Skip to main content
Download PDF
Download ePub

Lipoproteins and metabolites in diagnosing and predicting Alzheimer’s disease using machine learning

Lipids in Health and Disease volume 23, Article number: 152 (2024) Cite this article

Abstract

Background

Alzheimer’s disease (AD) is a chronic neurodegenerative disorder that poses a substantial economic burden. The Random forest algorithm is effective in predicting AD; however, the key factors influencing AD onset remain unclear. This study aimed to analyze the key lipoprotein and metabolite factors influencing AD onset using machine-learning methods. It provides new insights for researchers and medical personnel to understand AD and provides a reference for the early diagnosis, treatment, and early prevention of AD.

Methods

A total of 603 participants, including controls and patients with AD with complete lipoprotein and metabolite data from the Alzheimer’s disease Neuroimaging Initiative (ADNI) database between 2005 and 2016, were enrolled. Random forest, Lasso regression, and CatBoost algorithms were employed to rank and filter 213 lipoprotein and metabolite variables. Variables with consistently high importance rankings from any two methods were incorporated into the models. Finally, the variables selected from the three methods, with the participants’ age, sex, and marital status, were used to construct a random forest predictive model.

Results

Fourteen lipoprotein and metabolite variables were screened using the three methods, and 17 variables were included in the AD prediction model based on age, sex, and marital status of the participants. The optimal random forest modeling was constructed with “mtry” set to 3 and “ntree” set to 300. The model exhibited an accuracy of 71.01%, a sensitivity of 79.59%, a specificity of 65.28%, and an AUC (95%CI) of 0.724 (0.645–0.804). When Mean Decrease Accuracy and Gini were used to rank the proteins, age, phospholipids to total lipids ratio in intermediate-density lipoproteins (IDL_PL_PCT), and creatinine were among the top five variables.

Conclusions

Age, IDL_PL_PCT, and creatinine levels play crucial roles in AD onset. Regular monitoring of lipoproteins and their metabolites in older individuals is significant for early AD diagnosis and prevention.

Background

Alzheimer’s disease (AD) is a chronic neurodegenerative disorder and a protein-conformational disease that is primarily caused by abnormal processing and aggregation of normally soluble proteins [1,2,3]. In the brains of patients with AD, β-amyloid proteins (Aβ) aggregated into plaques, and tau proteins aggregated abnormally within neurons [4,5,6]. The presence of these aggregates not only disrupts the normal functioning of neurons but also leads to cell death and brain tissue degeneration [7,8,9]. The pathological changes in AD are closely associated with the imbalanced metabolism of Aβ, leading to the formation of senile plaques, as well as the excessive phosphorylation of tau protein, which results in the formation of neurofibrillary tangles in neurons [10,11,12].

Some studies have reported metabolic disturbances in patients with AD, suggesting a close association between AD onset and metabolic imbalance [13, 14]. Normal lipid metabolism is crucial for maintaining the proper functioning of the brain [15,16,17]. Lipid metabolism disorders can lead to synaptic loss and, ultimately, memory impairment through pathways such as inflammatory response, oxidative stress, blood–brain barrier damage, mitochondrial dysfunction, and neuronal signaling pathway damage [18]. For example, cholesterol is a major component of cell membranes and myelin sheaths, playing a crucial role in maintaining synaptic integrity and neuronal function. Amyloid precursor protein (APP) trafficking, proteolytic cleavage, and Aβ aggregation related to the core pathological process of AD are all related to biofilms and are affected by membrane components [19, 20]. Whether it is increased cholesterol levels, ApoE proliferation, trace amounts of sodium sulfate, or decreased plasmalogens, they all contribute to dysfunction in Aβ processing, ultimately leading to AD. Even the slightest changes in lipid concentrations can have a significant impact on the progression and severity of AD [20]. Therefore, lipid distribution and metabolism may affect the pathogenesis of AD. Moreover, cleavage products of APP accumulate at mitochondria-associated endoplasmic reticulum membranes (MAMs), where it impairs mitochondrial bioenergetics, disrupts cellular lipid homeostasis, and leads to alterations in membrane lipid components common in AD pathogenesis [21, 22].

Lipoproteins are globular particles composed of a hydrophobic core rich in sterol lipids and triglycerides and an outer shell composed of proteins, phospholipids, cholesterol, etc. Lipoproteins can be divided into chylomicrons (CM), very-low-density lipoproteins (VLDL), intermediate-density lipoproteins (IDL), low-density lipoproteins (LDL), and high-density lipoproteins (HDL). Metabolites are small molecules of substrates, intermediates, and products in cellular metabolic processes. Abnormal lipid metabolism may lead to excessive accumulation of Aβ and abnormal phosphorylation of tau protein [23, 24]. This has a certain suggestive effect on the prediction of AD. Researchers have explored the study of Aβ and tau proteins, making significant contributions to early disease diagnosis and monitoring disease progression. However, the exact mechanisms underlying AD onset remain unclear [25], and the increasing prevalence of AD poses a substantial economic burden on nations and affected families [26].

Random forest is effective in predicting AD [27, 28]. By integrating the predictions of multiple decision trees, model accuracy can be enhanced, making random forests suitable for complex classification problems [29, 30]. Moreover, random forests are highly resilient against noise and anomalies [31], making them capable of effectively handling complex real-world data. We hypothesize that machine learning methods such as random forests can effectively screen out important variables from high-dimensional data, and some key lipoproteins and metabolites may play an important role in the development of AD. Therefore, this study aimed to use various machine learning methods to screen risk factors for AD and analyze key lipid proteins and metabolite factors that influence the onset of AD using a random forest model.

Methods

Data sources

Data collected from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (https://adni.loni.usc.edu/) were used to assist in the early diagnosis and tracking of AD. Based on the status of the participants and availability of lipoprotein and metabolite data, 603 participants were recruited between 2005 and 2016, including 294 controls and 309 patients with AD. To achieve the research objectives, the lipoprotein and metabolite data of the participants were filtered, excluding variables with missing values. Ultimately, 213 lipoprotein and metabolite variables were included in this study. All 603 participants in this study had complete records for the 213 lipoprotein and metabolite variables.

Variable screen and predictive model construction methods

In this study, three machine learning methods, random forest, CatBoost algorithm, and Lasso regression, were used to screen key lipoproteins and metabolites. First, the 213 variables were sorted and filtered based on their importance, with any variables (lipoproteins and metabolites) ranking high in importance in ≥ 2 machine learning methods included in the model. Subsequently, the selected variables, along with the age, sex, and marital status of the participants, were used to construct a random forest predictive model to identify the key factors influencing AD onset. The random forest is trained by randomly selecting a subset of features, and only some features in each decision tree participate in the partitioning process, which effectively reduces the correlation between features and provides a more reliable feature importance ranking. The CatBoost algorithm can automatically discover and utilize the interaction between features in the process of building a decision tree, better capture the nonlinear relationship between features, and improve the expressive ability of the model.

In this study, while using the Lasso regression to screen variables, the original data were first standardized with Z-scores to ensure that each feature in the model was properly standardized, thereby improving the stability of the model and reliability of the results. The result of standardization was such that the mean of each variable was 0, and the standard deviation was 1. However, the random forest and CatBoost algorithms are tree model-based algorithms unaffected by the feature scale; therefore, no standardization was applied to the data, and the analysis was conducted using raw data.

The categorical variables used in the stochastic forest prediction model were assigned different values according to their attributes. Sex was coded 1 for males and 0 for females. Marital status was coded 1 for married individuals and 0 for those who were unmarried, widowed, or divorced. Apart from sex and marital status, the other variables were continuous. The main variables and their abbreviations are presented in Table 1.

Table 1 Names and abbreviations of the main variables in this study
Full size table

In this study, a non-replacement random sampling method was used to build the random forest predictive model. The research cohort was divided into training and testing datasets in an 8:2 ratio. The training set was used to establish the random forest model, and the testing set was used to evaluate the model’s performance. Optimizing model parameters aimed to enhance the model’s ability to capture the complexity of the data [32], which was achieved by calculating the optimal number of features “mtry” and decision trees “ntree” used in each tree of the model [33]. Out-of-bag error is an estimate of the model’s performance on unused data by using out-of-bag data. The mean error rate based on out-of-bag error is an important metric used to evaluate the performance of classification models. It indicates the proportion of errors in the prediction process, and a lower mean error rate usually indicates that the model has higher accuracy and generalization performance. In this study, the minimum mean error rate of the model was calculated to determine the optimal number of feature “mtry” used by each tree in a random forest. Then, by plotting the relationship between the model error and the number of decision trees, the optimal number of decision trees used by the model is determined. Finally, utilizing the two optimal parameters, along with the settings of importance = true and num_class = 2, a random forest model is constructed.

Statistical analysis

In this study, SPSS 21.0 (IBM Corp., Armonk, NY, USA) software was used to compare the differences in basic information between the NC group and the AD group by using two independent samples t-test or chi-square test or Mann–Whitney test, and the test level was α = 0.05. The random forest package in R version 4.2.2 was used to construct the random forest model, glmnet package for Lasso regression, and CatBoost package for the CatBoost model.

Results

Basic information about the participants

A total of 603 participants were recruited for this study: 294 controls and 309 in the group with AD. The average age of the controls was (74.72 ± 5.93 years), consisting of 136 males (46.26%) and 158 females (53.74%). Among them, 201 (68.37%) were married, and 93 (31.63%) were unmarried. The mean age of the group with AD was (74.34 ± 7.68 years), including 175 males (56.63%) and 134 females (43.37%). In the group with AD, 260 individuals (84.14%) were married, and 49 (15.86%) were unmarried. There was no statistically significant age difference between the two groups, but there were statistically significant differences in gender and marital status. Since this study mainly aimed to explore the key lipoproteins and metabolites that affect the incidence of AD, only the three basic variables of age, gender, and marital status were included in the final modeling, and the three variables of ethnicity, education scores, and MMSE were not included in the study. The basic information of the two groups is shown in Table 2.

Table 2 Baseline data of the NC group and AD group
Full size table

Variable screening results based on the random forest method

In the random forest model, the Mean Decrease Accuracy (MDA) and Mean Decrease Gini (MDG) are crucial indicators for assessing variable importance. The MDA evaluates the contribution of each variable to the model accuracy, whereas the MDG measures the improvement in Gini impurities during the decision-tree splitting process. Higher values of these indicators correspond to more significant variables.

This study used two indices to rank the importance of the 213 lipoproteins and metabolites. The top five important variables selected using the MDA indicator were glycoprotein acetylation (GLYCA), phospholipids to total lipids ratio in intermediate-density lipoproteins (IDL_PL_PCT), phospholipids to total lipids ratio in small high-density lipoproteins (S_HDL_PL_PCT), percentage of cholesteryl esters within low-density lipoprotein (L_LDL_CE_PCT), and cholesteryl esters to total lipids ratio in large HDL (L_HDL_CE_PCT). The top five most important variables selected using the MDG indicators were GLYCA, creatinine, IDL_PL_PCT, acetate, and L_LDL_CE_PCT. The top 20 variables screened using the two indicators are presented in Fig. 1 and Table 3.

Fig. 1
figure 1

Variables screened using the two indicators based on the random forest model

Full size image
Table 3 Variables screened using the two indicators based on the random forest model
Full size table

Variables screened using the Lasso regression model

The Lasso regression model was used to rank 213 lipid and metabolite variables. The coefficients of the variables with high importance were retained, whereas those with low importance were dismissed. In the final selection, 19 variables were retained (Fig. 2). The absolute values of the coefficients indicated the variables’ contributions to the model, with larger values signifying greater contributions. Among the retained variables, the top five with higher absolute coefficients were IDL_PL_PCT, L_HDL_PL_PCT, GLYCA, creatinine, and acetate. The ranking of the importance of these 19 variables is presented in Table 4.

Fig. 2
figure 2

Top 19 variables screened using the Lasso regression model

Full size image
Table 4 Top 19 variables screened using the Lasso regression model
Full size table

Variables screened using the CatBoost algorithm

The CatBoost algorithm was used to assess the importance of 213 lipid and metabolite variables. The magnitude of the gain was used to measure each variable’s contribution to the model performance, with higher gain values indicating greater contributions. The variables were ranked based on the magnitude of their gains. The top five variables were GLYCA, acetoacetate, creatinine, L_HDL_PL_PCT, and glucose. The top 20 variables screened using the CatBoost algorithm are presented in Fig. 3 and Table 5.

Fig. 3
figure 3

Top 20 variables screened using the CatBoost algorithm

Full size image
Table 5 Top 20 variables screened using the CatBoost algorithm
Full size table

Random forest prediction model construction

Based on the random forest, LASSO regression, and CatBoost algorithms, 14 variables were selected: acetate, acetone, albumin, creatinine, glucose, GLYCA, IDL_PL_PCT, L_HDL_CE_PCT, L_HDL_PL_PCT, L_LDL_CE_PCT, L_LDL_PL_PCT, LDL_size, omega_3, and free cholesterol to total lipids ratio in very large HDL. The aforementioned 14 variables, along with age, sex, and marital status, were included in the random forest prediction model. Figure 4 depicts the 14 selected variables.

Fig. 4
figure 4

Venn diagram of the overlap of the selected variables using the three methods

Full size image

The random forest model achieved optimal results with “mtry” (number of features used in each tree) set to 3 and “ntree” (number of trees in the model) set to 300. The model exhibited an accuracy of 71.01%, a sensitivity of 79.59%, a specificity of 65.28%, and an AUC (95% CI) of 0.724 (0.645–0.804). The importance rankings of the variables in the random forest prediction model based on MDG and MDA showed some differences. The top five variables selected using the MDA were age, IDL_PL_PCT, creatinine, marital status, and L_HDL_CE_PCT. The top five variables selected using the MDG were age, IDL_PL_PCT, GLYCA, creatinine, and acetate. Notably, age, IDL_PL_PCT, and creatinine were among the top five in both rankings, underscoring their crucial roles in AD onset. The variable importance rankings in the random forest prediction model are presented in Fig. 5 and Table 6.

Fig. 5
figure 5

Variable importance ranking in the random forest prediction model based on the Mean Decrease Accuracy and Mean Decrease Gini

Full size image
Table 6 Top 17 variables screened using the random forest model
Full size table

Discussion

Recently, machine learning techniques have been applied in various fields. Medically, researchers have utilized machine learning to analyze extensive healthcare data [34, 35] for more accurate diagnoses and disease predictions. The random forest algorithm performs well in AD [36, 37], effectively enhancing the precision of predictive models. The uniqueness of the random forest algorithm lies in constructing each decision tree through random sampling, allowing it to model based on different subsets of data and features [37]. This helps mitigate the risk of overfitting, improves the predictive accuracy of the model, and ensures good generalization performance. Additionally, the random forest model can indicate the relative importance of variables and enhance the interpretation of the results.

Zhang et al. [38] conducted a study investigating the association between serum total cholesterol (TC) levels and neuropsychological performance, as well as intrinsic functional networks in non-demented older adults. They utilized ANCOVA analysis, adjusting for age, gender, and education years to compare neuropsychological performance between the two groups. The study found that in nondemented older adults, higher serum cholesterol levels were associated with disrupted functional connectivity in the salience network (SN). Proitsi et al. [39] conducted a study involving nontargeted lipidomics analysis of plasma samples from 148 patients with AD and 152 elderly controls. They utilized both univariate and multivariate analysis methods and found that blood lipids hold promise as potential biomarkers for AD. This discovery may lead to the emergence of new therapeutic strategies. In a study conducted by Chung et al. [40], they evaluated the relationship between LDL cholesterol (LDL-C) and gray-matter volume (GMV) in a community-based population without stroke or dementia. Using multiple linear regression analysis, they found that low circulatory LDL-C levels, in combination with hypertension, appeared to have a combined detrimental effect on posterior cingulate GMV, white matter hyperintensities (WMH), and verbal memory.

This study initially used the random forest, Lasso regression, and CatBoost algorithms to rank and filter the importance of the 213 lipid and metabolite variables included in this study. While determining the final modeling variables, the results of variable selection from using these three machine learning methods were comprehensively considered, which could reduce the probability of occasional variable selection using a single method and ensure the reliability of the modeling results. Among the top 20 variables selected using each method, seven variables, including acetate, albumin, creatinine, glucose, GLYCA, IDL_PL_PCT, and L_HDL_PL_PCT, were co-screened using all three methods, indicating the significant role of these variables in predicting AD onset. Finally, by employing the random forest model and generating a visualization of variable importance rankings, we visualized the importance ranking of the variables contributing to AD onset.

The results of the AD risk prediction model in this study indicated that the importance rankings of the variables age, IDL_PL_PCT, and creatinine, after being selected using the MDA and MDG indices, all ranked within the top five. This suggests that these three variables play crucial roles in AD onset. AD is one of the most common age-related neurodegenerative diseases [41], and its incidence significantly increases with age [42]. Removing triglycerides from very low-density lipoprotein in muscles and adipose tissues can lead to the formation of cholesterol-rich IDL particles, and these IDL particles can promote atherosclerosis [43]. The central nervous system is rich in lipids, primarily located on biological membranes, maintaining the structure and function of the central nervous system [16]. Changes in the lipid composition of the brain and plasma have been widely observed in patients with AD [44]. Phospholipids are key components of the cell membrane and can lead to synaptic dysfunction in patients with AD [45]. Physical activity and exercise prevent or delay AD onset [46], with lipid levels in patients with AD who exercise regularly lower than in those without AD [47]. Creatinine concentrations in the cerebrospinal fluid of patients with AD were significantly higher than in those without AD, which may be related to the overuse of creatine phosphate [48]. This result is consistent with the conclusions of this study. Creatinine is a byproduct of phosphocreatine [49], which stores high-energy phosphate bonds and releases energy when the glucose supply is insufficient. These results suggest that a significant increase in creatinine concentration in the cerebrospinal fluid of patients with AD causes an imbalance in their energy metabolism.

In conclusion, age, IDL_PL_PCT, and creatinine were the key factors identified that influence AD onset. Clinical screening and regular monitoring of lipoproteins and their metabolites in older patients can provide new perspectives for early AD diagnosis and prevention.

Strengths and limitations

The strength of this study was that three machine learning methods were used for the preliminary screening of variables, which reduced the probability of chance in the screening of variables using a single method. Finally, the screening results of the three methods were combined for modeling, which improved the accuracy of the model to an extent.

However, this study had some limitations. First, this was a cross-sectional study; therefore, we only focused on the measurements of key lipoproteins and metabolites that affect AD pathogenesis at a certain point in time and did not consider the relationship between the longitudinal dynamic trajectories of these factors and AD pathogenesis. Second, no follow-up period was observed; therefore, investigating the impact of these key lipoproteins and metabolites on AD morbidity was impossible. Third, no standard range of health values for the studied variables was observed.

Conclusion

This study identified age, IDL_PL_PCT, and creatinine as key factors closely associated with the onset of AD. Age is a well-established risk factor for AD, and this study further substantiates its significance in AD development. The discovery of IDL_PL_PCT provides novel insights into the relationship between lipid metabolism abnormalities and AD, offering researchers and clinicians potential avenues for investigation. The elevated concentration of creatinine in patients with AD suggests a potential link to energy metabolism imbalance. It can serve as an adjunct diagnostic marker, enhancing our understanding of AD pathogenesis and presenting potential therapeutic targets. For older individuals, closer monitoring and assessment should be carried out to facilitate early detection and increase the likelihood of diagnosing AD. Therefore, lipid metabolism management should be emphasized in patient care, including diet control, moderate exercise, and potential pharmacological interventions to maintain healthy lipid status.

Availability of data and materials

The datasets generated and analyzed during the current study are available in the [ADNI] database.

Abbreviations

AD:

Alzheimer’s disease

Aβ:

β-Amyloid proteins 

APP:

Amyloid precursor protein

MAMs:

mitochondria-associated endoplasmic reticulum membranes

CM:

chylomicrons

VLDL:

very-low-density lipoproteins

IDL:

intermediate-density lipoproteins

LDL:

low-density lipoproteins

HDL:

high-density lipoproteins

ADNI:

Alzheimer’s Disease Neuroimaging Initiative

GMV:

gray-matter volume

IDL_PL_PCT:

Phospholipids to total lipids ratio in intermediate-density lipoproteins

MDA:

Mean Decrease Accuracy

MDG:

Mean Decrease Gini

GLYCA:

Glycoprotein acetylation

S_HDL_PL_PCT:

Phospholipids to total lipids ratio in small high-density lipoproteins

L_LDL_CE_PCT:

Percentage of cholesteryl esters within low-density lipoprotein

L_HDL_CE_PCT:

Cholesteryl esters to total lipids ratio in large high-density lipoproteins

References

  1. Soria Lopez JA, González HM, Léger GC. Alzheimer’s disease. Handb Clin Neurol. 2019;167:231–55.

    Article  PubMed  Google Scholar 

  2. Graff-Radford J, Yong KXX, Apostolova LG, Bouwman FH, Carrillo M, Dickerson BC, et al. New insights into atypical Alzheimer’s disease in the era of biomarkers. Lancet Neurol. 2021;20(3):222–34.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Tiwari S, Atluri V, Kaushik A, Yndart A, Nair M. Alzheimer’s disease: pathogenesis, diagnostics, and therapeutics. Int J Nanomedicine. 2019;14:5541–54.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Lei P, Ayton S, Bush AI. The essential elements of Alzheimer’s disease. J Biol Chem. 2021;296:100105.

    Article  CAS  PubMed  Google Scholar 

  5. Ferrari C, Sorbi S. The complexity of Alzheimer’s disease: an evolving puzzle. Physiol Rev. 2021;101(3):1047–81.

    Article  CAS  PubMed  Google Scholar 

  6. Breijyeh Z, Karaman R. Comprehensive review on Alzheimer’s disease: causes and treatment. Molecules. 2020;25(24):5789.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Khan S, Barve KH, Kumar MS. Recent advancements in pathogenesis, diagnostics and treatment of Alzheimer’s disease. Curr Neuropharmacol. 2020;18(11):1106–25.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Tatulian SA. Challenges and hopes for Alzheimer’s disease. Drug Discov Today. 2022;27(4):1027–43.

    Article  PubMed  Google Scholar 

  9. Naseri NN, Wang H, Guo J, Sharma M, Luo W. The complexity of tau in Alzheimer’s disease. Neurosci Lett. 2019;705:183–94.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Guo C, Jeong HH, Hsieh YC, Klein HU, Bennett DA, De Jager PL, et al. Tau activates transposable elements in Alzheimer’s disease. Cell Rep. 2018;23(10):2874–80.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Rosenberg PB, Lyketsos CG. New clues to preclinical Alzheimer’s disease. Am J Psychiatry. 2018;175(6):493–4.

    Article  PubMed  Google Scholar 

  12. Masoumi J, Abbasloui M, Parvan R, Mohammadnejad D, Pavon-Djavid G, Barzegari A, et al. Apelin, a promising target for Alzheimer disease prevention and treatment. Neuropeptides. 2018;70:76–86.

    Article  CAS  PubMed  Google Scholar 

  13. Huo Z, Yu L, Yang J, Zhu Y, Bennett DA, Zhao J. Brain and blood metabolome for Alzheimer’s dementia: findings from a targeted metabolomics analysis. Neurobiol Aging. 2020;86:123–33.

    Article  CAS  PubMed  Google Scholar 

  14. Maszka P, Kwasniak-Butowska M, Cysewski D, Slawek J, Smolenski RT, Tomczyk M. Metabolomic footprint of disrupted energetics and amino acid metabolism in neurodegenerative diseases: perspectives for early diagnosis and monitoring of therapy. Metabolites. 2023;13(3):369.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Grao-Cruces E, Claro-Cala CM, Montserrat-de la Paz S, Nobrega C. Lipoprotein metabolism, protein aggregation, and Alzheimer’s disease: a literature review. Int J Mol Sci. 2023;24(3):2944.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Falabella M, Vernon HJ, Hanna MG, Claypool SM, Pitceathly RDS. Cardiolipin, mitochondria, and neurological disease. Trends Endocrinol Metab. 2021;32(4):224–37.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Watanabe Y, Hirao Y, Kasuga K, Tokutake T, Kitamura K, Niida S, et al. Urinary apolipoprotein C3 is a potential biomarker for Alzheimer’s disease. Dement Geriatr Cogn Dis Extra. 2020;10(3):94–104.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Kao YC, Ho PC, Tu YK, Jou IM, Tsai KJ. Lipids and Alzheimer’s disease. Int J Mol Sci. 2020;21(4):1505.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Wood WG, Schroeder F, Igbavboa U, Avdulov NA, Chochina SV. Brain membrane cholesterol domains, aging and amyloid beta-peptides. Neurobiol Aging. 2002;23(5):685–94.

    Article  CAS  PubMed  Google Scholar 

  20. Makki BE, Rahman S. Alzheimer’s disease in diabetic patients: a lipidomic prospect. Neuroscience. 2023;530:79–94.

    Article  PubMed  Google Scholar 

  21. Pera M, Larrea D, Guardia-Laguarta C, Montesinos J, Velasco KR, Agrawal RR, et al. Increased localization of APP-C99 in mitochondria-associated ER membranes causes mitochondrial dysfunction in Alzheimer disease. EMBO J. 2017;36(22):3356–71.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Pera M, Montesinos J, Larrea D, Agrawal RR, Velasco KR, Stavrovskaya IG, et al. MAM and C99, key players in the pathogenesis of Alzheimer’s disease. Int Rev Neurobiol. 2020;154:235–78.

    Article  CAS  PubMed  Google Scholar 

  23. Zang F, Zhu Y, Liu X, Fan D, Wang Q, Zhang Q, et al. Polygenic effects of the lipid metabolic pathway accelerated pathological changes and disrupted default mode network trajectory across the Alzheimer’s disease spectrum. J Clin Psychiatry. 2021;82(6):20m13739.

    Article  PubMed  Google Scholar 

  24. Walter J, van Echten-Deckert G. Cross-talk of membrane lipids and Alzheimer-related proteins. Mol Neurodegener. 2013;8:34.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Rostagno AA. Pathogenesis of Alzheimer’s disease. Int J Mol Sci. 2022;24(1):107.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Sun BL, Li WW, Zhu C, Jin WS, Zeng F, Liu YH, et al. Clinical research on Alzheimer’s disease: progress and perspectives. Neurosci Bull. 2018;34(6):1111–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Song M, Jung H, Lee S, Kim D, Ahn M. Diagnostic classification and biomarker identification of Alzheimer’s disease with random forest algorithm. Brain Sci. 2021;11(4):453.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Chang CH, Lin CH, Lane HY. Machine learning and novel biomarkers for the diagnosis of Alzheimer’s disease. Int J Mol Sci. 2021;22(5):2761.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Speiser JL, Miller ME, Tooze J, Ip E. A comparison of random forest variable selection methods for classification prediction modeling. Expert Syst Appl. 2019;134:93–101.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Velazquez M, Lee Y. Random forest model for feature-based Alzheimer’s disease conversion prediction from early mild cognitive impairment subjects. PLoS ONE. 2021;16(4):e0244773.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Nakapraves S, Warzecha M, Mustoe CL, Srirambhatla V, Florence AJ. Prediction of mefenamic acid crystal shape by random forest classification. Pharm Res. 2022;39(12):3099–111.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Fan H, Xie Q, Zhang Z, Wang J, Chen X, Qiu P. Chronological age prediction: developmental evaluation of DNA methylation-based machine learning models. Front Bioeng Biotechnol. 2021;9:819991.

    Article  PubMed  Google Scholar 

  33. Pellegrino E, Jacques C, Beaufils N, Nanni I, Carlioz A, Metellus P, et al. Machine learning random forest for predicting oncosomatic variant NGS analysis. Sci Rep. 2021;11(1):21820.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. She J, Su D, Diao R, Wang L. A joint model of random forest and artificial neural network for the diagnosis of endometriosis. Front Genet. 2022;13:848116.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. Hason L, Krishnan S. Spontaneous speech feature analysis for Alzheimer’s disease screening using a random forest classifier. Front Digit Health. 2022;4:901419.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Moore PJ, Lyons TJ, Gallacher J. Random forest prediction of Alzheimer’s disease using pairwise selection from time series data. PLoS ONE. 2019;14(2):e0211558.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  37. Xu L, Liang G, Liao C, Chen GD, Chang CC. k-Skip-n-Gram-RF: a random forest based method for Alzheimer’s disease protein identification. Front Genet. 2019;10:33.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Zhang T, Li H, Zhang J, Li X, Qi D, Wang N, et al. Impacts of high serum total cholesterol level on brain functional connectivity in non-demented elderly. J Alzheimers Dis. 2016;50(2):455–63.

    Article  CAS  PubMed  Google Scholar 

  39. Proitsi P, Kim M, Whiley L, Simmons A, Sattlecker M, Velayudhan L, et al. Association of blood lipids with Alzheimer’s disease: a comprehensive lipidomics analysis. Alzheimers Dement. 2017;13(2):140–51.

    Article  PubMed  Google Scholar 

  40. Chung CP, Chou KH, Peng LN, Liu LK, Lee WJ, Chen LK, et al. Associations between low circulatory low-density lipoprotein cholesterol level and brain health in non-stroke non-demented subjects. Neuroimage. 2018;181:627–34.

    Article  CAS  PubMed  Google Scholar 

  41. Kosyreva AM, Sentyabreva AV, Tsvetkov IS, Makarova OV. Alzheimer’s disease and inflammaging. Brain Sci. 2022;12(9):1237.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Vasic V, Barth K, Schmidt MHH. Neurodegeneration and neuro-regeneration-Alzheimer’s disease and stem cell therapy. Int J Mol Sci. 2019;20(17):4272.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Feingold KR. Introduction to lipids and lipoproteins. South Dartmouth: MDText.com, Inc.; 2000.

    Google Scholar 

  44. Naudí A, Cabré R, Jové M, Ayala V, Gonzalo H, Portero-Otín M, et al. Lipidomics of human brain aging and Alzheimer’s disease pathology. Int Rev Neurobiol. 2015;122:133–89.

    Article  PubMed  Google Scholar 

  45. Bennett SA, Valenzuela N, Xu H, Franko B, Fai S, Figeys D. Using neurolipidomics to identify phospholipid mediators of synaptic (dys)function in Alzheimer’s Disease. Front Physiol. 2013;4:168.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Gaitán JM, Moon HY, Stremlau M, Dubal DB, Cook DB, Okonkwo OC, et al. Effects of aerobic exercise training on systemic biomarkers and cognition in late middle-aged adults at risk for Alzheimer’s disease. Front Endocrinol (Lausanne). 2021;12:660181.

    Article  PubMed  Google Scholar 

  47. Sato Y, Nakamura T, Aoshima K, Oda Y. Quantitative and wide-ranging profiling of phospholipids in human plasma by two-dimensional liquid chromatography/mass spectrometry. Anal Chem. 2010;82(23):9858–64.

    Article  CAS  PubMed  Google Scholar 

  48. van der Velpen V, Teav T, Gallart-Ayala H, Mehl F, Konz I, Clark C, et al. Systemic and central nervous system metabolic alterations in Alzheimer’s disease. Alzheimers Res Ther. 2019;11(1):93.

    Article  PubMed  PubMed Central  Google Scholar 

  49. Wyss M, Kaddurah-Daouk R. Creatine and creatinine metabolism. Physiol Rev. 2000;80(3):1107–213.

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We thank all the participants and data managers of the ADNI database.

Funding

This study was partially supported by the National Natural Science Foundation of China (No. 81803337, 81872719, and 82003560), Shandong Provincial Youth Innovation Team Development Plan of Colleges and Universities (No. 2019-6-156, Lu-Jiao), and Shandong Provincial Natural Science Foundation (No. ZR2023MH313).

Author information

Author notes

  1. Fenglin Wang, Aimin Wang, Yiming Huang and Wenfeng Gao contributed equally to this study and are co-first authors.

  2. Suzhen Wang and Fuyan Shi are co-corresponding authors and are co-senior authors.

Authors and Affiliations

  1. Department of Health Statistics, School of Public Health, Shandong Second Medical University, Weifang, Shandong, 261053, China

    Fenglin Wang, Aimin Wang, Yiming Huang, Yaqi Xu, Wenjing Zhang, Guiya Guo, Wangchen Song, Yujia Kong, Qinghua Wang, Suzhen Wang & Fuyan Shi

  2. Department of Rheumatology and Immunology, Affiliated Hospital of Shandong Second Medical University, Weifang, Shandong, 261031, China

    Wenfeng Gao

Authors
  1. Fenglin Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Aimin Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Yiming Huang
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Wenfeng Gao
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Yaqi Xu
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Wenjing Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Guiya Guo
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Wangchen Song
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Yujia Kong
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Qinghua Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Suzhen Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Fuyan Shi
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

Research idea and study design: SW and FS. Data acquisition: YX, WZ, GG, and WS. Data analysis and interpretation: FW, AW, YH, and WG. Manuscript writing: FW, AW, YH, and WG. Supervision or mentorship: YK, QW, SW, and FS.

Corresponding authors

Correspondence to Suzhen Wang or Fuyan Shi.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Wang, F., Wang, A., Huang, Y. et al. Lipoproteins and metabolites in diagnosing and predicting Alzheimer’s disease using machine learning. Lipids Health Dis 23, 152 (2024). https://doi.org/10.1186/s12944-024-02141-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12944-024-02141-w

Keywords

Lipids in Health and Disease

ISSN: 1476-511X