AbstractBackgroundHigh‐throughput proteomic platforms are an useful strategy to detect novel biomarkers and pathways involved in Alzheimer’s disease (AD) development. Due to the novelty of these techniques, an extensive quality control assessment is needed to ensure the reliability of those protein measurements. In this study, we aimed to assess the reproducibility and reliability of SomaScan and Olink proteomic platforms and select top‐performing proteins for further analysis.MethodTo assess CSF proteomic measurements, we used the SomaScan 7k platform (an aptamer‐based technique) measuring 7335 human proteins, and Olink proteomics platform (Explore panel, an antibody‐based technique), measuring 3072 proteins in a subset of 509 individuals from ACE Alzheimer Center Barcelona (Spain). To assess intra‐ and inter‐platform reproducibility and reliability, we evaluated 1) the correlations of aptamers of two studies carried out in the same samples with SomaScan platform and 2) the correlation between SomaScan vs Olink proteomics, respectively. Then, we created a metric of 6 categories of reproducibility combining both intra and inter‐ correlation analysis, considering good (r > 0.5), modest (0.2 > r ≥ 0.5), poor (r ≤ 0.2) correlation. We considered proteins in the top 3 categories as potential candidates to further analysis.ResultFrom the intra‐platform analysis, only 2402 proteins (35.45%) showed a Pearson’s correlation higher than 0.5. Regarding inter‐platform assessment, 2160 proteins were overlapping between both platforms. From these overlapping proteins, 712 proteins exhibited a Pearson’s r > 0.5 (26.67%). We considered potential candidate proteins those with good intra‐ and inter‐platform correlation (n = 687, 9.02%), only good intra‐platform correlation (n = 2015, 26.45%) and modest intra‐ and inter‐platform correlation (n = 5, 0.06%) for analysis.ConclusionThe validation of the reproducibility and reliability of the new proteomics platforms is essential for the correct analysis of the results. The classification of measured proteins in different categories could be useful to increase the confidence in these highly multiplexed proteomic panels.