Fig. 3

Integrated analysis of cortex and CSF proteomes unveils consistent CSF biomarkers in AD across independent studies. a Scheme for the integration of cortex and CSF proteomes. b Venn diagram shows overlap of quantified proteins in cortex discovery cohort (data i) and CSF discovery cohort (data v). c Integration of cortex and CSF proteomes identifies consistent CSF biomarkers in AD across independent studies. Proteins quantified in both cortex and CSF are plotted as a function of their Z score comparing AD to Ctl in cortex (x-axis) and their Z score comparing AD to Ctl in CSF (y-axis). Fourty-four DE proteins with Z score difference > 2 and FDR < 0.2 in both proteomes are plotted in black. Proteins consistently showed up as AD biomarkers in all three independent CSF studies are labeled in red, proteins stood out in this study and reference study 1 (Higginbotham L, BioRxiv, 2019) are labeled in blue, and proteins emerged in this study and reference study 2 (Sathe G, Proteomics Clin Appl. 2019) are labeled in turquoise. Red dashed lines indicate Z value difference > 2 in CSF and cortex. d Heatmap shows Z score transformed log2 level fold changes and -log10 FDR values of the 44 DE proteins comparing AD to control in cortex and CSF proteome datasets. e Integration of DE proteins and protein-protein interaction (PPI) database unveils enrichment of amyloid pathology and mitochondrial functions. Protein-protein interaction modules were derived from superimposing the 44 DE proteins along with APP and TAU on STRING PPI database. The interaction modules were build based only on the most confident interaction sources including experiments and database. The default statistic criteria of STRING with a cutoff of minimum interaction score of 0.4 was applied to derive PPI modules. The pairwise PPI interactions scores are displayed next to the edges