Deep learning analysis of UPLC-MS/MS-based metabolomics data to predict Alzheimer's disease

Mild cognitive impairment (MCI) is a clinical condition characterized by a decline in cognitive ability and progression of cognitive impairment. It is often considered a transitional stage between normal aging and Alzheimer's disease (AD). This study aimed to compare deep learning (DL) and traditional machine learning (ML) methods in predicting MCI using plasma proteomic biomarkers. A total of 239 adults were selected from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort along with a pool of 146 plasma proteomic biomarkers. We evaluated seven traditional ML models (support vector machines (SVMs), logistic regression (LR), naïve Bayes (NB), random forest (RF), k-nearest neighbor (KNN), gradient boosting machine (GBM), and extreme gradient boosting (XGBoost)) and six variations of a deep neural network (DNN) model-the DL model in the H2O package. Least Absolute Shrinkage and Selection Operator (LASSO) selected 35 proteomic biomarkers from the pool. Based on grid search, the DNN model with an activation function of "Rectifier With Dropout" with 2 layers and 32 of 35 selected proteomic biomarkers revealed the best model with the highest accuracy of 0.995 and an F1 Score of 0.996, while among seven traditional ML methods, XGBoost was the best with an accuracy of 0.986 and an F1 Score of 0.985. Several biomarkers were correlated with the APOE-ε4 genotype, polygenic hazard score (PHS), and three clinical cerebrospinal fluid biomarkers (Aβ42, tTau, and pTau). Bioinformatics analysis using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) revealed several molecular functions and pathways associated with the selected biomarkers, including cytokine-cytokine receptor interaction, cholesterol metabolism, and regulation of lipid localization. The results showed that the DL model may represent a promising tool in the prediction of MCI. These plasma proteomic biomarkers may help with early diagnosis, prognostic risk stratification, and early treatment interventions for individuals at risk for MCI.

Cross-sectional studies suggest self-reported cancer history is associated with decreased risk of Alzheimer's disease (AD). However, little is known about how self-reported cancer affects longitudinal AD progression, the primary outcome in clinical trials and observational studies. To determine self-reported cancer history's effect on longitudinal AD progression in an observational study. We utilized data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) to evaluate progression to AD by self-reported all-cancer, breast, prostate, colorectal, or non-melanoma skin cancer history. Linear mixed effects models were used to examine baseline differences and rates of progression on the Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) by self-reported cancer history. Age at AD onset was examined using consensus clinical diagnoses with Cox proportional hazards regression. Among 1,271 participants, models revealed no significant differences in progression over time but did reveal significantly lower baseline ADAS-Cog score, indicating better cognition at a given age in those with self-reported cancer history. Cox models indicated those with self-reported cancer history had significantly later age of AD onset (HR: 0.67, 95% CI: 0.53-0.85) after adjustment for covariates. Participants with self-reported cancer history entered ADNI with better cognition and later age of AD onset, but progressed similarly to participants without such history, indicating differences in AD between those with and without self-reported cancer history emerge early in the disease course. Such differences in longitudinal progression by self-reported cancer history could affect AD trials and observational studies, given the current focus on early disease course. Further investigation is warranted with detailed longitudinal assessment of cancer and AD.

IntroductionBrain glucose hypometabolism precedes cognitive decline in Alzheimer's disease, however its role in determining long-term cognitive trajectories remains under studied in current literature in a clear manner. We investigated whether baseline brain glucose metabolism predicts cognitive decline rates and disease conversion risk in a large longitudinal cohort.MethodsWe analyzed 4,732 participants (1,685 with longitudinal cognitive data) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) with fluorodeoxyglucose positron emission tomography (FDG-PET), Mini-Mental State Examination (MMSE), and Alzheimer's Disease Assessment Scale (ADAS) measurements over ten years. Mixed-effects models investigated time × brain glucose metabolism interactions on cognitive decline. Cross-validation assessed predictive accuracy for mild cognitive impairment and Alzheimer's disease conversion.ResultsBrain glucose metabolism was found to modulate the cognitive decline rates (time × FDG interaction: MMSE β = 0.746, P-value <0.001; ADAS β = −1.595, P-value <0.001). High versus low glucose metabolism demonstrated 1.49 MMSE points/year and 3.19 ADAS points/year protection. Among cognitively normal participants, low glucose metabolism increased Alzheimer's disease conversion risk four-fold (incidence rate ratio = 3.79, 95%CI: 2.94–4.88). Predictive models achieved high accuracy for Alzheimer's disease conversion (AUC = 0.826) with good calibration (Brier score = 0.092). High-metabolism individuals showed essentially stable cognition over ten years follow-up duration as evident in the ADNI dataset.ConclusionsWe found that brain glucose metabolism is a notable determinant of cognitive decline progression, especially for Alzheimer’s disease risk, providing quantifiable metabolic protection against decline. Our results demonstrate brain glucose hypometabolic findings from baseline FDG-PET as a precision biomarker for therapeutic stratification and support further interventions for cognitive preservation for further research, validation and development purposes that should be further evaluated and investigated in further clinical trials that may lead to better preventive and therapeutic benefits for cognitive functions preservation and prevention of cognitive decline process in high-risk individuals.

This study modeled predementia Alzheimer's disease clinical trials. Longitudinal data from cognitively normal (CN) and mild cognitive impairment (MCI) participants in the Alzheimer's Disease Neuroimaging Initiative were used to calculate sample size requirements for trials using outcome measures, including the Clinical Dementia Rating scale sum of boxes, Mini-Mental State Examination, Alzheimer's Disease Assessment Scale-cognitive subscale with and without delayed recall, and the Rey Auditory Verbal Learning Task. We examined the impact on sample sizes of enrichment for genetic and biomarker criteria, including cerebrospinal fluid protein and neuroimaging analyses. We observed little cognitive decline in the CN population at 36 months, regardless of the enrichment strategy. Nonetheless, in CN subjects, using Rey Auditory Verbal Learning Task total as an outcome at 36 months required the fewest subjects across enrichment strategies, with apolipoprotein E genotype ε4 carrier status requiring the fewest (n = 499 per arm to demonstrate a 25% reduction in disease progression). In MCI, enrichment reduced the required sample sizes for trials, relative to estimates based on all subjects. For MCI, the Clinical Dementia Rating scale sum of boxes consistently required the smallest sample sizes. We conclude that predementia clinical trial conduct in Alzheimer's disease is enhanced by the use of biomarker inclusion criteria.

PURPOSEChronic neuro‐inflammation is an important mechanism underlying normal aging and Alzheimer's disease (AD) that is thought to potentiate medial temporal lobe (MTL) atrophy and memory decline. Currently, there are not enough prognostic biomarkers that can track the development and progression of normal aging versus AD, as well as differing effects on the brain. Our study examined which pro‐ or anti‐inflammatory cerebrospinal fluid (CSF) biomarkers best predicted AD neuropathology over 24 months.METHODSBaseline mass spectrometry data from two hundred and eighty‐five adults was obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Our focus was on examining biomarkers directly associated with inflammatory processes. For our study, we utilized T1‐weighted medial temporal gray matter volumes at baseline and months 6, 12, and 24 after baseline. Statistical analyses were conducted using SPSS 23.0 software. We used linear mixed models as well as least significant differences (LSD) post‐hoc tests to analyze how variables differed by baseline diagnosis (normal aging, pre‐AD, AD) or pre‐AD patients that did or did not convert to AD. We used stepwise regression analyses to analyze which inflammatory biomarkers best predicted atrophy and memory decline over 24 months. We specifically focused on CSF levels of Neuronal Pentraxin 2 (NPTX2) and Chitinase‐3‐like‐protein‐1 (C3LP1), as they are the inflammatory biomarkers that significantly predicted MTL atrophy and memory decline. NPTX2 is a biomarker of synaptic plasticity. C3LP1 is a biomarker of microglial activation and chronic neuro‐inflammation. We also analyzed relative risk ratios for diagnosis with NPTX2 and C3LP1 with multinomial logistic regression analyses. Preliminary results for memory indicate that, regardless of baseline diagnosis, higher NPTX2 levels strongly predict higher memory scores by month 24 (Figure 1).RESULTSNPTX2 and C3LP1, respectively beneficial and adverse biomarkers of synaptic plasticity and microglial activation, loaded significantly. By month 24, higher baseline NPTX2 corresponded to less MTL atrophy [R2= .287] and less memory decline [R2= .560]. Conversely, higher C3LP1 modestly predicted more MTL atrophy [R2=.083].CONCLUSIONSOur study suggests that NPTX2 is a promising biomarker for predicting longitudinal changes in medial temporal volume, especially memory performance across the AD spectrum. Baseline NPTX2 performs better than C3LP1 in predicting MTL atrophy and memory decline over time. C3LP1 was a comparatively modest predictor of medial temporal atrophy and memory decline. For future directions, studies should investigate if clinical trials using anti‐inflammatory drugs/nutritional therapy increase or inhibit NPTX2 and/or C3LP1.Support or Funding InformationThis study was funded by Iowa State University and NIH K99 AG047282. Data collection and sharing for this project were funded by the ADNI (National Institutes of Health Grant U01‐AG‐024904) and Department of Defense ADNI (award number W81XWH‐12‐2‐0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the Alzheimer's Association and the Alzheimer's Drug Discovery Foundation. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private‐sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer's Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. The data used in the preparation of this article were obtained from the ADNI database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report.

Although cerebrospinal fluid (CSF) amyloid-β42 peptide (Aβ42) and phosphorylated tau (p-tau) and blood p-tau are valuable for differential diagnosis of Alzheimer's disease (AD) from cognitively normal (CN) there is a lack of validated biomarkers for mild cognitive impairment (MCI). This study sought to determine how plasma and CSF protein markers compared in the characterization of MCI and AD status. This cohort study included Alzheimer's Disease Neuroimaging Initiative (ADNI) participants who had baseline levels of 75 proteins measured commonly in plasma and CSF (257 total, 46 CN, 143 MCI, and 68 AD). Logistic regression, least absolute shrinkage and selection operator (LASSO) and Random Forest (RF) methods were used to identify the protein candidates for the disease classification. We observed that six plasma proteins panel (APOE, AMBP, C3, IL16, IGFBP2, APOD) outperformed the seven CSF proteins panel (VEGFA, HGF, PRL, FABP3, FGF4, CD40, RETN) as well as AD markers (CSF p-tau and Aβ42) to distinguish the MCI from AD [area under the curve (AUC) = 0.75 (plasma proteins), AUC = 0.60 (CSF proteins) and AUC = 0.56 (CSF p-tau and Aβ42)]. Also, these six plasma proteins performed better than the CSF proteins and were in line with CSF p-tau and Aβ42 in differentiating CN versus MCI subjects [AUC = 0.89 (plasma proteins), AUC = 0.85 (CSF proteins) and AUC = 0.89 (CSF p-tau and Aβ42)]. These results were adjusted for age, sex, education, and APOEϵ4 genotype. This study suggests that the combination of 6 plasma proteins can serve as an effective marker for differentiating MCI from AD and CN.

The Alzheimer’s disease Neuroimaging Initiative (ADNI) Private Partners Scientific Board (PPSB) encompasses members from industry, biotechnology, diagnostic, and non-profit organizations that have until recently been managed by the Foundation for the National Institute of Health (FNIH) and provided financial and scientific support to ADNI programs. In this article, we review some of the major activities undertaken by the PPSB, focusing on those supporting the most recently completed National Institute on Aging grant, ADNI3, and the impact it has had on streamlining biomarker discovery and validation in Alzheimer’s disease. We also provide a perspective on the gaps that may be filled with future PPSB activities as part of ADNI4 and beyond.

Develop a cerebrospinal fluid biomarker signature for mild Alzheimer's disease (AD) in Alzheimer's Disease Neuroimaging Initiative (ADNI) subjects. Amyloid-beta 1 to 42 peptide (A beta(1-42)), total tau (t-tau), and tau phosphorylated at the threonine 181 were measured in (1) cerebrospinal fluid (CSF) samples obtained during baseline evaluation of 100 mild AD, 196 mild cognitive impairment, and 114 elderly cognitively normal (NC) subjects in ADNI; and (2) independent 56 autopsy-confirmed AD cases and 52 age-matched elderly NCs using a multiplex immunoassay. Detection of an AD CSF profile for t-tau and A beta(1-42) in ADNI subjects was achieved using receiver operating characteristic cut points and logistic regression models derived from the autopsy-confirmed CSF data. CSF A beta(1-42) was the most sensitive biomarker for AD in the autopsy cohort of CSF samples: receiver operating characteristic area under the curve of 0.913 and sensitivity for AD detection of 96.4%. In the ADNI cohort, a logistic regression model for A beta(1-42), t-tau, and APO epsilon 4 allele count provided the best assessment delineation of mild AD. An AD-like baseline CSF profile for t-tau/A beta(1-42) was detected in 33 of 37 ADNI mild cognitive impairment subjects who converted to probable AD during the first year of the study. The CSF biomarker signature of AD defined by A beta(1-42) and t-tau in the autopsy-confirmed AD cohort and confirmed in the cohort followed in ADNI for 12 months detects mild AD in a large, multisite, prospective clinical investigation, and this signature appears to predict conversion from mild cognitive impairment to AD.

When assessments are primarily used to measure change over time, it is important to evaluate items according to their sensitivity to change, specifically. Items that demonstrate good sensitivity to between-person differences at baseline may not show good sensitivity to change over time, and vice versa. In this study, we applied a longitudinal factor model of change to a widely used cognitive test designed to assess global cognitive status in dementia, and contrasted the relative sensitivity of items to change. Statistically nested models were estimated introducing distinct latent factors related to initial status differences between test-takers and within-person latent change across successive time points of measurement. Models were estimated using all available longitudinal item-level data from the Alzheimer's Disease Assessment Scale-Cognitive subscale, including participants representing the full-spectrum of disease status who were enrolled in the multisite Alzheimer's Disease Neuroimaging Initiative. Five of the 13 Alzheimer's Disease Assessment Scale-Cognitive items demonstrated noticeably higher loadings with respect to sensitivity to change. Attending to performance change on only these 5 items yielded a clearer picture of cognitive decline more consistent with theoretical expectations in comparison to the full 13-item scale. Items that show good psychometric properties in cross-sectional studies are not necessarily the best items at measuring change over time, such as cognitive decline. Applications of the methodological approach described and illustrated in this study can advance our understanding regarding the types of items that best detect fine-grained early pathological changes in cognition. (PsycINFO Database Record

Missing data is prevalent in the Alzheimer's Disease Neuroimaging Initiative (ADNI). It is common to deal with missingness by removing subjects with missing entries prior to statistical analysis; however, this can lead to significant efficiency loss and sometimes bias. It has yet to be demonstrated that the imputation approach to handling this issue can be valuable in some longitudinal regression settings. The purpose of this study is to demonstrate the importance of imputation and how imputation is correctly done in ADNI by analyzing longitudinal Alzheimer's Disease Assessment Scale -Cognitive Subscale 13 (ADAS-Cog 13) scores and their association with baseline patient characteristics. We studied 1,063 subjects in ADNI with mild cognitive impairment. Longitudinal ADAS-Cog 13 scores were modeled with a linear mixed-effects model with baseline clinical and demographic characteristics as predictors. The model estimates obtained without imputation were compared with those obtained after imputation with Multiple Imputation by Chained Equations (MICE). We justify application of MICE by investigating the missing data mechanism and model assumptions. We also assess robustness of the results to the choice of imputation method. The fixed-effects estimates of the linear mixed-effects model after imputation with MICE yield valid, tighter confidence intervals, thus improving the efficiency of the analysis when compared to the analysis done without imputation. Our study demonstrates the importance of accounting for missing data in ADNI. When deciding to perform imputation, care should be taken in choosing the approach, as an invalid one can compromise the statistical analyses.

The Alzheimer's Disease Neuroimaging Initiative (ADNI) is an ongoing, longitudinal, multicenter study designed to develop clinical, imaging, genetic, and biochemical biomarkers for the early detection and tracking of Alzheimer's disease (AD). The study aimed to enroll 400 subjects with early mild cognitive impairment (MCI), 200 subjects with early AD, and 200 normal control subjects; $67 million funding was provided by both the public and private sectors, including the National Institute on Aging, 13 pharmaceutical companies, and 2 foundations that provided support through the Foundation for the National Institutes of Health. This article reviews all papers published since the inception of the initiative and summarizes the results as of February 2011. The major accomplishments of ADNI have been as follows: (1) the development of standardized methods for clinical tests, magnetic resonance imaging (MRI), positron emission tomography (PET), and cerebrospinal fluid (CSF) biomarkers in a multicenter setting; (2) elucidation of the patterns and rates of change of imaging and CSF biomarker measurements in control subjects, MCI patients, and AD patients. CSF biomarkers are consistent with disease trajectories predicted by β-amyloid cascade (Hardy, J Alzheimers Dis 2006;9(Suppl 3):151-3) and tau-mediated neurodegeneration hypotheses for AD, whereas brain atrophy and hypometabolism levels show predicted patterns but exhibit differing rates of change depending on region and disease severity; (3) the assessment of alternative methods of diagnostic categorization. Currently, the best classifiers combine optimum features from multiple modalities, including MRI, [(18)F]-fluorodeoxyglucose-PET, CSF biomarkers, and clinical tests; (4) the development of methods for the early detection of AD. CSF biomarkers, β-amyloid 42 and tau, as well as amyloid PET may reflect the earliest steps in AD pathology in mildly symptomatic or even nonsymptomatic subjects, and are leading candidates for the detection of AD in its preclinical stages; (5) the improvement of clinical trial efficiency through the identification of subjects most likely to undergo imminent future clinical decline and the use of more sensitive outcome measures to reduce sample sizes. Baseline cognitive and/or MRI measures generally predicted future decline better than other modalities, whereas MRI measures of change were shown to be the most efficient outcome measures; (6) the confirmation of the AD risk loci CLU, CR1, and PICALM and the identification of novel candidate risk loci; (7) worldwide impact through the establishment of ADNI-like programs in Europe, Asia, and Australia; (8) understanding the biology and pathobiology of normal aging, MCI, and AD through integration of ADNI biomarker data with clinical data from ADNI to stimulate research that will resolve controversies about competing hypotheses on the etiopathogenesis of AD, thereby advancing efforts to find disease-modifying drugs for AD; and (9) the establishment of infrastructure to allow sharing of all raw and processed data without embargo to interested scientific investigators throughout the world. The ADNI study was extended by a 2-year Grand Opportunities grant in 2009 and a renewal of ADNI (ADNI-2) in October 2010 through to 2016, with enrollment of an additional 550 participants.

Alzheimer's Disease Neuroimaging Initiative biomarkers as quantitative phenotypes: Genetics core aims, progress, and plans

INTRODUCTIONBiomarkers for Alzheimer's disease neuropathologic change (ADNC) have been instrumental in developing effective disease‐modifying therapeutics. However, to prevent/treat dementia effectively, we require biomarkers for non‐AD neuropathologies; for this, neuropathologic examinations and annotated tissue samples are essential.METHODSWe conducted clinicopathologic correlation for the first 100 Alzheimer's Disease Neuroimaging Initiative (ADNI) Neuropathology Core (NPC) cases.RESULTSClinical syndromes in this cohort showed 95% sensitivity and 79% specificity for predicting high/intermediate ADNC, a 21% false positive rate, and a ∼44% false negative rate. In addition, 60% with high/intermediate ADNC harbored additional potentially dementing co‐pathologies.DISCUSSIONThese results suggest that clinical presentation imperfectly predicts ADNC and that accurate prediction of high/intermediate ADNC does not exclude co‐pathology that may modify presentation, biomarkers, and therapeutic responses. Therefore, new biomarkers are needed for non‐AD neuropathologies. The ADNI NPC supports this mission with well‐characterized tissue samples (available through ADNI and the National Institute on Aging) and “gold‐standard” diagnostic information (soon to include digital histology).HighlightsThe Alzheimer's Disease Neuroimaging Initiative (ADNI) Neuropathology Core (NPC) brain donation cohort now exceeds 200 cases.ADNI NPC data in National Alzheimer's Coordinating Center format are available through the Laboratory of Neuro Imaging.Digitized slide files from the ADNI NPC will be available in 2025.Requests for ADNI brain tissue samples can be submitted online for ADNI/National Institute on Aging evaluation.Clinical diagnoses of Alzheimer's disease (AD)/AD and related dementias (ADRD) do not always predict post mortem neuropathology.Neuropathology is essential for the development of novel AD/ADRD biomarkers.

To investigate the longitudinal impact of type 2 diabetes mellitus (T2DM) on the prodromal and dementia stages of Alzheimer's disease (AD), focusing on diabetes duration and other comorbidities. 1395 dementia-free individuals aged 55-90 years with a maximum 15-year follow-up data were enrolled from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Cox proportional hazards regression models were used to estimate hazard ratios (HRs) of the incidence of prodromal or dementia stages of AD. Longer T2DM duration (≥5 years) (multi-adjusted HR=2.19, 95% CI=1.05-4.58), but not shorter T2DM duration (<5 years), was associated with a significantly increased risk of incident prodromal AD over a mean follow-up of 4.8 years. APOE ε4 alleles (HR=3.32, 95% CI=1.41-7.79) and comorbid coronary artery disease (CAD) (HR=3.20, 95% CI=1.29-7.95) further increased the risk of incident prodromal AD in patients with T2DM. No significant association was observed between T2DM and conversational risk of progression from prodromal AD to AD dementia. T2DM, which is characterized by a longer duration, increases the incidence risk of prodromal AD but not AD dementia. APOE ε4 allele and comorbid CAD strengthen the relationship between T2DM and prodromal AD. These findings highlight T2DM characteristics and its comorbidities as predictors for accurate prediction of AD and screening of at-risk populations.

While the greatest risk factor for Alzheimer's disease (AD) is aging, women are disproportionately affected by the disease. Interestingly, the hippocampus and cerebellum exhibit gender-specific cytoarchitecture differences, which are associated with AD, despite the absence of a role in animal reproductive behavior or hormonal signaling. This study investigates the potential association of sex differences associated with AD by interrogating cerebellar and hippocampal volume in preclinical (MCI) as well as clinical phases of AD compared to cognitively normal patients (CN) and in an animal model of AD, the streptozotocin (STZ)-induced sporadic AD model. In order to investigate putative changes in cerebellum and hippocampus in a rat model of AD, we used a STZ-induced sporadic AD model at three different time points (2, 4, and 8 weeks) after surgery in male and female rats. Previous studies have reported hippocampal-dependent changes as well as sex-dependent behavioral and signaling effects in the STZ animal model of sporadic AD while our current study showed involvement of cerebellum-mediated changes. To interrogate the role of cerebellar volume in AD progression within the human context, we analyzed data available through the Alzheimer's Disease Neuroimaging Initiative (ADNI). In a cross-sectional analysis, we observed that levels of peripheral Glial Acidic Fibrillary Protein (GFAP) (astrocytic protein) were associated negatively with cerebellar and hippocampal volumes (β = -0.002, p-value = 0.04; β = -6.721, p-value < 0.0001) and were associated with sex specific differences in males. Our analysis identified that the effect on hippocampal volume was earlier in disease stage, reinforcing the relevance of longitudinal alterations of cerebellum and hippocampus volume over time. The STZ animal model of sporadic AD, corroborated the progressive changes in hippocampal volume and more minor and temporally delayed involvement of the cerebellum volume changes which were dependent on sex. This suggests that cerebellar involvement may be secondary to hippocampal neurodegeneration, and both regional differences were dependent on sex. Due to the association with GFAP, our findings may be due to network astrocyte connection spread regardless of primary pathology. Overall, our study uncovers a novel role for cerebellum in AD in a model and in the human context.