Applying the same preprocessing procedure resulted in 156 CNV regions with 419 genes associated with rare inherited events in autistic children. Using the data described above, Selleck Ruxolitinib we identified statistically significant gene clusters affected by de novo CNV events associated with autistic individuals. Significant clusters detected using either one-gene-per-CNV (p value = 0.02) or the two-genes-per-CNV (p value = 0.02) clustering are shown in Figure 2. If genes forming the high scoring clusters were masked, no other significant clusters were detected in the data. In contrast, no statistically significant clusters were obtained using ultrarare inherited CNVs from affected individuals (the

best cluster p value = 0.6) The absence of a significant cluster in the ultrarare inherited data set, which had a comparable number of genes to the de novo events, suggests that the inherited CNVs contain a significantly smaller fraction of casual genes, i.e., genes associated with autistic phenotype. This conclusion is also supported by the observation by Levy et al. (2011) that there is less bias in transmission of ultrarare find more inherited events. In other words, autistic children were almost as likely as their unaffected siblings to inherit an ultrarare event. This is in sharp contrast to the de novo events, which were nearly four times more frequent in the autistic children

(7.9% in autistic children versus 2.0% in unaffected siblings). The contribution of each gene to the cluster score, i.e., its functional connection to other cluster genes, is not uniform. To capture each gene’s contribution to the cluster score we performed Markov Chain Monte Carlo (MCMC) simulations, sampling clusters based on their scores. The size

of each gene (node) in Figure 2 not is proportional to the each gene’s membership in high-scoring clusters during the sampling simulations (see Experimental Procedures, Supplemental Experimental Procedures). Similar node sizes were also obtained based on the average connection strength from each gene to the other genes in the cluster (Pearson’s r = 0.8, p value = 4∗10−11). Interestingly, we found that genes affected by de novo CNVs observed in females are significantly more important for the overall cluster score than genes affected by CNVs in males, i.e., female genes have stronger average connections with other genes from the identified network (see Figure S2; one-tail Mann-Whitney test, female > male, p value = 0.013). This observation is illuminating because one of the striking phenotypic characteristics of autism is the male-to-female incidence ratio of more than 5:1 for high-functioning ASD (Newschaffer et al., 2007). It has been previously suggested (Zhao et al., 2007) that stronger genetic perturbations are required, on average, to trigger an autistic phenotype in females than males due to currently unknown compensatory mechanisms.