CNV-propensity score parameters.

Each of the three parameters significantly improves the likelihood model predicting whether a gene overlaps a rare deletion within the meta-controls. We tested three parameters: (1) the gene size, which represents a straightforward parameter representing the length of the gene in Mb, (2) the gene neighborhood density score is the average number of additional genes that a CNV overlapping the gene overlaps, and (3) the gene structure score which is the proportion of CNVs overlapping a gene that is fully intronic. For each parameter we list the multivariate logistic regression beta and 95% confidence interval. We also list the statistical significance of the change in log-liklihood by removing the variable from the model.