Fig. 6

Genome-wide distribution of predicted CNVs in neurons. A Overview of cell and CNVs elimination steps in the uncorrected and corrected data. We applied the same criteria for excluding CNVs to the PCA-corrected results. *We discarded CNVs if their breakpoints were within the three base pairs window around each other. B Bar plots represent the number of CNVs per cell for AD and control groups. Blue area: 1-somy (deletions); pink area: 3-somy (duplications). The number of unique individuals with at least one cell identified with a reliable CNV is indicated within the panel. C The table shows the ploidy levels of cells in corrected and uncorrected versions of the data. D The table shows several combinations of fixed factors that were used to test the difference between AD and control in the frequency of CNVs (“Freq of CNVs”). The fixed factors of the model included: diagnoses (AD and control), chromosomes (“Chr”, 22 autosomes as categorical variable), sex (male and female), brain regions (“Brain reg”, temporal cortex, hippocampus CA1, hippocampus CA3, cerebellum, entorhinal cortex) and coverage. Considering correlation due to repeated measurements on the same subject, individual (“Indiv”) effects were added as a random factor. In addition to the interaction of chromosomes and diagnoses (i.e. “Chr * Diagnosis + Coverage”), the effects of the chromosomes were also tested individually (i.e. “Chr + Diagnosis + Coverage”. The results were not significant across any of the seven GLMM models tested, in any of the analyses datasets. In the main text, we report the lowest “diagnoses” p-value across the seven tests. Note that we did not apply any multiple testing correction. E The heatmap shows the genome-wide copy number profile of cells analyzed in the corrected data (n = 10) with at least one reliable CNV. CNVs, brain regions and diagnoses are illustrated in different colors (see the color key on the left of the figure). Each row shows a cell and each column shows a chromosome