Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2022 Jul 18;23(14):7907.
doi: 10.3390/ijms23147907.

Prognosis of Alzheimer's Disease Using Quantitative Mass Spectrometry of Human Blood Plasma Proteins and Machine Learning

Alexey S Kononikhin  1 , Natalia V Zakharova  2 , Savva D Semenov  3 , Anna E Bugrova  1   2 , Alexander G Brzhozovskiy  1 , Maria I Indeykina  1   2 , Yana B Fedorova  4 , Igor V Kolykhalov  4 , Polina A Strelnikova  2   3 , Anna Yu Ikonnikova  5 , Dmitry A Gryadunov  5 , Svetlana I Gavrilova  4 , Evgeny N Nikolaev  1
Affiliations

Prognosis of Alzheimer's Disease Using Quantitative Mass Spectrometry of Human Blood Plasma Proteins and Machine Learning

Alexey S Kononikhin et al. Int J Mol Sci. .

Abstract

Early recognition of the risk of Alzheimer's disease (AD) onset is a global challenge that requires the development of reliable and affordable screening methods for wide-scale application. Proteomic studies of blood plasma are of particular relevance; however, the currently proposed differentiating markers are poorly consistent. The targeted quantitative multiple reaction monitoring (MRM) assay of the reported candidate biomarkers (CBs) can contribute to the creation of a consistent marker panel. An MRM-MS analysis of 149 nondepleted EDTA-plasma samples (MHRC, Russia) of patients with AD (n = 47), mild cognitive impairment (MCI, n = 36), vascular dementia (n = 8), frontotemporal dementia (n = 15), and an elderly control group (n = 43) was performed using the BAK 125 kit (MRM Proteomics Inc., Canada). Statistical analysis revealed a significant decrease in the levels of afamin, apolipoprotein E, biotinidase, and serum paraoxonase/arylesterase 1 associated with AD. Different training algorithms for machine learning were performed to identify the protein panels and build corresponding classifiers for the AD prognosis. Machine learning revealed 31 proteins that are important for AD differentiation and mostly include reported earlier CBs. The best-performing classifiers reached 80% accuracy, 79.4% sensitivity and 83.6% specificity and were able to assess the risk of developing AD over the next 3 years for patients with MCI. Overall, this study demonstrates the high potential of the MRM approach combined with machine learning to confirm the significance of previously identified CBs and to propose consistent protein marker panels.

Keywords: Alzheimer’s disease; machine learning; mass spectrometry; multiple reaction monitoring; targeted proteomics.

PubMed Disclaimer

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

Figure 1
Figure 1
The results of study participants using psychometric tests and scales: (A) Mini Mental State Examination (MMSE), (B) Boston naming test (BNT), (C) Mattis dementia rating scale (MDRS) and categorical associations subtest. MCI—mild cognitive impairment (nc—non-converter, c—converter); AD—Alzheimer’s disease (mild, moderate, and severe subgroups); VD—vascular dementia and FTD—frontotemporal dementia. Lines inside the boxes show medians; box flanges—25–75 percentiles; whisker range ± SD. Significantly different results, with p-values < 0.01 according to the Mann–Whitney U test, are shown with the following symbols: * significantly different to the control group; ** to the control and MCI groups; *** to the control, MCI and AD groups; **** to the control and AD groups; ***** to the control, MCI, AD and VD groups; # to the control and MCI-nc; “F”—to FTD; “V”—to VD; “!”—significant differences between subgroups.
Figure 2
Figure 2
Differentiating proteins between the control and AD groups with uncorrected p-values ≤ 0.01 according to the Mann–Whitney U test. Lines inside boxes—medians; box flanges—25–75 percentiles; whisker range ± SD.
Figure 3
Figure 3
The relative ratings of the quantified proteins (AD vs. control). (A) The ordering of 22 proteins with uncorrected p-values of <0.05 by the effect sizes (Cohen’s d). (B) The ordering by the mean feature importance of the proteins included as classifiers in this study. The abbreviation for the name of proteins along the Y axis is given in accordance with Table 2. The color saturation represents the value of feature importance/p-value of the feature.
Figure 4
Figure 4
Differentiation of the AD and control samples using the obtained protein panels and developed classifiers. (A) ROC curves for the differentiation of AD vs. the controls for classifiers generated with 4 algorithms and 2 sets of proteins. (B) Distribution of the control and AD samples according to the probability of AD using two classifiers with the best performances.
Figure 5
Figure 5
The distribution of MCI, VD and FTD patients according to the probability of AD development using the RF-31 classifier. (A) Distribution of MCI patient samples with a known three-year prognosis (samples of MCI-nc patients with <3 years of observation are excluded). (B) The distribution of the VD and FTD samples.
Figure 6
Figure 6
Venn diagram showing proteins differing between AD, FTD and the control at the uncorrected p-value of ≤ 0.05. Bold type shows proteins with a p-value ≤ 0.01.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Prince M., Bryce R., Albanese E., Wimo A., Ribeiro W., Ferri C.P. The global prevalence of dementia: A systematic review and metaanalysis. Alzheimer’s Dement. 2013;9:63–75. doi: 10.1016/j.jalz.2012.11.007. - DOI - PubMed
    1. Alzheimer’s Association 2016 Alzheimer’s disease facts and figures. Alzheimer’s Dement. 2016;12:459–509. doi: 10.1016/j.jalz.2016.03.001. - DOI - PubMed
    1. Perrin J., Fagan A.M., Holtzman D.M. Multimodal techniques for diagnosis and prognosis of Alzheimer’s disease. Nature. 2009;461:916–922. doi: 10.1038/nature08538. - DOI - PMC - PubMed
    1. Mollenhauer B., Bibl M., Wiltfang J., Steinacker P., Ciesielczyk B., Trenkwalder C., Otto M. Total tau protein, phosphorylated tau (181p) protein, β-amyloid1-42, and β-amyloid1-40 in cerebrospinal fluid of patients with dementia with Lewy bodies. Clin. Chem. Lab. Med. 2006;44:192–195. doi: 10.1515/CCLM.2006.035. - DOI - PubMed
    1. Ritchie C., Smailagic N., Ladds E.C., Noel-Storr A.H., Ukoumunne O., Martin S. CSF tau and the CSF tau/ABeta ratio for the diagnosis of Alzheimer’s disease dementia and other dementias in people with mild cognitive impairment (MCI) Cochrane Database Syst. Rev. 2013;3:CD10803. doi: 10.1002/14651858.CD010803. - DOI - PMC - PubMed