressed in our simulations of recurrent outlier detection by PAC: The EGFRvIII isoform in GBM67 was detected only once it was present as a unique outlier among 14 samples, whereas it had not been detected in our original PAC screen that included five other EGFRvIII expressing glioblastomas. However, this sub optimal PAC performance appeared not related to the recurrence of outliers, as recurrent outliers were easily identified among cell lines 2 even when present in five out of six cell lines. The simulation experiments also revealed that two cell lines were sufficient to reliably detect outlier exons and that more than eight cell lines did not further improve PAC performance, whereas for clinical tumor samples ten appeared the minimum but twenty would be preferred. How efficient might PAC be in detecting mutations in cancer genomes?