As the results above were somewhat intriguing, but not necessarily all pointing to the same direction, the last step I took was to try to find correlations between CNV burden and different phenotypic variables. Brain variables weren't as good as I'd like, mostly because there are only 25 kids (out of the possible 51 in the simplex study) with imaging, and that's even before taking QC into consideration. Wendy confirmed that.
Still, in Figure \ref{526465} I show an exploratory search of correlations between different gene burden pipelines (Y) and phenotypes (X). By pipelines I mean different combinations of the things I tried (see below). I'm not too worried about multiple comparison corrections yet, as most of the pipelines in the Y axis can be removed. But I do need to calculate appropriate p-values for each cell in the heatmaps as the N for each phenotype varies. All in all, there seems to be some interesting phenotypes correlated with CNV burden. I think we're in a good place that, if we add a few more phenotypes and burden pipelines, some cooler stuff will come up.