Fig. 1

Comprehensive integration of ultra-deep cortex, CSF and serum proteome datasets for biomarker analyses in Alzheimer’s disease. a Workflow for data integration. The human brain cortex proteomes consist of 4 datasets including 2 discovery cohorts (data i and ii) that were validated by 2 reference cohorts (data iii and iv). The human cerebrospinal fluid (CSF) proteomes consist of 4 individual datasets including one discovery cohort (data v) that was validated by 3 reference cohorts (data vi, vii, viii). Differential expression (DE) of CSF proteome was carried out through LIMMA R package, and then integrated with cortex proteome. Next, the human CSF proteome was compared with mouse CSF (data ix). Finally, the cortex and CSF proteomes were integrated with the serum proteome (data x). b Principle component analyses (PCA) of discovery proteomes. Dot plots show two-dimensional principle component analyses of all quantified proteins in the representative discovery datasets including human cortex (ii), Human CSF (v), and human serum (x). Protein expression values of all datasets were log2-transformed for PCA analyses. c Advanced tissue proteome profiling pipeline achieves ultra-deep proteome coverage in cortex. The unique proteins quantified in the cortex discovery cohorts were combined and then compared to the cortex transcriptome with consensus normalized expression (NX) values > 1 in the Human Protein Atlas database. d Advanced biofluid proteome profiling platform achieves ultra-deep proteome coverage in human CSF and serum. The CSF proteome was compared to the two deepest MS-based CSF proteome studies in AD so far, Reference study 1 (data vi, Higginbotham L, BioRxiv, 2019) and Reference study 2 (data viii, Sathe G, Proteomics Clin Appl. 2019), similarly the serum proteome data was compared to the two recent MS-based AD serum protein biomarker studies, Reference study 3 (Ashton N, Science Advances, 2019) and Reference study 4 (Lan J, Journal of Proteome Research, 2018)