2nd Workshop on Research Definitions for Reserve and Resilience in Cognitive Aging & Dementia
September 14-15, 2020
Online
PROGRAM
TUESDAY, September 15, 2020
2nd Workshop Pilot Project Presentations and Analytical Plans
Below are the 12 awarded pilot project introductions and analytical plans (available from the toggle bars below the video files).
ANALYTICAL PLAN - Melis Anaturk
- Study Title
Modelling brain and cognitive age to study cognitive reserve and resilience
- Study Investigators
First Name | Last Name | Organization | |
PI | James | Cole | University College London |
Co-PI | Melis | Anatürk | University College London |
Co-PI | Ann-Marie | de Lange | University of Oxford |
Co-PI | Klaus | Ebmeier | University of Oxford |
- Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
The proposed project will involve a secondary analysis of three pre-collected datasets: the UK Biobank study (mean age = 64 years; n = 42,067; 2 MRI assessments for approximately 1,500 individuals), Lothian Birth Cohort 1936 (mean age = 73 years, n = 669; 4 MRI assessments), and Insight 46 (mean age = 71 years, n = 500; 2 MRI assessments). Data on structural, diffusion-weighted and fluid-attenuated inversion recovery (FLAIR) MRI, cognitive function and demographic/lifestyle variables (e.g., physical activity levels, smoking status and alcohol intake) will be examined in all three cohorts. We plan to utilize both cross-sectional and longitudinal data from these cohorts. Heritability will be examined based on genome-wide association studies (GWAS), using data from UK Biobank.
- Key variables analyzed/measured in the study
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
The key outcome variables of this paper are level and changes in the brain age gap (BAG), cognitive age gap (CAG) and CAG (residualized for BAG). SNP-based heritability of BAG and CAG will also be examined as an outcome. BAG will be computed based on the difference between brain age and chronological age, with brain age estimated using image-derived phenotypes. CAG will be calculated using the difference between cognitive age and chronological age, with cognitive age estimated using latent factors derived from neuropsychological data. The predictors of interest are education, premorbid IQ and a lifestyle score (composite score integrating physical activity, smoking status and alcohol intake). Co-variates will consist of baseline measures of chronological age, sex, ethnicity, body mass index, scanner site and scanner motion (for brain age estimates), with exploratory analyses introducing further co-variate adjustments. Further, for the analysis of CAG, BAG will also be included as an additional co-variate.
- Operational definitions of concepts
Brain maintenance will be operationalized as BAG, a neuroimaging-derived metric from machine learning models of healthy brain aging. BAG estimated from T1-weighted, diffusion-weighted and FLAIR-weighted MRI reflects the difference between estimated brain age and chronological age and is sensitive to both the preservation and decline in brain structure. Negative values reflect preserved brain structure relative to what is expected based on healthy trajectories. Following this rationale, we aim to operationalize cognitive maintenance with a novel metric, CAG. CAG reflects the difference between cognitive age and chronological age, with negative values indexing preserved cognition as compared to what is expected based on normative trajectories. A discrepancy between BAG and CAG, for instance a younger cognitive age relative to brain age, would indicate preserved cognitive function in the presence of aging-related brain decline, i.e., cognitive reserve. Hence, cognitive reserve will be captured as the variance in CAG not explained by BAG.
In summary, brain maintenance will be measured using BAG, cognitive maintenance will be evaluated using CAG and cognitive reserve will be captured as CAG (residualized for BAG).
- Specific Aims and Hypotheses
The primary aim of this study is to use machine learning to derive estimates of brain age and cognitive age using MRI and cognitive data. The secondary aim is to compute the brain age gap (BAG) and cognitive age gap (CAG) and evaluate these metrics for their heritability and genomic correlations. Baseline and longitudinal associations will also be examined. The tertiary aim is to examine whether education, premorbid IQ and a composite lifestyle score, relate to BAG, CAG and CAG (residualized for BAG).
We hypothesize that: (i) BAG and CAG are partially heritable traits and will correlate genomically, suggesting genetic overlap (ii) level and changes in BAG and CAG will be associated (iii) education and premorbid IQ will associate with BAG, CAG and CAG (residualized for BAG) and (iv) a healthier lifestyle will associate with BAG, CAG and CAG (residualized for BAG) independently.
- Statistical plan
The XGBoost regressor model (https://xgboost.readthedocs.io) will be used to run predictions based on MRI and cognitive data, with brain age prediction based on image-derived phenotypes. Sixty percent of the UK Biobank sample will be used as the training set. To ensure that brain age and cognitive age are estimated based on healthy aging patterns, the training set will be restricted to individuals without a diagnosis of any neurological or psychiatric diseases, such as Alzheimer’s disease or depression. The rest of UK Biobank sample will be used as a validation (20%) set and test set (20%). The model based on the healthy training set will be applied to the other cohorts, allowing us to estimate brain and cognitive age and compute BAG and CAG. Prior to brain age estimation, the MRI images from all cohort studies will be pre-processed using established pipelines in FSL and FreeSurfer and imaging data will be harmonized using tools such as ComBAT. As there are inter-study differences in the cognitive batteries administered, latent factors derived from the cognitive data will be used for harmonisation purposes.
In the UK Biobank sample, genome-wide complex trait analysis will be employed to assess single nucleotide polymorphism (SNP) heritability, and genomic restricted maximum likelihood will be used to compute genetic correlations between BAG and CAG. For the subsequent analyses conducted across all cohorts, latent change models will be used to evaluate the associations between baseline levels and changes in BAG and CAG. Further, education, premorbid IQ and lifestyle scores (a composite measure of physical activity, smoking status and alcohol intake) will be regressed onto baseline level and changes in BAG and CAG, to evaluate how these factors associate with our outcomes of interest. Additional co-variates, including chronological age, sex, ethnicity, body mass index, scanner site and scanner motion will be added to these models, with BAG introduced as an additional co-variate in the analysis of CAG.
FDR corrections will be applied to control the Type I error rate, and robust permutation and cross-validation procedures will be used to ensure reproducibility and generalisability of our findings. The main analyses will be restricted to individuals with complete data.
- Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Rory Boyle
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Rory | Boyle | Trinity College Institute of Neuroscience, Trinity College Dublin |
Co-PI | Robert | Whelan | Trinity College Institute of Neuroscience, Trinity College Dublin |
3. Brief explanation of existing data or new data collection
In both research aims, human data will be used. Cross-sectional neuroimaging and clinical data, along with longitudinal cognitive data, from the Irish Longitudinal Study on Ageing (TILDA), will be used. TILDA is a randomly sampled, nationally representative sample. To-date, TILDA data has been collected for five time points/waves. For the purposes of this project, Wave 3 (only time point to-date with neuroimaging data) will be considered the baseline time point, and Wave 5 will be considered the follow-up time point.
In Aim 1, after listwise deletion, the analyses will have an N of 351 participants with a mean age of 67.9 years (SD = 7.3, range = 49 – 87) and with 51% females. In Aim 2, after listwise deletion, the analyses will have an N of 309 participants with a mean age of 67.5 years (SD = 7.3, range = 49 – 87) and with 50% females.
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
In both research aims, there are two separate predictor variables: cognitive reserve (CR) and brain maintenance (BM). CR will be measured using CR network strength which will be calculated from the application of a robust machine learning model, connectome-based predictive modelling, to functional MRI data. BM will be measured using the brain-predicted age difference, which is calculated from the application of robust machine learning techniques to structural MRI data. To minimise repetition, the specific detail describing these measure are outlined in Section 6 Operational definitions of concepts.
Three separate cognitive outcome variables measured at follow-up will be used in both research aims, with each variable reflecting different domains of cognitive function. These will include global cognition (MMSE), executive function/verbal fluency (Animal Naming Test), and episodic memory (composite measure of immediate and delayed recall of 10-word list).
In Aim 1, covariates will be comprised of the following variables measured at baseline:
• Age
• Sex
• cognitive function
• brain structure measured using Alzheimer’s disease signature cortical thickness. This reflects the mean cortical thickness across entorhinal, inferior temporal, middle temporal, and fusiform cortices and is not confounded by total intracranial volume (Jack Jr. et al., 2015; Neth et al., 2020).
In Aim 2, in addition to the covariates included in aim 1, the following baseline covariates will be added:
• white matter hyperintensity volume measured using FSL BIANCA
• white matter microstructural integrity measured by fractional anisotropy of the genu of the corpus collosum (Neth et al., 2020; Vemuri et al., 2018) generated using ExploreDTI
• physical frailty measured using grip strength and 5 metre walking speed
• alcohol use (as measured using the CAGE questionnaire)
• tobacco exposure (as measured using packyears which corresponds to the number of packs of cigarettes smoked per day by the number of years the person has smoked)
• body mass index
• depressive symptoms (as measured using the CES-D scale)
• systemic vascular health (as measured using a composite score of Cardiovascular and Metabolic Conditions; (2018)). The composite score of Cardiovascular and Metabolic Conditions score is calculated by summing the presence of the following conditions: hyperlipidemia, cardiac arrhythmia, coronary artery disease, congestive heart failure, diabetes mellitus, and stroke. Whereas Vemuri et al (2018) accounted for the presence of these conditions within the last five years, the present project will account for the presence of these conditions at the current or previous time-point which amounts to a three year period.
5. Operational definitions of concepts
CR will be operationalised as CR network strength. CR network strength will be developed using a machine learning model, connectome-based predictive modelling (Rosenberg et al., 2015; Shen et al., 2017), applied to a task-based connectivity matrix obtained from the Paper Folding task in the CR/RANN dataset. CR network strength reflects the connectivity strength of connections between different brain regions-of-interest that are positively associated with CR. Using 10-fold cross validation, the summed connectivity strengths will be used to predict a proxy measure of CR, which will be defined by the residual from a multiple regression predicting episodic memory from hippocampal volume, age, and sex. This model will then be externally validated on resting-state fMRI data in TILDA.
BM will be operationalised as a robust machine learning residual, the brain-predicted age difference, which compares an individual’s structural brain health, reflected by their voxel-wise grey matter density, to the state typically expected at that individual’s age. This measure was developed using a robust machine learning model, the Elastic Net with 10-fold cross validation and a data resampling ensemble approach, applied to open-access structural MRI data from 1,359 participants. In this model, voxel-wise grey matter density values were used to predict chronological age. This model was then applied to three independent datasets, including the TILDA dataset, where it accurately predicted chronological age in each dataset (Boyle et al., 2020). Brain-predicted age difference scores were then created by subtracting the brain-predicted age from chronological age, with positive scores reflected worse brain health, or BM, and negative scores reflecting better brain health, or BM.
6. Specific Aims and Hypotheses
The key criterion for satisfying an operational definition of CR is that a candidate CR measure must exert a moderation effect on cognitive function, such that it moderates the relationship between brain structure or pathology and clinical status or cognitive function (Stern, Arenaza-Urquijo, et al., 2018). This criterion can be examined in a moderated multiple regression model where cognitive function is predicted using variables representing brain structure, a CR measure, and the interaction term comprising the product of brain structure and the CR measure.
Moderation effects suffer from greatly reduced statistical power when there is measurement error in the variables in the interaction term and/or when either variable are themselves associated with the outcome variable (McClelland & Judd, 1993; Whisman & McClelland, 2005). This is pertinent for functional neuroimaging measures of CR, which will inevitably contain considerable levels of noise (Murphy et al., 2013) and because cognitive function is often associated with brain structure variables (Hedden et al., 2014; Tsapanou et al., 2019) and neuroimaging measures of CR (Stern, Gazes, et al., 2018; van Loenhoud et al., 2020). Moreover, longitudinal neuroimaging studies suggest that CR delays onset of cognitive decline but might not alter the rate of decline (Soldan et al., 2020) which casts doubt on the existence of moderation effects for CR. As such, the moderation effect criterion may be too strict and could result in promising candidate CR measures being too easily discounted.
A less rigorous criterion, the independent effect, holds that a candidate CR measure should demonstrate a positive association with cognitive function, after accounting for brain structure (Stern, Arenaza-Urquijo, et al., 2018). However, the closely related concept of BM has been also positively associated with cognitive function (Boyle et al., 2020; Cole et al., 2018; Elliott et al., 2019; Habeck et al., 2017; Liem et al., 2017). Consequently, this criterion may not dissociate CR from BM, which could hinder attempts to uncover their neural bases.
In sum, the current operational definition of CR is impeded by a strict criterion which may dissociate between CR and BM but might not be statistically feasible and a lenient criterion which can be fulfilled but does not dissociate between CR and BM. If the moderation effect cannot be satisfied, one solution is to refine the independent effect criterion such that a measure of CR must be positively associated with cognitive function after accounting for additional brain structure variables as well as physiological and clinical health measures that are related to BM (Cole et al., 2018; Elliott et al., 2019; Franke et al., 2013, 2014; Han et al., 2019; Kolenic et al., 2018; Lange et al., 2020; Ning et al., 2020; Ronan et al., 2016) and cognitive function (Anstey et al., 2007; Auyeung et al., 2011; Cronk et al., 2010; Hooghiemstra et al., 2017; Sabia et al., 2014; Samieri et al., 2018; Vemuri et al., 2018; Wilson et al., 2004; Yaffe, 2007).
The specific aims of the research proposal are to address the following research questions:
Aim 1: Are the current moderation effect and independent effect criteria appropriately rigorous for operational definitions of CR such that they can dissociate CR from BM?
• Hypothesis 1a: The moderation effect criterion will be satisfied by a measure of CR, but not by a measure of BM, in the presence of appropriate statistical conditions.
• Hypothesis 1b: The independent effect criterion will be satisfied by a measure of CR, and by a measure of BM, and as such it will not be rigorous enough to dissociate CR from BM.
Aim 2: Can the independent effect criterion be refined such that it will be satisfied by a measure of CR but not by a measure of BM?
• Hypothesis 2: The independent effect criterion, adjusting for physiological and clinical health covariates related to BM and cognitive function, will be satisfied by a CR measure, but not by a measure of BM.
Confirmation of hypothesis 1a would demonstrate that the current independent effect criterion is too lenient such that it cannot dissociate between measures of CR and BM. Confirmation of hypothesis 1b would restrict the operational definition of neuroimaging measures of CR to measures that demonstrate moderation effects. This refined operational definition would enable measures of BM and CR to be dissociated. Given the noted difficulties in detecting moderation effects (McClelland & Judd, 1993), confirmation of the hypothesis 2 would refine the independent effect criterion so that the operational definition of CR can be satisfied without demonstrating moderation effects but that can dissociate putative CR measures from BM measures. This would allow candidate CR measures to be validated against a statistically achievable criterion which can dissociate CR from the closely related concept of BM. Ultimately, this would enable researchers to apply such measures in order to study the neural basis of CR without mistakenly focusing on the neural basis of BM.
7. Statistical plan
For Aim 1, hierarchical moderated multiple regressions predicting longitudinal cognitive function from CR and BM, respectively, will be conducted. In the first step to assess the ‘independent effect’, covariates will include brain structure, as measured by Alzheimer’s disease signature cortical thickness, a validated biomarker of neurodegeneration that is not confounded by total intracranial volume (Jack Jr. et al., 2015; Neth et al., 2020), gender, baseline age, and baseline cognitive function. In the second step to assess the ‘moderation effect’, the interaction term for CR or BM and brain structure will be included in the model. This analysis will have a sample size of 351.
For Aim 2, multiple regressions predicting longitudinal cognitive function from CR and BM, respectively, will be conducted. Standard covariates will include gender, baseline age, and baseline cognitive function. Brain structure will be measured using Alzheimer’s disease signature cortical thickness, white matter hyperintensity volume, generated with FSL BIANCA (Griffanti et al., 2016), and by fractional anisotropy of the genu of the corpus callosum (Neth et al., 2020; Vemuri et al., 2018), generated with ExploreDTI. Additional baseline covariates will include systemic vascular health (Cardiovascular and Metabolic Conditions composite score; (Vemuri et al., 2018), markers of physical frailty (grip strength and walking speed; (Killane et al., 2013; Orr et al., 2020)), alcohol use (CAGE Questionnaire; (Ewing, 1984)), tobacco exposure (packyears; (Sheridan et al., 2018)), body mass index, and depressive symptoms (CES-D Scale; (Radloff, 1977)). This analysis will have a sample size of 309.
For both aims, cognitive dependent variables will include global cognition (MMSE), executive function (Animal Naming Test), and episodic memory (composite measure of immediate and delayed recall). 12 statistical models will be analysed in total (3 cognitive dependent variables each for CR and BM separately in Aims 1 and 2). To control for multiple comparisons, a maximum statistic correction will be applied which is more appropriate than other methods (e.g., Bonferroni) when there are correlated dependent measures (Conneely & Boehnke, 2007; Dudoit et al., 2003). All continuous variables will be z-scored to reduce multicollinearity. To attenuate the effect of outliers, a robust Winsorization technique will be used based on the median absolute deviation (Leys et al., 2013).
8. Add wiki / GitHub if available for the code
Python code for hierarchical regressions (including diagnostic/assumption tests): https://github.com/rorytboyle/hierarchical_regression
Python code for robust outlier detection: https://github.com/rorytboyle/robust_stats
ANALYTICAL PLAN - Daniel Gray
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Daniel | Gray | University of Arizona |
Co-PI | Carol | Barnes | University of Arizona |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
Ages: Young (6-8 mo); Aged (23-25 mo)
Sex: Male
Cross Sectional
New data on existing tissue
Expected completion – 9/1/2021
4. Key variables analyzed/measured in the study
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
Outcome measurements: Perineuronal net density, number of NPTX2 clusters
5. Operational definitions of concepts
6. Specific Aims and Hypotheses
Specific Aims: To immunohistochemically-label NPTX2, PNNs, and PV neurons in the CA3 region of the hippocampus in high-, average-, and low-performing, cognitively assessed aged rodents.
Hypothesis: Given the role of both NPTX2 and PNNs in synaptic plasticity and their known interactions at PV-expressing interneurons, we hypothesize that older animals with better hippocampus-dependent cognition will have higher PNN densities and more NPTX2 clusters on PV interneurons.
7. Statistical plan
Cognitive Assessment: Performance measure on the spatial version of the Morris watermaze will be used to classify rats as high-, average-, or low-performing for their age group. For rats in each age group, Barnes lab historical data were used to calculate the 25th and 75th percentiles for that group, and these cutoffs were used to distinguish low-, average- and high-performing rats. For the current experiment we will compare young “average-“, old “high-“, “old average-” and “old low-” performing groups.
Power Analysis: A power analysis (0.8 power; α = 0.05; 2-sided) using preliminary data from PV-PNN stained macaque tissue from our laboratory indicates that 12 rats from each group will be sufficient to reliably detect group differences.
Analysis of Anatomy data: All anatomical outcome measures will be statistically assessed using Multi-factor Analysis of Variance analyses.
8. Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Colin Groot
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Colin | Groot | Amsterdam UMC |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
WP1
– Participants for our discovery sample will be amyloid-positive human individuals from ADNI (estimated age, sex and race; 70±10 years, 50% female, primarily non-Hispanic white). Our replication sample will include amyloid-positive human individuals from the Amsterdam Dementia cohort (ADC) selected to match the discovery sample (estimated age, sex and race; 70±10 years, 50% female, primarily non-Hispanic white).
– Existing data to be collected is longitudinal cognitive data of at least 5 timepoints prior to diagnosis, at least two timepoints with CSF and MRI data prior to diagnosis.
WP2
– Participants will be selected based on the same criteria as in WP1. ADNI will again be used as the discovery sample and ADC as the replication sample
– Existing data to be collected will be longitudinal data from at least 2 timepoints after AD diagnosis with cognitive, CSF and MRI data
4. Key variables analyzed/measured in the study
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
WP1:
Predictor variable measurement
– Structural properties of the brain
o Gray matter atrophy
o White matter hyperintensities
– AD pathology
o CSF amyloid
o CSF total tau
o CSF p-tau
Outcome measurement
The following characteristics of the inflection point (see statistical plan) will be used as outcomes:
– Concurrent level of cognitive performance at inflection point
– Time to dementia from inflection point
Covariates:
Age, sex, intracranial volume (for structural properties of the brain)
WP2:
Predictor variable measurement
– Years of education, dichotomized into high education vs low education using a mean split
Outcome measurement
– Cognitive decline after AD diagnosis
– Longitudinal structural properties of the brain
o Gray matter atrophy
o White matter hyperintensities
– AD pathology
o CSF amyloid
o CSF total tau
o CSF p-tau
Covariates
Age, sex, intracranial volume (for structural properties of the brain)
5. Operational definitions of concepts
WP1:
Cognitive reserve will be operationalized by the concurrent level of cognitive performance at the inflection point and by time to dementia from the inflection point. Higher performance and longer time to dementia will indicate more reserve. Brain reserve will be operationalized as structural brain properties and brain maintenance will be operationalized as change over time in structural properties and AD pathology.
WP2
For this work package, cognitive reserve will be operationalized by years of education, and will be dichotomized into low and high reserve using a mean split for education. Brain reserve will again be operationalized as structural brain properties and brain maintenance will again be operationalized as change over time in structural properties and AD pathology.
6. Specific Aims and Hypotheses
WP1
Validation step: We hypothesize that higher education will be related to higher inflection points (higher initial performance) as well as delayed inflection points (longer time to dementia).
Hypothesis WP1: We hypothesize that structural brain properties (brain reserve) are more strongly associated with higher initial performance (higher inflection points) and that reduced pathological build-up (brain maintenance, e.g. CSF-tau) plays a larger role in delayed onset of cognitive decline (longer time to dementia).
WP2
Validation step: We hypothesize that higher education will be associated with faster cognitive decline, thereby replicating findings with regard to increased decline with higher reserve
Hypothesis WP2: If cognitive decline in individuals with high reserve is only related to baseline structural properties of the brain and not with change over time, this would support the notion of static effects of brain reserve. However, we hypothesize that accelerated cognitive decline in individuals with high education is coupled with increased rates of degeneration and build-up of AD pathology, supporting the notion of a (loss of) brain maintenance capability after cognitive decline has set in.
7. Statistical plan
WP1:
Longitudinal cognitive measures (at least 5 timepoints) will be used to establish individual inflection point towards cognitive decline using Bayesian change point analysis. The concurrent level of cognitive performance at inflection point and time to dementia from inflection point are used to operationalize different aspects of reserve (see Operational definitions of concepts). In a validation step, we will first relate the inflection point characteristics to education using linear regression analyses. We will then use linear mixed model analyses to predict height and onset of inflection points from (changes in) structural properties of the brain (gray matter atrophy, white matter hyperintensities) and AD pathology (CSF-amyloid-beta 42, total tau and p-tau) using linear mixed effects models.
WP2:
First, we will try to replicate earlier findings regarding increased decline with higher reserve by assessing the association between education and rates of cognitive decline using linear mixed effects models. Dichotomized high reserve vs low reserve will then be used to assess the association between cognitive decline and high/low reserve, longitudinal degeneration of structural brain properties and longitudinal change in AD pathology using linear mixed effects models.
8. Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Christian Habeck
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Christian | Habeck | Columbia University |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
1. We will use existing fMRI activation and functional-connectivity data from the Reference Ability Neural Network (RANN) study in our laboratory as the discovery sample to derive functional activation patterns. We will also use this data for quasi-replication of any findings.
2. For genuine replication in de-novo data sets, we will use fMRI activation data from several other tasks collected in our laboratory, a verbal Working-Memory task, a set-switching task.
4. Key variables analyzed/measured in the study
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
Data inputs (not constructs or analysis outcomes):
Brain Structure: 68 Regions of interest (ROIs) of cortical thickness and volume; 18 White-Matter tract integrity values
Functional brain activation: fMRI maps from first-level time-series modeling
Cognitive performance: percentage accuracy or reaction time of in-scanner performance
Demographics: age, education, gender, verbal intelligence
Derived Cognitive-Reserve construct: an fMRI activation pattern assigning loadings to every voxel location in the brain
5. Operational definitions of concepts
Cognitive Reserve is operationalized as an fMRI activation pattern that (1) accounts for cognitive performance and (2) is uncorrelated with brain structure and age (rather than just accounting for cognitive performance beyond brain structure and age which is a weaker constraint).
We will check whether this pattern also correlates with traditional CR proxies, and whether it shows unique correlations with cognitive performance beyond brain structure and age in held-out data.
6. Specific Aims and Hypotheses
1. fMRI activation patterns can be derived which account for cognitive performance and which are orthogonal to brain structure and age. (Lack of confirmation would be constituted by: PCA-regression to derive such activation patterns fails to produce one at p< 0.05.)
2. The derived CR-pattern from step 1 shows a significant correlation with traditional CR proxies like education and verbal intelligence. (Lack of confirmation would be constituted by: Pearson correlations fails to show a positive correlation at p< 0.05, hinting at orthogonal contribution to these CR-proxies.)
3. The derived CR-pattern also accounts for cognitive performance beyond brain structure and age in held-out data. (Lack of confirmation would be constituted by: Partial correlation between pattern score and cognitive performance, using brain structure and age as covariates, does not yield a positive correlation at p< 0.05.)
7. Statistical plan
The full plan is developed with mathematical notation in the grant narrative, so I will give only a brief conceptual version.
Derivation of functional MRI activation patterns:
1. Orthogonalize fMRI data and cognitive outcome with respect to brain structure and age;
2. Use PCA-regression (Scaled Subprofile Modeling; SSM) to derive a residual activation pattern that accounts for residual cognitive performance; the pattern score will now be orthogonal to brain structure and age by design
Association of pattern score with CR proxies:
We will test the associations of pattern score with traditional CR proxies like education and verbal intelligence using standard Pearson correlation.
Out-of-sample replication:
We will forward apply the derived patterns(s) to held-out data and test whether the resulting pattern score accounts for cognitive performance beyond age and brain structure.
8. Add wiki / GitHub if available for the code
Code and data will be available on request as Matlab code and Matlab archive, respectively, with annotations.
I will look into a WiKi to supply pseudo code to make the analytic stream clearer.
ANALYTICAL PLAN - Holly Hunsberger
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Holly | Hunsberger | Columbia University |
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
Age: 2 months, 8 months (longitudinal)
New data: Completion date 9/1/2021
Independent variable: Training/Education in the repeated acquisition and performance chamber along with natural aging.
Dependent variable/outcome: Performance in cognitive contextual and spatial tasks (behavior). To determine the neural correlates of cognitive reserve and brain maintenance, we will tag a contextual memory in our male and female ArcCreERT2 x EYFP mouse model at 8 months of age. These mice will either have gone through 2 months of RAPC or have aged normally. As a marker of brain aging, we will also quantify senescent cells.
5. Operational definitions of concepts
Cognitive reserve: The brain’s ability to buffer against neuropathology and cognitive stressors that lead to degeneration. Here, education or behavior training will be defined as boosting cognitive reserve.
Brain maintenance: A compliment of CR, refers to the relative absence of change in the brain over time with age. To test brain maintenance, we will have a control group that ages without training
6. Specific Aims and Hypotheses
AIM 1: Test the hypothesis that educational training buffers against age-related cognitive decline (ARCD) in ArcCreERT2 x EYFP mice: I hypothesize that education will protect against ARCD and therefore, enhance CR.
We have included a wide-array of behavioral paradigms and are not reliant on one task to accurately assess cognition. However, we are attempting training for the first time in the RAPC and this specific timeline may not produce robust results. If this occurs, we have the option to let the animals age until 12 or 24 months. Our AD mouse line is also bred to the ArcCreERT2, which we can test as an alternate model. We can also increase training to 8 weeks. These parameters can be modified. Additionally, we can use other types of training tasks such as the radial arm maze and hole board
AIM 2: Test the hypothesis that there is a specific neural signature of cognitive reserve in ArcCreERT2 x EYFP mice: I predict that memory trace activation in trained mice will be increased throughout the hippocampus. The number of reactivated memory trace cells will represent a specific neural signature of CR.
Although unlikely, it is possible that training does not affect memory trace cells in the hippocampus. If we do not observe differences in the hippocampus, then we will investigate other brain regions. We can also assess differences in encoding or retrieval alone; it is also possible that trained mice do not exhibit differences in whole brain memory traces during retrieval, but instead exhibit enhanced encoding of the memory. If this is the case, we can examine cells immediately after encoding of the memory
7. Statistical plan
-Ymaze: working memory
-Novel object: short-term memory
-Contextual fear conditioning: Long-term memory
Possible other tests could include anxiety/depression and social measures.
AIM 2: T-tests will be used to compare memory trace activation in trained and non-trained mice. To compare trained/untrained and male and female, a Two-Way ANOVA will be performed using a tukey posthoc. . Multivariate correlations and R-studio network analysis will be used to identify novel brain regions involved in CR.
8. Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Sarah-Naomi James
1. Study Title
First Name | Last Name | Organization | |
---|---|---|---|
PI | Sarah-Naomi | James | University College London |
Co-PI | Marcus | Richards | m.richards@ucl.ac.uk |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
Predictor variable measurement:
1. Early social circumstances (Paternal occupational status [coded as: (0) non-manual occupation (1) manual occupation], Mother’s education (coded as: (0) primary only (1) Secondary or any formal qualifications) and material home conditions [previously coded as (0) very good (1) good (2) modest (3) poor (4)](21).
2. Childhood cognitive ability from tests assessed at age 8 (standardised to sample)
3. Education attainment up to age 43 (coded as: (0) no qualification (1) vocational and ordinary secondary qualifications (2) advanced secondary qualifications and higher)
4. Own occupational status up to age 53 (coded as: (0) higher and lower managerial, administrative and professional; and small own account worker occupations (1) lower supervisory and technical occupations, semi-routine and routine occupations)
5. Crystallised ability using the National Adult Reading Test (NART) at age 53 (continuous)
Outcome
At age 69-71 we assessed a range of cognitive tests including the Mini Mental State Examination (MMSE); logical verbal episodic memory from the Wechsler Memory Scale-Revised (WMS-R); digit-symbol substitution test (DSST) from the Wechsler Adult Intelligence Scale-Revised (WAIS-R); the 12-item Face-Name test (FNAME-12). We additionally used these measures to derive the Preclinical Alzheimer Cognitive Composite (PACC) which will be the main outcome measure (17).
Neuroimaging outcomes measures
Neuroimaging at age 69-71 was performed on a single Biograph mMR 3T PET/MRI scanner (Siemens Healthcare, Erlangen), with simultaneous acquisition of dynamic PET/MR data; the full imaging protocol has been described previously (13) (17). The following variables will be assessed and have already been generated using validated pipelines (detail in appendix).
1. PET F18florbetapir Aβ status (Aβ+ or Aβ- status)
2. Whole white matter hyperintensity (WMH) volume (continuous)
3. Whole and hippocampal brain volume (continuous)
4. Cortical thickness measures – frontal, temporal and AD signatures (continuous)
covariates:
age at scan, sex
5. Operational definitions of concepts
6. Specific Aims and Hypotheses
A) Address which typical CR proxy exposures are associated with the development of a range of pathologies in the pre-symptomatic window, and to what extent the associated pathologies mediate the relationship between the CR proxy and cognition.
Outcome: In the case of evidence emerging of a mediation effect, the proxy would be considered to be associated on the causal pathway to pathology development and subsequent cognitive function. If so, when considering the pathology of interest, the proxy would be linked to the concept of brain maintenance, not the classical cognitive reserve model.
Hypothesis: A) Typical CR proxy exposures will be associated with the development of some, but ,not all, pathologies present in the preclinical window, and these associated pathologies will mediate the relationship between CR proxy exposure and cognition. For example, lower education attainment will be associated with greater white matter hyperintensity (WMH) volume and will partly mediate the association between education and cognition; but education will not be associated with Aβ status.
B) Address which, and to what extent, typical CR proxies moderate the cognitive expression from a range of pathologies in the pre-symptomatic window that do not show evidence of a direct association with CR proxy exposures (Question A).
Outcome: Significant interactions will indicate that the exposure does moderate the negative expression from the pathology measure of interest. If so, when considering the pathology of interest, the exposure would be in line with the classic cognitive reserve concept outlined as it captures sustained cognitive performance in the face of a specific disease.
Hypothesis: B) For pathologies that do not show evidence of a direct association from CR proxy exposures (from question A), we hypothesise that CR proxies will moderate their cognitive expression, in line with a classical CR model. For example, there will be a significant interaction between educational attainment and Aβ status on cognition.
7. Statistical plan
B) To investigate which, and to what extent, CR proxy exposures moderate the expression of specific pathologies that were not shown to be directly associated with CR proxy exposures (question A), we will conduct linear regression models using cognitive function as an outcome and test for interaction terms between each CR proxy exposure and each pathology measure of interest, independently.
8. Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Anna Marseglia
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Anna | Marseglia | Karolinska Institutet |
Co-PI | Eric | Westman | Karolinska Institutet |
Co-PI | Gustav | Mårtensson | Karolinska Institutet |
Co-PI | Daniel | Ferreira | Karolinska Institutet |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
Cross-sectional and longitudinal data from two existing human cohort studies:
1) UK Biobank, including ~9 500 people, aged 50-90 yrs., without dementia, neuropsychiatric disorders or other diseases (identified with the International Classification of Diseases 9/10th versions), having good-to-excellent self-reported health.
2) Gothenburg H70/H85 studies, including 790 people aged 70 years and 190 people aged 85 years without dementia or neuropsychiatric disorders with available 3T brain MRI data. Follow-up data is available for a subsample of 200 people from H70 and the full H85 cohort at age 75 and 88 years, respectively.
4. Key variables analyzed/measured in the study
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
BRAIN MRI MEASURES. For the UK Biobank, details on the acquisition protocols are provided here: https://biobank.ctsu.ox.ac.uk/crystal/crystal/docs/brain_mri.pdf. The UK Biobank and H70/H85 studies share very similar MRI protocols. In both datasets, using multiple structural MRI modalities—T1-weighted, T2-weighted FLAIR, diffusion-weighted tensor imaging (DTI), and susceptibility-weighted imaging (SWI)—the following brain volumetric measures will be derived: cortical thickness, total brain tissue (grey matter and white mater), white matter hyperintensities, white matter integrity, and microbleeds. Images are analysed using FreeSurfer (version 6.0; http://freesurfer.net/) through the-Hive database system, located in our research group at Karolinska Institute (PI: Prof. Westman E). This database system provides ultra-fast processing of brain MRI data so that the amount of data can be analysed within the time-frame of the current project.
COGNITIVE DOMAINS. Standardized cognitive tests sensitive to normal and pathological aging —Memory-in-Reality (free recall), Thurstone Picture Memory, Ten Word Memory List, Digit Span, Verbal fluency, Figure Identification, Koh’s Block test—are available at baseline and follow-up in the H70/H85 studies. These test scores cover the major cognitive domains (attention/processing speed, memory, executive function, visuospatial abilities, verbal fluency), which will be derived as composite domain based on tests’ z-scores.
CSF BIOMARKERS. In the H70/H85 studies, CSF biomarkers of AD neuropathology (concentrations of amyloid-β42 [Aβ42], total-tau, tau phosphorylated at threonine 181 [p-tau]) are available for a subset of participants (n=322) via lumbar puncture.
PROXIES OF COGNITIVE RESERVE. In this pilot, we will use a well-established socio-behavioural proxy of CR, education, in terms of: 1) years of education, and 2) educational level attained (primary, secondary, and high school/university). We opted for education because it is the most commonly used proxy for CR in the literature, and because of the project feasibility (to be completed within one year).
5. Operational definitions of concepts
OPERATIONALIZATION OF RESILIENCE AND BRAIN MAINTENANCE WITH DEEP LEARNING
1) Training set. In the UK Biobank, multimodal MRI images from healthy individuals (without dementia, cognitive impairment, neuropsychiatric & other disorders) will be labelled with the person’s chronological age (CA). A convolutional neural network will be trained to predict the age of relatively healthy individuals from brain MRI images. The model implicitly learns to capture the healthy aging process, including the substantial inter-subject variability, which can be a confounder in many models. The trained model will be validated on a hold-out set from the UK Biobank cohort and will then be applied to the brain images of the individuals in the test set.
2) Test set. The baseline H70/85 cohorts will be used to generate the predicted brain age (PBA). We will apply the deep learning model, trained and developed on the training set, to cognitively healthy participants’ brain scans (14,23).
3) In cognitively healthy older people, resilience and brain maintenance will be defined as the differences between a person’s PBA and chronological age (CA). Specifically, resilience will be identified when a person’s PBA is greater than their CA (PBA>CA), whereas a PBA smaller than the CA will indicate brain maintenance mechanism (PBA< CA).
6. Specific Aims and Hypotheses
HYPOTHESIS. We hypothesized that discrepancies between PBA and CA in cognitively healthy older adults may reflect mechanisms: 1) Older-appearing brains (PBA>CA) in absence of cognitive dysfunction would indicate the presence of resilience mechanisms–despite ongoing neuropathological processes, the individual has preserved cognitive function; 2) Younger-appearing brains (PBA< CA) would indicate the presence of mechanisms of brain maintenance that might be preventing or slowing down the development of brain atrophy.
SPECIFIC AIMS
Aim 1. Developing biological measures of resilience and brain maintenance (RE/BM) based on multimodal structural MRI (T1, fluid-attenuated inversion recovery sequence [FLAIR], and diffusion MRI) using deep learning methods.
Aim 2. Examining the relation between the generated measures (RE/BM) and a common socio-behavioural proxy for cognitive reserve (education), in-vivo cerebrospinal fluid (CSF) biomarkers of Alzheimer’s disease-related neuropathology, other risk factor for aging (e.g. vascular risk factors, or cardiometabolic disease), as well as with cognitive changes.
7. Statistical plan
Alongside deep learning, we will also use a simpler method, Orthogonal Partial Least Squares to latent structures (OPLS) analysis. We will compare the findings using CR measures obtained with both methods. This is because while OPLS may provide less accurate predictions than deep learning methods, it is more transparent and can offer better interpretability. In addition:
1) Linear or logistic regression models will be used to examine cross-sectional associations between RE/BM and education as well as between RE/BM and individual biomarkers of AD-pathology, and to identify positive and negative factors associated with RE/BM.
2) In a subset with longitudinal cognitive data, linear mixed-effect models will be used to address associations between RE/BM and cognitive outcome and between individual biomarkers of AD-pathology and cognitive outcome.
3) Whether and to what extent RE/BM modifies the association between CSF biomarkers of brain neuropathology and cognitive outcome using interaction analysis, stratification, and/or mediation analyses.
The role of factors (e.g. sex, genetic risk factor for dementia) influencing the aforementioned mentioned associations will be considered in data analysis.
The deep learning model will be developed in Python using the machine learning library PyTorch. SIMCA and Stata-SE version 16 statistical software programs will be used in the OPLS analysis and in the analyses described in steps 1-3 above.
ANALYTICAL PLAN - Lidia Vaque-Alcazar
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Lídia | Vaqué-Alcázar | University of Barcelona |
Co-PI | Lars | Nyberg | Umeå University |
Co-PI | Anders | Fjell | University of Oslo |
Co-PI | Álvaro | Pascual-Leone | Hebrew SeniorLife, Harvard Medical School |
Co-PI | David | Bartrés-Faz | University of Barcelona |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
4. Key variables analyzed/measured in the study
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
5. Operational definitions of concepts
Broadly, based on the well-known associations between episodic memory and hippocampus, we will apply an iterative process based on a decision tree to identify whether the MRI substrates underlying stable episodic memory are related to the lack of longitudinal changes as regards hippocampal structure and/or functionality, or this maintenance of performance is sustained by other mechanisms typically associated with CR (i.e. efficiency or compensation).
6. Specific Aims and Hypotheses
Our specific goals are:
1. To identify those subjects who show correlated episodic memory and hippocampal volume stability (defined as BM) versus those who show episodic memory stability despite progressive hippocampal volume shrinkage (termed here for convenience ‘deviant cases’).
2. To analyze whether longitudinal episodic memory stability among the ‘deviant cases’ at structural level is associated with functional mechanisms that may serve as compensatory brain changes, specifically changes in i) resting-state functional MRI hippocampal connectivity or ii) episodic memory task-based activated areas.
3. To elucidate different levels of functional efficiency and compensation (i.e. CR-related mechanisms) underlying stable episodic memory in the context of hippocampal atrophy and changes in memory systems brain functionality (hippocampal and medial temporal lobe connectivity).
4. To investigate the role of education and other lifestyle factors as possible mediators of the categorizations identified.
5. To replicate our findings on a middle-aged cohort of healthy participants (N=1000).
We predict that our results will allow us to establish a new categorization within the concept of BM (discerning between structural and functional component) and that will better define CR longitudinally in the context of atrophy, by identifying different pathways within this construct. Related to the study of CR-related measures, we hypothesize that a similar proportion of high and low educated elders will be identified in the BM subgroups, while the CR classifications will be more frequently assigned to elders with a higher level of education (i.e. higher CR).
7. Statistical plan
1. Age-adjusted residual approach in the association between hippocampal volume and episodic memory to identify BM pathway 1 (x>0 and y>0, positive change in episodic memory and hippocampus) and ‘deviant cases’ (x>0 and y< 0, positive change in episodic memory and negative in hippocampus) groups.
2. Age-adjusted residual approach in the association between hippocampal functional connectivity or typical episodic memory task-related areas and episodic memory performance stability across the ‘deviant cases’ group to identify 2 groups: BM pathway 2 (x>0 and y>0, positive association between changes in hippocampal function and episodic memory), and subjects who need to use other circuits to maintain performance, and therefore will be employing CR-related mechanisms (x>0 and y< 0, no relationship between episodic memory and hippocampal structure or functionality/connectivity longitudinal changes).
3. We will analyze whether episodic memory is sustained by CR mechanisms related to reorganization/expansion of episodic memory task-related areas or by a general independent CR task-invariant network expression when the typical neural substrates are damaged (at both structural and functional level).
4. In parallel, additional analyses will be undertaken to investigate, if classical proxies considered within the CR theory are associated with the identified terminological categories.
5. Finally, we will follow the same procedure in an independent sample of middle-aged and older adults (N=1000).
8. Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Eero Vuoksimaa
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Eero | Vuoksimaa | University of Helsinki |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
4. Key variables analyzed/measured in the study
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
Predictors of episodic memory trajectory are 1) brain maintenance (BM) as measured with cortical thickness and cortical surface area change in Alzheimer’s Disease signature region-of-interest from VETSA 1 to VETSA 3; and 2) lifetime years of education (EduYears) at VETSA 1 & young adult cognitive ability (GCA) as measured with the Armed Forces Qualification Test at a mean age of 20 years, and also 3) EduYears-BM and GCA-BM interactions. Outcome will be episodic memory trajectory as defined by the group membership (stable vs. decliner) based on latent growth mixture modeling. Covariates include baseline age, scanner and MRI follow-up time.
5. Operational definitions of concepts
Cortical reconstruction in Freesurfer will be used to yield cortical thickness (CT) and surface area (SA) measures. Desikan-Killiany cortical parcellation will be used to derive regions-of-interests (ROIs). As suggested by Schwarz et al. (2016) the following ROIs will form the bilateral cortical signature of AD (AD-ROI): entorhinal cortex, inferior temporal gyri, middle temporal gyri, inferior parietal cortex, fusiform, and precuneus. BM will be operational by calculating brain change in AD-ROI in CT and SA. AD-ROI will be adjusted for global measures of mean CT and total SA.
CT maintenance = (AD-ROI CT V3 / mean CT V3) – (AD-ROI CT V1 / mean CT V1),
SA maintenance = (AD-ROI SA V3 / total SA V3) – (AD-ROI SA V1 / total SA V1),
where V1 = baseline at VETSA 1, V3 = follow-up data from VETSA 3.
Proxy measures of cognitive reserve are lifetime years of education at VETSA 1 and young adult general cognitive ability at a mean age of 20 years as measured with Armed Forces Qualification Test (percentile scores).
Episodic memory trajectories will be derived by using latent growth mixture modeling. We will use a composite episodic memory score consisting of six episodic memory tests: (1) Logical Memory immediate and (2) delayed free recall, (3) Visual Reproductions immediate and (4) delayed free recall, and (5) California Verbal Learning Test-II total words in trials 1-5 and (6) delayed free recall. We will use two groups: stable and decliners. In case of multiple declining groups we will combine these groups as one decliner group. Stable group will be those without episodic memory change from VETSA1 to VETSA3 or alternatively the group with smallest decline (in case all sub-groups have significant decline).
6. Specific Aims and Hypotheses
The specific study questions are: 1) can those with less relative brain change (greater BM) sustain their EM performance better than those with more relative brain change (poorer BM)? 2) can those with higher CR maintain their EM performance I) in aging and II) in the face of Alzheimer’s Disease related brain changes better than those with lower CR?
Hypotheses are: 1) those with less EM change have less cortical thinning and cortical contraction than those with more EM decline. 2) I) those with less EM change have higher education and higher young adult GCA compared to those with more EM decline. II) education and young adult GCA moderate the association between brain change and EM trajectory such that those with higher education/young adult GCA can tolerate more brain changes than those with lower education/young adult GCA.
7. Statistical plan
We will use the following linear regression models:
Model 1 “BM”, where predictor = AD-ROI (relative change from VETSA1 to VETSA3) and dependent variable = EM trajectory
Model 2 “CR – aging”, where predictor = Education / GCA (controlling for AD-ROI) and dependent variable = EM trajectory
Model 3 “CR – brain insult”, where predictor = education / GCA x AD-ROI interaction and dependent variable = EM trajectory
Dependent variable in all analyses is episodic memory class (stable vs. decliners based on trajectories). Co-variates in all models are baseline age, scanner and MRI follow-up time. Standard errors will be adjusted for family structure.
8. Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Thomas Foster/Brittney Yegla
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Brittney | Yegla | University of Florida |
Co-PI | Thomas | Foster | University of Florida |
3. Brief explanation of existing data or new data collection
1) Measures of aging
2) Variability in cognition
3) Gene expression within the medial prefrontal cortex and hippocampal regions (CA1, CA3, dentate gyrus).
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
5. Operational definitions of concepts
Brain maintenance: Preserved cognition associated with a delay in brain changes normally observed with aging. In this case, preserved cognition is associated with less of a difference in gene expression between age-unimpaired and young, and larger differences in gene expression are observed between age-impaired and young.
Network level cognitive reserve is related to the activity within networks in relation to cognitive function. We will measure expression of immediate early genes as a measure of circuit activity.
6. Specific Aims and Hypotheses
Aim1. Transcriptional markers of senescence. The goal of Aim1 is to operationally define brain aging genes in order to have a subset of genes for analysis in Aim2. Public databases (GEO) will be used to operationally define senescent genes as genes that consistently change with age in the same direction (increasing or decreasing) across multiple studies.
Aim2. Transcriptional markers of maintenance and cellular resilience. The goal of Aim2 is to operationally define maintenance and cellular resilience by re-examining our behavioral and transcriptional data using the list of senescent genes formed in Aim1.
Aim3. Transcriptional markers of network efficiency and compensation in preserving spatial memory. Aim3 will focus on a recent study, which characterized attentional set-shift behavior and spatial memory in young and aged F344 rats, and reanalyze gene transcription in the hippocampus and mPFC, focusing on hippocampal-dependent spatial memory.
7. Statistical plan
In initial analysis for Aim1 will involve directional filtering of gene expression. Using the mean expression from GEO data sets for young and aged animas we will determine genes that increase or decrease expression with aging, >75% of the time across studies. This will provide a subset of brain aging genes within each area for analysis in Aim2.
Aim2 will employ statistical tests (ANOVAs or t-tests) with behavioral characterization (young, age-impaired, age-unimpaired) to determine is expression of the subset of brain aging genes exhibit statistical differences between young and age-impaired or age-unimpaired.
Aim3 will employ statistical tests and correlations to determine if expression of immediate early genes in aged animals varies by cognitive function.
8. Add wiki / GitHub if available for the code
ANALYTICAL PLAN - Gabriel Ziegler
1. Study Title
2. Study Investigators
First Name | Last Name | Organization | |
---|---|---|---|
PI | Gabriel | Ziegler |
3. Brief explanation of existing data or new data collection
Please clarify the species (human, mice, or others specified), and when applicable, the age (age range, and/or average age), sex (% female), race/ethnicity, cross-sectional/longitudinal/both, existing data or new data collection, the expected collection completion date (if new data collection).
Please specify the following variables: predictor variable measurement, outcome measurement, covariates.
We aim at application of machine learning techniques where the classic distinction of indpendent and dependent variables is partially not appropriate. There is a feature set, based on functional brain contrasts of the FADE fMRI-tasks. However we also consider for comparisons morphometric features (based on T1-weighted data) like voxel-based tissue probabilities (GM, WM, CSF). The to be applied ML algorithms aim at outlier-detection scheme and therefore do not primarily have an outcome measurement. The proposed alternative will be classical discrimative modelling using AD progression index as an outcome measurement. AD progression index (1D) will be based on a continous implementation of the ATN system (A-T-N-< A+T-N-< A+T+N-< A+T+N+) using CSF 42/40 Amyloid, p-tau and hippocampal atrophy as markers.
5. Operational definitions of concepts
6. Specific Aims and Hypotheses
Our core hypothesis is that multivariate indices are potentially more sensitive and reliable than the traditionally approaches. Here we aim to explore and formalize the notion of functional brain maintenance using activity pattern similarity (or deviation) in form of a one-class classification approach resulting in a single continuous index of maintenance. In this project using the longitudinal DZNE DELCODE cohort, the focus will be set on the following questions:
2. Can we validate the new maintenance index with respect to biomarkers, risk due to vascular pathology and measures of reserve such as global and local brain volume and cognitive performance ? We hypothesize independent contributions of maintenance and reserve to cognitive performance differences and a moderator effect of structural brain reserve on brain pathology indicator’s (Aβ/tau and WMH) effect on cognition.
In order to benchmark the functional to a structural-based maintenance index we compare the validation against biomarkers and cognition to the BrainAGE index (Franke et al., 2010). Although this is highly exploratory work, we expect a stronger association of BrainAGE index to existing cognitive differences and a higher associatioo of the function-based index to Amyloid biomarker status, respectively.
3. Are young adults or high maintenance elderly individuals more suited as a reference (training) sample for inference about unseen subjects at risk of Alzheimer’s disease ?
4. How does the one-class approach compare to alternative two-class discriminative approaches ?
7. Statistical plan
Work package 1: Operational definition of maintenance using a normative model of functional brain activity
Here we follow the established principle of casting patient classification as an outlier detection problem (Mourao-Miranda, 2011). Using multivariate metrics, measures of departure from normal brain activity (as originally proposed in Düzel et al., 2011) could extended using Euclidian or Mahalanobis distance from group mean activity in a young reference sample. However, the extremely large number of dimensions characteristic of neuroimaging data (hundreds of thousands of voxels) renders this difficult or impossible due to the small sample size typically available for neuroimaging datasets. One way of efficiently addressing this problem is to compute the probability density of brain activity contrast in a specific (high maintenance) reference class. When a new participant’s brain activity contrast falls below some density threshold this new example is considered abnormal (or more unusual). This is expected to happen when the spatial pattern of local networks recruited during task-related activity does change compared to the high maintenance reference sample. In addition, the distance of new unseen participant’s brain activity to the boundary (learning in the reference) can be used to quantify the degree of abnormality (or deviation) in single continuous index. Then, the latter index can be correlated with clinical outcomes (such as cognition and biomarkers) for validation. In this work package we aim to start adapt the previously established one-class Support Vector Machines (SVM) approach (Mourao-Miranda, 2011) for the quantification of brain maintenance using the FADE task paradigm.
More specifically, machine learning (or AI) kernel methods are used to define a similarity measure of whole brain functional maintenance using the novelty during encoding contrast two individuals performing the task. Linear or non-linear kernel mappings might be potentially useful to characterize different aspects activity pattern similarity relevant for the definition of maintenance. We therefore explore linear as well as non-linear relationships in the data. Using a nonlinear kernel matrix is equivalent to mapping the contrast maps from the original input space into a high dimensional feature space where the distance become more meaningful or the separation between the two classes can be easier (using a plain linear boundary).
Once the decision boundary is fitted during a model training process using a young reference sample, it can be used to classify and characterize unseen elderly participant’s maintenance as being either low (if they fall outside the boundary) or high (normal non-outliers). Alternative multivariate soft-margin (Shawe-Taylor & Christianini, 2004) and probabilistic approaches accounting for full model uncertainties (such as Gaussian Process Classification, Rasmussen & Williams, 2006) and Deep Learning (Pinaya et al., 2018) are explored and compared in this project. Reliability and robustness estimates of the proposed maintenance index will be assessed using longitudinal follow-up data in very low risk subsamples (of the well characterized DELCODE cohort). Accounting for known and undesired covariates is another important issue for single-subject inference methods to be considered in this project phase. The empirically optimized one-class machine learning approach using FADE paradigm’s novelty contrast will result in a new maintenance index using a sample of n=99 young (high maintenance) adults with ages 20-30 performing the scene-encoding (FADE) task.
Work package 2: Validation using elderly individuals with high maintenance and probing the hypothesis of brain reserve
In this work package we test whether the functional brain maintenance index as introduced in WP1 can be validated against a clinically defined high maintenance reference sample. These are older adults who are CSF biomarker and ApoE4 negative, do not have vascular lesions (based on White Matter Hyperintensities (WMH) and vascular risk scores), show absence of significant atrophy and remain cognitively stable (over follow-ups, which are available in DELCODE for several years). Based on previous findings it can be expected that this subsample shows high maintenance (Cabeza et al., 2018, Düzel et al., 2011). Furthermore we test whether there is evidence that it can be used as a determinant of preserved cognition (i.e. using individual level longitudinal cognitive trajectories over follow-ups in DELCODE).
In context of this call cognitive reserve is defined as multiple properties of the brain that might allow for sustained cognitive performance in the face of age-related changes and brain insult or disease (such as Aβ/tau or vascular pathology). Here we specifically focus on the potential reserve variables such as global and local (e.g. hippocampal network) gray matter volumes. In addition to the above validation, this WP explores the contribution (or dissociation) of several factors (when considered simultaneously) to cross-sectional cognitive performance in a single regression model. We include the predictors (1) functional maintenance; (2) structural brain reserve, (3) demographics, and the (4) presence amyloid/tau and vascular pathology. We further hypothesize independent contributions of maintenance and reserve to cognitive performance differences and a moderator effect of structural brain reserve on brain pathology indicator’s (Aβ/tau and WMH) effect on cognition.
Work package 3: Assessing the role of the reference sample.
One class approaches might be fundamentally biased by a somewhat arbitrary choice of the reference sample (and population). A crucial question is whether brain activity contrast differences induced by age and (often mixed) brain pathologies do align in the considered feature space. A related issue is whether one should assume that no functional reorganization e.g. as a consequence of decades of experience and structural brain ageing in high maintenance elderly individuals has happened. Both assumptions are implicitly made in the traditional FADE score (Düzel et al., 2011) and might play out differentially (1) when trying to assess maintenance in healthy functional ageing; or (2) for early identification of brain pathology.
Consequently, by using the traditional approach when extrapolating (from functionally healthy young individuals) ‘long-distance’ towards pathological ageing (i.e. by varying age & pathology simultaneously), we might not obtain an optimal characterization of individual differences and subtle changes (over follow-ups) for one or even both of these important scientific goals (1) and (2). As introduced above, the non-linear dependency of (early) hyper- and (later) hypo-activation along the healthy ageing to AD disease spectrum might also challenge performance of a naive (linear) extrapolation based on young individuals as a reference. In this work package we compare the above approach outlined in WP1 using a low risk (assumed high maintenance) elderly sample (n=100) as an alternative reference for the individual characterization of brain maintenance in unseen subjects. We consider both biomarker, vascular lesion load and cognition for stratification and compare those competing methods.
Work package 4: Does learning a discriminative model of functional brain maintenance offer a promising alternative ?
Our normative (one-class) approach to brain maintenance proposed in WP1 solves a more general (and difficult) problem than conventional two-class machine learning problems (e.g. classifying patients into NO vs. AD). However, two-class and one-class classification approaches address fundamentally different questions. The first finds the discriminative boundary between two classes while the second finds the boundary enclosing a specific class in relation to which activity patterns belonging to other classes can be detected as outliers. In clinical practice, if one is interested in training a classifier to discriminate two well-defined and homogeneous classes with high accuracy, the standard two class approach is expected to have the best performance. However, the proposed one-class approach will be more advantageous in situations where one class being more homogeneous and well defined and the other(s) being highly heterogeneous and/or with small sample sizes.
In this work-package we finally aim to compare the above one-class novelty detection approach to maintenance with an alternative approach explicitly learning the (non-linear) function characterizing the step by step transition of brain (activity) contrast maps from healthy ageing (Amyloid-,Tau-) towards pathological ageing individuals (Amyloid+, Tau+). Multiple kernel-based approaches (such as SVM or GPC) allow accounting for non-linear effects and assessing the predicted ‘distance travelled’ of a patient at risk during progression towards AD. We finally intend to validate both approaches by comparing their potential for early detection of those DELCODE participants that show cognitive decline over longitudinal follow-ups.
To compare the above function-based index with a brain structure-based maintenance index, we aim to complement the performance in the validation steps (WP2 & WP4) with the BrainAge index. The latter is described in detail in Franke et al., 2010, Neuroimage. The pre-trained regression-method will be applied to DELCODE subjects generating their individual Brain-Age which will be entered in all validation analyses using CSF biomarkers and cognition described above.
8. Add wiki / GitHub if available for the code