Supplementary Figure 10: AD-associated circRNAs improve sensitivity and specificity of logistic models predicting AD case status.

The base, genetic-demographic model includes the differential expression covariates (post mortem interval, transcript integrity number, age of death, batch, sex, genetic ancestry) as well as number of APOE4 alleles. The Circ model includes normalized counts for the top 10 circRNAs most significantly associated with CDR on meta-analysis. The Circ+Base model combined the previous two models together. PCtx: Parietal discovery dataset (ncontrol = 13, nAD = 83); BM10: Brodmann Area 10 replication dataset (ncontrol = 46, nDefinite AD = 108); BM22: Brodmann Area 22 replication dataset (ncontrol = 38, nDefinite AD = 97); BM36: Brodmann Area 36 replication dataset (ncontrol = 41, nDefinite AD = 91); BM44: Brodmann Area 44 replication dataset (ncontrol = 40, nDefinite AD = 89).