Forkhead box (Fox) transcription factors (TFs) mediate multiple conserved cardiogenic processes in both mammals and Drosophila. Our prior work identified the roles of two Drosophila Fox genes, jumeau (jumu) and Checkpoint suppressor 1-like (CHES-1-like), in cardiac progenitor cell specification and division, and in the proper positioning of cardiac cell subtypes. Fox TF binding sites are also significantly enriched in the enhancers of genes expressed in the heart, suggesting that these genes may play a core regulatory role in one or more of these cardiogenic processes. We identified downstream targets of Jumu by comparing transcriptional expression profiles of flow cytometry-sorted mesodermal cells from wild-type embryos and embryos completely lacking the jumu gene and found that genes with functional annotation and ontological features suggesting roles in cell division were overrepresented among Jumu targets. Phenotypic analysis of a subset of these targets identified 21 jumu-regulated genes that mediate cardiac progenitor cell division, one of which, Retinal Homeobox (Rx), was characterized in more detail. Finally, the observation that many of these 21 genes and/or their orthologs exhibit genetic or physical interactions among themselves indicates that Jumu is a master regulator acting as a hub of a cardiac progenitor cell division-mediating network.

Forkhead (Fkh/Fox) domain transcription factors (TFs) mediate multiple cardiogenic processes in both mammals and Drosophila. We showed previously that the Drosophila Fox gene jumeau (jumu) controls three categories of cardiac progenitor cell division-asymmetric, symmetric, and cell division at an earlier stage-by regulating Polo kinase activity, and mediates the latter two categories in concert with the TF Myb. Those observations raised the question of whether other jumu-regulated genes also mediate all three categories of cardiac progenitor cell division or a subset thereof. By comparing microarray-based expression profiles of wild-type and jumu loss-of-function mesodermal cells, we identified nebbish (neb), a kinesin-encoding gene activated by jumu. Phenotypic analysis shows that neb is required for only two categories of jumu-regulated cardiac progenitor cell division: symmetric and cell division at an earlier stage. Synergistic genetic interactions between neb, jumu, Myb, and polo and the rescue of jumu mutations by ectopic cardiac mesoderm-specific expression of neb demonstrate that neb is an integral component of a jumu-regulated subnetwork mediating cardiac progenitor cell divisions. Our results emphasize the central role of Fox TFs in cardiogenesis and illustrate how a single TF can utilize different combinations of other regulators and downstream effectors to control distinct developmental processes.

Enrichment analysis has been widely applied in the genome-wide association studies (GWAS), where gene sets corresponding to biological pathways are examined for significant associations with a phenotype to help increase statistical power and improve biological interpretation. In this work, we expand the scope of enrichment analysis into brain imaging genetics, an emerging field that studies how genetic variation influences brain structure and function measured by neuroimaging quantitative traits (QT). Given the high dimensionality of both imaging and genetic data, we propose to study Imaging Genetic Enrichment Analysis (IGEA), a new enrichment analysis paradigm that jointly considers meaningful gene sets (GS) and brain circuits (BC) and examines whether any given GS-BC pair is enriched in a list of gene-QT findings. Using gene expression data from Allen Human Brain Atlas and imaging genetics data from Alzheimer’s Disease Neuroimaging Initiative as test beds, we present an IGEA framework and conduct a proof-of-concept study. This empirical study identifies 12 significant high level two dimensional imaging genetics modules. Many of these modules are relevant to a variety of neurobiological pathways or neurodegenerative diseases, showing the promise of the proposal framework for providing insight into the mechanism of complex diseases.

Statistical shape analysis is a fundamental topic in biomedical image computing, and plays important roles in numerous applications in brain imaging. This chapter describes typical shape analysis methods for modeling and analyzing 3D surface data in brain imaging studies. These studies examine various structures of interest in the brain, and aim to identify morphometric differences associated with a particular condition to aid diagnosis and treatment. We first present classic computational methods for modeling and registering 3D surfaces, and then discuss advanced methods that take into consideration subfield information on the surface. After that, we describe techniques for shape analysis of registered surface models, including traditional general linear models as well as a newly proposed statistical learning model for performing surface-based morphometry. Finally, we provide a real world neuroimaging application to demonstrate the effectiveness of these techniques.

MOTIVATION:Structured sparse canonical correlation analysis (SCCA) models have been used to identify imaging genetic associations. These models either use group lasso or graph-guided fused lasso to conduct feature selection and feature grouping simultaneously. The group lasso based methods require prior knowledge to define the groups, which limits the capability when prior knowledge is incomplete or unavailable. The graph-guided methods overcome this drawback by using the sample correlation to define the constraint. However, they are sensitive to the sign of the sample correlation, which could introduce undesirable bias if the sign is wrongly estimated.RESULTS:We introduce a novel SCCA model with a new penalty, and develop an efficient optimization algorithm. Our method has a strong upper bound for the grouping effect for both positively and negatively correlated features. We show that our method performs better than or equally to three competing SCCA models on both synthetic and real data. In particular, our method identifies stronger canonical correlations and better canonical loading patterns, showing its promise for revealing interesting imaging genetic associations.AVAILABILITY AND IMPLEMENTATION:The Matlab code and sample data are freely available at http://www.iu.edu/∼shenlab/tools/angscca/CONTACT:shenli@iu.eduSUPPLEMENTARY INFORMATION:Supplementary data are available at Bioinformatics online.

Background: Recently, structured sparse canonical correlation analysis (SCCA) has received increased attention in brain imaging genetics studies. It can identify bi-multivariate imaging genetic associations as well as select relevant features with desired structure information. These SCCA methods either use the fused lasso regularizer to induce the smoothness between ordered features, or use the signed pairwise difference which is dependent on the estimated sign of sample correlation. Besides, several other structured SCCA models use the group lasso or graph fused lasso to encourage group structure, but they require the structure/group information provided in advance which sometimes is not available.Results: We propose a new structured SCCA model, which employs the graph OSCAR (GOSCAR) regularizer to encourage those highly correlated features to have similar or equal canonical weights. Our GOSCAR based SCCA has two advantages: 1) It does not require to pre-define the sign of the sample correlation, and thus could reduce the estimation bias. 2) It could pull those highly correlated features together no matter whether they are positively or negatively correlated. We evaluate our method using both synthetic data and real data. Using the 191 ROI measurements of amyloid imaging data, and 58 genetic markers within the APOE gene, our method identifies a strong association between APOE SNP rs429358 and the amyloid burden measure in the frontal region. In addition, the estimated canonical weights present a clear pattern which is preferable for further investigation.Conclusions: Our proposed method shows better or comparable performance on the synthetic data in terms of the estimated correlations and canonical loadings. It has successfully identified an important association between an Alzheimer's disease risk SNP rs429358 and the amyloid burden measure in the frontal region.

Depressive symptoms are common in older adults and are particularly prevalent in those with or at elevated risk for dementia. Although the heritability of depression is estimated to be substantial, single nucleotide polymorphism-based genome-wide association studies of depressive symptoms have had limited success. In this study, we performed genome-wide gene- and pathway-based analyses of depressive symptom burden. Study participants included non-Hispanic Caucasian subjects (n = 6,884) from three independent cohorts, the Alzheimer's Disease Neuroimaging Initiative (ADNI), the Health and Retirement Study (HRS), and the Indiana Memory and Aging Study (IMAS). Gene-based meta-analysis identified genome-wide significant associations (ANGPT4 and FAM110A, q-value = 0.026; GRM7-AS3 and LRFN5, q-value = 0.042). Pathway analysis revealed enrichment of association in 105 pathways, including multiple pathways related to ERK/MAPK signaling, GSK3 signaling in bipolar disorder, cell development, and immune activation and inflammation. GRM7, ANGPT4, and LRFN5 have been previously implicated in psychiatric disorders, including the GRM7 region displaying association with major depressive disorder. The ERK/MAPK signaling pathway is a known target of antidepressant drugs and has important roles in neuronal plasticity, and GSK3 signaling has been previously implicated in Alzheimer's disease and as a promising therapeutic target for depression. Our results warrant further investigation in independent and larger cohorts and add to the growing understanding of the genetics and pathobiology of depressive symptoms in aging and neurodegenerative disorders. In particular, the genes and pathways demonstrating association with depressive symptoms may be potential therapeutic targets for these symptoms in older adults.

Identifying associations between genetic variants and neuroimaging quantitative traits (QTs) is a popular research topic in brain imaging genetics. Sparse canonical correlation analysis (SCCA) has been widely used to reveal complex multi-SNP-multi-QT associations. Several SCCA methods explicitly incorporate prior knowledge into the model and intend to uncover the hidden structure informed by the prior knowledge. We propose a novel structured SCCA method using Graph constrained Elastic-Net (GraphNet) regularizer to not only discover important associations, but also induce smoothness between coefficients that are adjacent in the graph. In addition, the proposed method incorporates the covariance structure information usually ignored by most SCCA methods. Experiments on simulated and real imaging genetic data show that, the proposed method not only outperforms a widely used SCCA method but also yields an easy-to-interpret biological findings.

We used whole‐exome sequencing to identify variants other than APOE associated with the rate of hippocampal atrophy in amnestic mild cognitive impairment. An in‐silico predicted missense variant in REST (rs3796529) was found exclusively in subjects with slow hippocampal volume loss and validated using unbiased whole‐brain analysis and meta‐analysis across 5 independent cohorts. REST is a master regulator of neurogenesis and neuronal differentiation that has not been previously implicated in Alzheimer's disease. These findings nominate REST and its functional pathways as protective and illustrate the potential of combining next‐generation sequencing with neuroimaging to discover novel disease mechanisms and potential therapeutic targets. Ann Neurol 2015;77:547–552

Brain imaging genetics is an emergent research field where the association between genetic variations such as single nucleotide polymorphisms (SNPs) and neuroimaging quantitative traits (QTs) is evaluated. Sparse canonical correlation analysis (SCCA) is a bi-multivariate analysis method that has the potential to reveal complex multi-SNP-multi-QT associations. Most existing SCCA algorithms are designed using the soft threshold strategy, which assumes that the features in the data are independent from each other. This independence assumption usually does not hold in imaging genetic data, and thus inevitably limits the capability of yielding optimal solutions. We propose a novel structure-aware SCCA (denoted as S2CCA) algorithm to not only eliminate the independence assumption for the input data, but also incorporate group-like structure in the model. Empirical comparison with a widely used SCCA implementation, on both simulated and real imaging genetic data, demonstrated that S2CCA could yield improved prediction performance and biologically meaningful findings.

Identifying genetic variants other than APOE ε4 associated with rate of hippocampal volume loss, a key AD endophenotype, has been an important goal. Combining whole exome sequencing (WES) and imaging genetics we recently identified two functional single nucleotide variants (SNVs) in participants with APOE-ε3/ε3 genotypes (Nho 2013). Minor alleles were associated with more rapid hippocampal volume loss. In further analyses of the same WES data, we identified protective functional SNVs. WES with an extreme trait design was used to compare matched unrelated APOE-ε3 homozygous male Caucasian participants with mild cognitive impairment (MCI) from ADNI selected on the basis of extremes of the 2-year longitudinal change distribution of hippocampal volume (8 subjects with rapid rates of atrophy and 8 with slow/stable rates of atrophy). To validate and extend the exome findings in a larger sample, we conducted quantitative trait analysis including whole-brain search and meta-analysis across 5 independent cohorts (ADNI-1/2, AddNeuroMed, MIRAGE, and Indiana Memory and Aging Study). We identified 56 non-synonymous SNVs that were found exclusively in at least 4 of 8 subjects in the slow atrophy group, but not in any of the 8 subjects in the rapid atrophy group. Among these SNVs, the variant that accounted for the greatest group difference and was predicted in silico as a “functional” missense variant was rs3796529 (REST; RE1-Silencing Transcription factor, Chr. 4q12). Quantitative trait analysis including whole-brain search in the remaining ADNI APOE-ε3/ε3 group (N=315) showed that genetic variation within REST was significantly associated with hippocampal volume at baseline (Table 1 and Fig. 1). Meta-analysis across 5 independent cross-sectional cohorts indicated that rs3796529 is significantly associated with hippocampal volume in APOE-ε3/ε3 individuals (N=923) (Fig. 2). The minor allele of rs3796529 confers a protective effect on hippocampal volume loss. REST, also known as neuron-restrictive silencer factor, is a master negative transcriptional regulator of adult hippocampal neurogenesis (Gao 2011) that recruits chromatin-modifying enzymes indicating a potential epigenetic role. These findings warrant further investigation and illustrate the potential of next generation sequencing combined with quantitative imaging phenotypes for discovery of disease mechanisms and novel therapeutic targets. Association of WES-identified variant (rs3796529) with brain structure using surface- based analysis. rs3796529 demonstrated a positive effect on cortical thickness at baseline. Statistical maps computed using SurfStat were thresholded using random field theory (RFT) at P-corrected < 0.05. Left figure is left hemisphere. Plots of meta-analysis results for rs3796529 using 5 independent cohorts: The initial phase of ADNI (ADNI-1), its subsequent phases (ADNI-GO and ADNI-2), the AddNeuroMed Consortium, Multi-Institutional Research on Alzheimer Genetic Epidemiology (MIRAGE) Study, Indiana Memory and Aging Study (IMAS).

Increasing evidence suggests that inflammation is one pathophysiological mechanism in Alzheimer’s disease (AD). Recent studies have identified an association between the poly (ADP-ribose) polymerase 1 (PARP1) gene and AD. This gene encodes a protein that is involved in many biological functions, including DNA repair and chromatin remodeling, and is a mediator of inflammation. Therefore, we performed a targeted genetic association analysis to investigate the relationship between the PARP1 polymorphisms and brain microglial activity as indexed by [11C]PBR28 positron emission tomography (PET). Participants were 26 non-Hispanic Caucasians in the Indiana Memory and Aging Study (IMAS). PET data were intensity-normalized by injected dose/total body weight. Average [11C]PBR28 standardized uptake values (SUV) from 6 bilateral regions of interest (thalamus, frontal, parietal, temporal, and cingulate cortices, and whole brain gray matter) were used as endophenotypes. Single nucleotide polymorphisms (SNPs) with 20 ε4 status as a covariate. Associations with [11C]PBR28 SUVs from thalamus and cingulate were significant at corrected p<0.014 and <0.065, respectively. Subsequent multi-marker analysis with cingulate [11C]PBR28 SUV showed that individuals with the “C” allele at rs6677172 and “A” allele at rs61835377 had higher [11C]PBR28 SUV than individuals without these alleles (corrected P<0.03), and individuals with the “G” allele at rs6677172 and “G” allele at rs61835377 displayed the opposite trend (corrected P<0.065). A previous study with the same cohort showed an inverse relationship between [11C]PBR28 SUV and brain atrophy at a follow-up visit, suggesting possible protective effect of microglial activity against cortical atrophy. Interestingly, all 6 AD and 2 of 3 LMCI participants in the current analysis had one or more copies of the “GG” allele combination, associated with lower cingulate [11C]PBR28 SUV, suggesting that this gene variant warrants further investigation.

BACKGROUND:Apolipoprotein E (APOE) ε4 allele's role as a modulator of the relationship between soluble plasma amyloid beta (Aβ) and fibrillar brain Aβ measured by Pittsburgh compound B positron emission tomography ([(11)C]PiB PET) has not been assessed. METHODS:Ninety-six Alzheimer's Disease Neuroimaging Initiative participants with [(11)C]PiB scans and plasma Aβ1-40 and Aβ1-42 measurements at the time of PET scanning were included. Regional and voxelwise analyses of [(11)C]PiB data were used to determine the influence of APOE ε4 allele on association of plasma Aβ1-40, Aβ1-42, and Aβ1-40/Aβ1-42 with [(11)C]PiB uptake. RESULTS:In APOE ε4- but not ε4+ participants, positive relationships between plasma Aβ1-40/Aβ1-42 and [(11)C]PiB uptake were observed. Modeling the interaction of APOE and plasma Aβ1-40/Aβ1-42 improved the explained variance in [(11)C]PiB binding compared with using APOE and plasma Aβ1-40/Aβ1-42 as separate terms. CONCLUSIONS:The results suggest that plasma Aβ is a potential Alzheimer's disease biomarker and highlight the importance of genetic variation in interpretation of plasma Aβ levels.

Proteins, widely studied as potential biomarkers, play important roles in numerous physiological functions and diseases. Genetic variation may modulate corresponding protein levels and point to the role of these variants in disease pathophysiology. Effects of individual single nucleotide polymorphisms (SNPs) within a gene were analyzed for corresponding plasma protein levels using genome-wide association study (GWAS) genotype data and proteomic panel data with 132 quality-controlled analytes from 521 Caucasian participants in the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort. Linear regression analysis detected 112 significant (Bonferroni threshold p=2.44×10(-5)) associations between 27 analytes and 112 SNPs. 107 out of these 112 associations were tested in the Indiana Memory and Aging Study (IMAS) cohort for replication and 50 associations were replicated at uncorrected p<0.05 in the same direction of effect as those in the ADNI. We identified multiple novel associations including the association of rs7517126 with plasma complement factor H-related protein 1 (CFHR1) level at p<1.46×10(-60), accounting for 40 percent of total variation of the protein level. We serendipitously found the association of rs6677604 with the same protein at p<9.29×10(-112). Although these two SNPs were not in the strong linkage disequilibrium, 61 percent of total variation of CFHR1 was accounted for by rs6677604 without additional variation by rs7517126 when both SNPs were tested together. 78 other SNP-protein associations in the ADNI sample exceeded genome-wide significance (5×10(-8)). Our results confirmed previously identified gene-protein associations for interleukin-6 receptor, chemokine CC-4, angiotensin-converting enzyme, and angiotensinogen, although the direction of effect was reversed in some cases. This study is among the first analyses of gene-protein product relationships integrating multiplex-panel proteomics and targeted genes extracted from a GWAS array. With intensive searches taking place for proteomic biomarkers for many diseases, the role of genetic variation takes on new importance and should be considered in interpretation of proteomic results.

Alzheimer's disease (AD) is characterized by gradual neurodegeneration and loss of brain function, especially for memory during early stages. Regression analysis has been widely applied to AD research to relate clinical and biomarker data such as predicting cognitive outcomes from MRI measures. In particular, sparse models have been proposed to identify the optimal imaging markers with high prediction power. However, the complex relationship among imaging markers are often overlooked or simplified in the existing methods. To address this issue, we present a new sparse learning method by introducing a novel network term to more flexibly model the relationship among imaging markers. The proposed algorithm is applied to the ADNI study for predicting cognitive outcomes using MRI scans. The effectiveness of our method is demonstrated by its improved prediction performance over several state-of-the-art competing methods and accurate identification of cognition-relevant imaging markers that are biologically meaningful.

Amyloid imaging with [11 C]Pittsburgh Compound-B (PiB) provides in vivo data on plaque deposition in those with, or at risk for, Alzheimer's disease (AD). We performed a gene-based association analysis of 15 quality-controlled amyloid-pathway associated candidate genes in 103 Alzheimer's Disease Neuroimaging Initiative participants. The mean normalized PiB uptake value across four brain regions known to have amyloid deposition in AD was used as a quantitative phenotype. The minor allele of an intronic SNP within DHCR24 was identified and associated with a lower average PiB uptake. Further investigation at whole-brain voxel-wise level indicated that non-carriers of the minor allele had higher PiB uptake in frontal regions compared to carriers. DHCR24 has been previously shown to confer resistance against beta-amyloid and oxidative stress-induced apoptosis, thus our findings support a neuroprotective role. Pathway-based genetic analysis of targeted molecular imaging phenotypes appears promising to help elucidate disease pathophysiology and identify potential therapeutic targets.

Large-scale genome-wide association studies have identified common variants associated with late-onset Alzheimer's disease (LOAD). However, a substantial portion of LOAD heritability remains unexplained by known susceptibility genes, the most robust being the APOE E 4 allele. We performed a study to identify novel variants beyond APOE E 4 in protein-coding regions associated with rate of hippocampal neurodegeneration in APOE E 3/E 3 patients with amnestic mild cognitive impairment (MCI). In the discovery phase, whole-exome sequencing was performed on blood-derived DNA samples from 16 non-Hispanic Caucasian APOE E 3/E 3 males, with baseline diagnosis of MCI from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Matched participants were selected at the extremes of the 2-year longitudinal change distribution of Freesurfer-derived hippocampal volume (8 slow/stable and 8 rapid decliners). Short-reads were mapped to the NCBI reference human genome (build 37.64) with the BWA aligner. GATK was used for single nucleotide variant (SNV) calling and filtering. In a validation phase, known variants (dbSNP135) identified from the discovery phase were imputed using MACH in 750 non-Hispanic Caucasian ADNI participants, and were assessed in relation to selected neuroimaging phenotypes from 314 APOE E 3/E 3 participants, excluding those used in sequencing. We identified 90,386 targeted SNVs, of which 6,276 were novel (dbSNP135). Analyses focused on 50,903 exonic SNVs to identify variants associated with rate of hippocampal volume loss. 57 non-synonymous variants were found exclusively in more than 4 of 8 rapid decliners but not 8 slow decliners. Two SNVs in PARP1 (Poly (ADP-ribose) polymerase-1) and CARD10 (Caspase Recruitment Domain) were each detected in 6 of 8 rapid decliners and predicted as “probably damaging” missense variants. These two SNVs were significantly associated with neuroimaging phenotypes in the analysis of the remaining ADNI APOE E 3/E 3 group. Two SNVs in PARP1 and CARD10 were associated with rate of hippocampal neurodegeneration in APOE E 3/E 3 participants. PARP1 overactivation by oxidative stress has been associated with cognitive impairment 1. NF-kappaB pathway through CARD10 has been implicated in AD 2. Combining next generation sequencing and quantitative imaging phenotypes holds promise for discovery of rare and novel variants involved in neurodegeneration. Reference: 1. Strosznajder et al. (2012)2. Wang et al. (2001)

Copy number variants (CNVs) are DNA regions that have gains (duplications) or losses (deletions) of genetic material. CNVs may encompass a single gene or multiple genes and can affect their function. They are hypothesized to play an important role in certain diseases. We previously examined the role of CNVs in late-onset Alzheimer's disease (AD) and mild cognitive impairment (MCI) using participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study and identified gene regions overlapped by CNVs only in cases (AD and/or MCI) but not in controls. Using a similar approach as ADNI, we investigated the role of CNVs using 794 AD and 196 neurologically evaluated control non-Hispanic Caucasian NIA-LOAD/NCRAD Family Study participants with DNA derived from blood/brain tissue. The controls had no family history of AD and were unrelated to AD participants. CNV calls were generated and analyzed after detailed quality review. 711 AD cases and 171 controls who passed all quality thresholds were included in case/control association analyses, focusing on candidate gene and genome-wide approaches. We identified genes overlapped by CNV calls only in AD cases but not controls. A trend for lower CNV call rate was observed for deletions as well as duplications in cases compared to controls. Gene-based association analyses confirmed previous findings in the ADNI study (ATXN1, HLA-DPB1, RELN, DOPEY2, GSTT1, CHRFAM7A, ERBB4, NRXN1) and identified a new gene (IMMP2L) that may play a role in AD susceptibility. Replication in independent samples as well as further analyses of these gene regions is warranted.

Genetic mapping of hippocampal shape, an under-explored area, has strong potential as a neurodegeneration biomarker for AD and MCI. This study investigates the genetic effects of top candidate single nucleotide polymorphisms (SNPs) on hippocampal shape features as quantitative traits (QTs) in a large cohort. FS+LDDMM was used to segment hippocampal surfaces from MRI scans and shape features were extracted after surface registration. Elastic net (EN) and sparse canonical correlation analysis (SCCA) were proposed to examine SNP-QT associations, and compared with multiple regression (MR). Although similar in power, EN yielded substantially fewer predictors than MR. Detailed surface mapping of global and localized genetic effects were identified by MR and EN to reveal multi-SNP-single-QT relationships, and by SCCA to discover multi-SNP-multi-QT associations. Shape analysis identified stronger SNP-QT correlations than volume analysis. Sparse multivariate models have greater power to reveal complex SNP-QT relationships. Genetic analysis of quantitative shape features has considerable potential for enhancing mechanistic understanding of complex disorders like AD.

Multi-modal neuroimaging and biomarker data provide exciting opportunities to enhance our understanding of phenotypic characteristics associated with complex disorders. This study focuses on integrative analysis of structural MRI data and proteomic data from an RBM panel to examine their predictive power and identify relevant biomarkers in a large MCI/AD cohort. MRI data included volume and thickness measures of 98 regions estimated by FreeSurfer. RBM data included 146 proteomic analytes extracted from plasma and serum. A sparse learning model, elastic net logistic regression, was proposed to classify AD and MCI, and select disease-relevant biomarkers. A linear support vector machine coupled with feature selection was employed for comparison. Combining RBM and MRI data yielded improved prediction rates: HC vs AD (91.9%), HC vs MCI (90.5%) and MCI vs AD (86.5%). Elastic net identified a small set of meaningful imaging and proteomic biomarkers. The elastic net has great power to optimize the sparsity of feature selection while maintaining high predictive power. Its application to multi-modal imaging and biomarker data has considerable potential for discovering biomarkers and enhancing mechanistic understanding of AD and MCI.