Vince D Calhoun - US grants

DESCRIPTION (provided by applicant): Each brain imaging modality reports on a different aspect of the brain (e.g. gray matter integrity, blood flow changes, white matter integrity) and each has strengths and weaknesses. However the vast majority of studies analyzes each imaging modality separately and interprets the results independently of one another. Many mental illnesses, such as schizophrenia, bipolar disorder, depression, and others, currently lack definitive biological markers and rely primarily on symptom assessments for diagnosis. One area which can benefit greatly from the combination of multimodal data is the study of schizophrenia. The brain imaging findings in schizophrenia are widespread and heterogeneous and have limited replicability. We show evidence that, in part, the lack of consistent findings is because most models do not combine imaging modalities in an integrated manner and miss important changes which are partially detected by each modality separately. Combining multimodal imaging data is not easy since, among other reasons, the combination of multiple data sets consisting of thousands of voxels or time points yields a very high dimensional problem, requiring appropriate data reduction strategies. There are two important areas that we focus on in this new phase of the project. First, we will focus on developing data fusion strategies that will leverage our initial success in developing ICA-based tools for combining multiple tasks and modalities. We will develop and validate approaches which can scale easily from one to many different data types. In the first funding period we focused mainly on pair-wise combinations of multimodal data. However, the results have convinced us that allowing higher order relationships is also important (e.g. we show pilot data in which using 3-way relationships improves our ability to discriminate schizophrenia and control groups). In this proposal we will significantly expand this work and develop novel methods to efficiently exploit high-order joint information not just pair-wise. Next, we will develop new tools that will identify correspondences among modalities. For example we show that structural and functional patterns of covariation are in some cases remarkably similar to one another and in others cases quite distinct and these relationships can predict diagnosis. We will thoroughly test our approach using a well characterized data set involving multiple illnesses that have overlapping symptoms and which can sometimes be misdiagnosed and treated with the wrong medications for months or years (schizophrenia, bipolar disorder, and unipolar depression). As before, we will provide open source tools and release data throughout the duration of the project via a web portal and the NITRIC repository, hence enabling other investigators to compare their own methods with our own as well as to apply them to a large variety of brain disorders. 36

DESCRIPTION (provided by applicant): Data driven methods are being increasingly used to analyze brain imaging data. FMRI analyses can be put on an analytic spectrum with heavily model-based approaches (like the general linear model (GLM) implemented in the SPM software) on one end and flexible data-driven approaches like independent component analysis (ICA), principal component analysis (PCA), or clustering on the other end. In between there is a gap, which we and others have been trying to fill. In particular, methods such as ICA are particularly useful for reducing the multivariate fMRI problem down to one that is both tractable and also enables the incorporation of prior information. In the first period of this competing renewal, we focused our efforts upon developing ICA of fMRI methods which would be suitable for making group inferences, and which would allow the incorporation of prior information, hence moving from a 'blind'ICA approach to a semi-blind ICA approach. Despite the progress we have made, there is still considerable work to be done in the analysis of fMRI data with ICA. In this competing renewal, we propose to continue and significantly expand this work. First, we will extend our semi-blind ICA (sbICA) framework to provide a general framework for incorporating prior information from multiple spatial and temporal sources. In the second aim we will focus upon statistical inference and develop a framework for integrating the relevant functional components. In the third aim, we will validate the algorithms in aims 1 and 2, including using fMRI data collected on multiple days from a variety of paradigms. In this aim we develop a decision mechanism for selecting the best combination of methods given a particular problem. For the fourth aim, we will apply our methods to data collected during four well-studied paradigms in healthy controls and patients with schizophrenia. Our final aim involves the continuing development of our GIFT toolbox, and incorporation of the above algorithms, constraint selection mechanisms, and visual interfaces into the software. The successful completion of this research will provide a powerful set of tools for the research community to increase the sensitivity and specificity of BOLD analysis methods by drawing upon the strengths of both model-based and data-driven approaches. These tools will also provide a way to study the inter-relationship among functional networks in a flexible manner. This has application not only in schizophrenia but in many other diseases such as Alzheimer's, attention deficit hyperactivity, and psychopathy.

This Center for Biomedical Research Excellence (COBRE) will examine the neural mechanisms of schizophrenia by integrating multiple neuroimaging methods with psychiatric and neuropsychological testing, and incorporating genetic testing. Its overarching theme is the study of schizophrenia as a disorder characterized by abnormalities in structural, functional, and effective connectivity between cortical and subcortical brain regions producing abnormalities in the integration of information across distributed brain circuits. The program is composed of four tightly integrated projects conceptualized as a hierarchy in which each independently investigates a major cognitive domain of dysfunction in schizophrenia, as identified by a panel of experts in a recent NIH-sponsored study. This dysfunction ranges from basic sensory to higherorder deficits, with attention, memory, concept formation and problem solving abilities (ie, intelligence) listed among the top cognitive deficits that detrimentally effected patients with schizophrenia. The plan begins at a basic level of sensory processing (auditory sensory gating; Proj.1), followed by multi-sensory integration (auditory and visual; Proj.2), to working memory and relational memory integration (transverse patterning; Proj.3), and, finally, generalized higher cognitive functioning (intelligence; Proj.4). Plans provide for data collection on up to 100 of the same patients with schizophrenia (SP) and 100 healthy normal volunteers (HNV) and a centralized data processing stream that has been implemented and is already in use. Proj.1 quantifies brain function and clinical pathology through multimodal imaging of sensory gating. Proj.2 studies the neural mechanisms underlying auditory and visual integration in SP and HNV using magnetoencephalography (MEG), electroencephalography (EEG) combined with anatomical magnetic resonance imaging (MRI), and functional MRI (fMRI). Proj.3 tests the fronto-temporal disconnection hypothesis in schizophrenia by addressing basic clinical and translational research questions. Proj.4 addresses whether general cognitive functioning in schizophrenia is related to particular white matter, metabolic, and volumetric changes in subcortical gray- and white-matter regions suggestive of frontosubcortical disconnection. These projects will produce a wealth of information about the nature of antomic and functional misconnections in schizophrenia and how they relate to the manifestation of the illness.

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. The unifying theme for our COBRE is the study of schizophrenia as a disorder characterized by abnormalities in structural, functional, or effective connectivity between cortical and subcortical brain regions, producing abnormalities in integration within distributed brain circuits. The resulting information will be used to better understand the neurophysiological, neuroanatomical and neurochemical underpinnings of the abnormalities associated with schizophrenia. This COBRE will combine a wide variety of imaging tools relevant to the study of neurotransmitters and tissue abnormalities [i.e., magnetic resonance spectroscopy (MRS)], along with measures of brain anatomy, electromagnetic activity and hemodynamics [structural MRI (sMRI), diffusion tensor imaging (DTI), magnetoencephalography (MEG), electroencephalography (EEG), and functional MRI (fMRI)], combined with psychiatric and behavioral measures. Each of the 4 projects will acquire and analyze data using at least two of these imaging techniques. The Image Analysis Core is responsible for: 1) providing a common set of image analysis tools needed to accomplish the scientific goals of the 4 projects; 2) providing staff and consultants experienced in relevant image analysis to assist project personnel in organizing and conducting the required analyses;and 3) providing education and training for Pis and other project personnel to accomplish the Specific Aims of the COBRE projects. The Imaging Cores have been separated into two different cores, Image Data Acquisition (IDA) and Image Analysis (IA), in order to more adequately deal with the complexity of issues necessary for these two phases of each project. For example, pulse sequence details necessary for sMRI, DTI, MRS, and fMRI is incorporated into the IDA Core while modeling and analysis issues and multimodal integration are included in the IA Core. Storage of imaging data (which uses relational database strategies) and biostatistical approaches for testing the output measures is covered in the STATNI Core. Genetic and clinical assessment data are discussed in the ACAS core. While these various cores will be described separately in this application, there is already a high degree of communication and coordination among their members.

DESCRIPTION (provided by applicant): This application addresses broad Challenge Area (03) Biomarker Discovery and Validation and specific Challenge Topic 03-MH-101* Biomarkers in Mental Disorders. Specifically, we plan to use a combination of genetic and neuroimaging tools to identify novel biomarkers of clinical severity in patients with schizophrenia. Schizophrenia is a chronic and severely debilitating mental disorder affecting approximately 1% of the world's population. Multiple neurotransmitter systems have been implicated, as well as both gray and white matter abnormalities. These structural alterations are thought to underlie both synaptic miscommunication at local neuronal circuits and functional disconnectivity among distributed brain regions. Given the role of myelin in sub-serving rapid long-distance communication, it has been proposed that a disruption of oligodendrocyte function and myelin integrity may contribute to some of the symptoms of this illness. Supporting this idea, an increasing number of neuropathological, neuroimaging and molecular genetic studies demonstrate the presence of white matter pathology in patients with schizophrenia. Although schizophrenia is not a dysmyelinating disorder, it is important to note that: a) the onset of symptoms usually coincides with the peak of myelination in the frontal and temporal lobes, b) patients with schizophrenia often show an impaired age- related increase in white matter volumes in the these brain regions and c) specific disruption of myelin structure during this critical period is often associated with schizophrenia-like symptoms. Over 90 articles in the past 5 years have used diffusion tensor imaging (DTI) to characterize white matter abnormalities in chronic and first episode patients. The studies demonstrated myelin integrity defects in the subcortical white matter, particularly in the frontal and temporal lobes. However given that these studies were performed using a small number of patients, and there were some discrepancies about the location and extent of white matter pathology identified, several questions remain about the prevalence of myelin pathology in the patient population. Furthermore, the genetic contributions to these alterations and the significance of white matter pathology to clinical severity remain to be established. As shown in the diagram above, in this challenge grant, we propose to assess influence of white matter alterations and genetic variation to the different symptoms of schizophrenia. The contributions of specific gene polymorphisms to measurements of white matter integrity will be evaluated using 500 patients and control subjects. These measurements will be correlated with disease severity using multivariate statistical methods developed by the PI. Specifically we plan to: Aim 1: Employ available DTI data and DNA samples, collected in 250 well-characterized schizophrenia patients and healthy controls, to identify the putative genetic underpinnings of white matter tract abnormalities and correlate these with several measurements of clinical severity. The data and samples have been collected as part of previously and currently funded studies at two sites: The Mind Research Network (MRN) and the Olin Neuropsychiatry Research Center (ONRC). Aim 2: Use additional data collected at both sites (N=250) to perform a confirmatory analysis validating the observations made under Aim 1. We will also release a set of software tools to the community. Why is a Challenge grant mechanism ideal for the proposed research? 1) Our goal of using innovative approaches to identify candidate biomarkers for mental disorders that are suitable for subsequent validation efforts matches the goals of this RFA. The proposed use of genetic, neuroimaging and statistical tools also matches the technological approaches described in the RFA and represents a new direction in the field. There are currently no reliable biomarkers for schizophrenia, so the proposed search for biomarkers that can predict disease severity is of high impact. 2) A two year grant award is ideal for the proposed work. We have already acquired most of the MRI data and collected saliva samples as part of other NIH funded studies that used different imaging modalities (fMRI and EEG) in the same groups of patients. Therefore, the project will focus on the genotyping and DTI analyses and the statistical methods to search for specific biomarkers. 3) We have assembled a team of investigators with unique expertise and an excellent record of effective past collaborations to pursue these studies. Our plan to hire and train new personnel, and to employ unique genome wide and bioinformatics technologies using US-based companies such as Illumina, Inc., will have the added benefit of stimulating the economy. PUBLIC HEALTH RELEVANCE: The goal of this Challenge grant application is to identify novel biomarkers of clinical severity in patients with schizophrenia. There are currently no reliable biomarkers for schizophrenia, so the proposed use of sophisticated genotyping, neuroimaging and biostatistical tools for searching biomarkers that can predict disease severity in two large cohorts of patient has a high clinical impact. The identification of such biomarkers will not only increase our knowledge of the pathophysiology of schizophrenia but also, and most importantly, may help predict an increased risk for this illness even before the onset of symptoms.

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. The Image Data Acquisition (IDA) CORE provides state-of-the art protocols and technology which can support a large clinical study and will be applied to the individual projects. Specifically, we highlight some of the contributions and advances that our IDA CORE has made. Two developments will be highlighted here. First, Dr. Charles Gasparovic is responsible for the development of spectroscopic MRI methods which support Project 4 (Jung, PI). Second, Dr. Cheryl Aine is responsible for developments in MEG/EEG methods which are utilized Projects 1-3 (Mayer, Stephen, Hanlon, PIs).

DESCRIPTION (provided by applicant): The economic impact of substance and schizophrenia amounts to hundreds of billions of dollars, not to mention the countless lives impacted both directly and indirectly. Schizophrenia (SZ) and substance use disorders (SUD) are both extremely complex, both with substantial genetic and environmental components and with some shared aspects. In addition, schizophrenia patients tend to have increased levels of substance use which further complicates our understanding of the diagnosis. Most studies of SZ and SUD which incorporate imag ing and genetic data still ignore most of the information provided by the data by only analyzing a small number of genetic factors or brain regions. To characterize the available information, we are in need of approaches that can deal with high-dimensional data exhibiting interactions at multiple levels, while providing interpretable solu- tions. In the previous funding period we developed methods for pairwise coupling of high-dimensional genetic and imaging data which provided a powerful way to analyze the full information in joint data sets. However this is just the tip of the iceberg because in order to understand the complex interchange of biological pathways, brain networks, and behavior we need approaches that can handle more than two types of data. In this pro- posal we will focus on three key areas. First-building on our previous successes-we will develop new meth- ods that can robustly capture complex relationships between multiple types of data (e.g. genetic codes -> methylation -> brain function -> behavior). Then, we will develop new approaches for the effective use of reliable prior information and provide a set of methods that optimally balance between prior information (model- based) and information readily available from the data (data-driven). And finally, we will combine the strengths of two domains of research, the tractability of data-driven decompositions such as independent component analysis (ICA) with the flexibility of multi-layered learning, to develop an approach we call deep independence networks. This will allow us to capture indirect, but important, relationships among modalities, while also taking advantage of the full data available. The methods we develop will provide a very desirable framework allowing investigators to infer relations in high dimensional data and will provide a much needed set of data analysis tools to the community. We will continue to focus on two important applications where integrating such data is especially important, schizophrenia and addiction, which also share some comorbidity. Focusing on two differ- ent disorders will help us to further generalize the algorithms developed and evaluate shared and distinct as- pects of these two disorders. By combining 1) the extensive data made available by our collaborators, 2) de- velopment of computational approaches for fusing high dimensional data, and 3) the conceptual models we have developed for schizophrenia and alcohol use disorder and results from ongoing studies, we are poised to fill an important gap in the field and produce new tools that have applicability to a wide variety of diseases.

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. Primary support for the subproject and the subproject's principal investigator may have been provided by other sources, including other NIH sources. The Total Cost listed for the subproject likely represents the estimated amount of Center infrastructure utilized by the subproject, not direct funding provided by the NCRR grant to the subproject or subproject staff. The Biostatistics &Neurolnformatics Core (STATNI) serves as a centralized resource for biostatistical consulting for all scientific projects proposed in this application. The unit provides support to each of the four research projects at all levels of investigation, beginning with the formulation of more specific hypotheses related to the overall unifying hypotheses, reviewing the design of studies, and evaluating the utility of measurement techniques. Support and consultation concludes with aiding in the interpretation, presentation, and final writing of results. The core also serves as a resource for data management and quality control through a set of comprehensive neuroinformatics tools.

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. Primary support for the subproject and the subproject's principal investigator may have been provided by other sources, including other NIH sources. The Total Cost listed for the subproject likely represents the estimated amount of Center infrastructure utilized by the subproject, not direct funding provided by the NCRR grant to the subproject or subproject staff. ACAS CORE Goals: The primary goal of the ACAS CORE is to serve as a central resource for clinical and neuropsychological data collection, as well as recruit all subjects and ensure clinical stability during the testing period. The ACAS CORE is tasked to develop the infrastructure and collection of genomic samples to be used by current and also future projects. The ACAS CORE is utilized by all four of the projects in the COBRE for recruitment, diagnosis, clinical stabilization, protocol coordination and management, support for outreach activities to the community, coordination of the activities of trainees and junior investigators, and a variety of other Center functions. It is also the responsibility of the ACAS CORE to implement the organizational plan outlined in the Research Plan.

Project Summary. Current studies are using genome wide scan (GWS) approaches to identify the numerous genes which might play a role in increased risk for psychosis. Structural neuroimaging measures implicate gray matter loss in schizophrenia; a pattern of regional loss in the medial frontal, temporal and insular gyri have been identified by us and others in schizophrenia. However, the recognition that case/control approaches are not perhaps the most useful, has led to an emphasis on cognitive constructs such as attention, memory, language, as in the Research Domain Criteria (RDoC) matrix, to identify cross-diagnostic mechanisms. This has left the psychotic symptoms per se without a clear connection to the neuroanatomical circuits and genetic mechanisms. Identifying the relationships among patterns of gray matter reduction, symptom co-occurrence patterns, and genetic profiles which exist across schizophrenia and bipolar disorder is the goal of this project. We propose a multivariate method for analyzing already existing GWS data, voxelwise measures of gray matter density, and symptom measures from an aggregated dataset of over 4000 individuals with diagnoses from the schizophrenia and bipolar spectrum. We will apply three way parallel ICA, with reference; this technique identifies patterns of spatial variation in the brain structure, symptom profiles, and patterns of genotypes which are linked. We begin with over 7,000 structural imaging, symptom scores, and GWS samples from cases and controls, from aggregated legacy data. We constrain the imaging and genetic analyses with reference vectors to incorporate a priori information. In Aims 1 and 2 we will develop initial a priori spatial patterns, structural networks using source- based morphometry methods, both alone and in conjunction with symptom measures; in Aim 3 we will determine the heritability and quantitative trait loci for these networks in independent family samples; in Aim 4 we use the quantitative trait loci as a priori constraints on the genetic data, and the heritable structural networks as constraints on the imaging data on our three-way parallel ICA analysis. We include a split-half analysis for replication and a follow-up high-density genotyping plan. Using these methods, we will determine the spatial patterns and genetic profiles that covary within our sample, and which show relationships with symptom profiles across schizophrenia and bipolar disorder; this forms the basis for linking the symptoms to the brain circuits and genetics ?units of analysis?. Using higher-order clustering on the identified patterns, we identify coherent sub- groupings of subjects using the genetics, brain structure, and symptom measures within the larger data matrix. The final results will be the combinations of genotypic networks which influence the patterns of structural brain effects in conjunction with variations in symptom clusters, refining the diagnostic categories based on biological evidence.

The algorithm and data analysis (ADA) core of this phase II COBRE will fulfill the need for centralized image analysis resources that will be used to support all five projects. These resources include tools designed for measurement and analysis of sMRI, MRS, fMRI, DTI, genetics, EEG and MEG data. The ADA Core will play a leading role in developing and providing software that is needed to solve basic image analysis problems that arise when working with MR and MEG/EEG data. This will be accomplished by providing a core set of tools and approaches for analysis of imaging and genetic data. The core set of resources includes expertise and tools for analyzing all first level-imaging data (automated pipeline preprocessing) as well as advanced algorithms for network-based functional and structural connectivity measures to address in a comprehensive way the scientific questions being asked in each of the projects. We will work with the tools developed locally as well as widely-used tools developed by other groups to enable network-based analysis, data-fusion of multimodal data, and prediction/classification approaches. Importantly, a key aspect of this COBRE and the ADA core is focused on combining multimodal data as each project will work with two or more modalities. An additional area of emphasis will be on the development of realistic simulation approaches, to enable comparisons of algorithms, optimization of parameters, and to provide intuition about how new algorithms work. Finally, the ADA core will also provide essential training to junior investigators about data analysis of brain imaging and genetic data. This will ensure junior investigators are informed about the various algorithms, understand how to make analysis choices given a particular hypothesis, and have a basic idea of how to implement such algorithms themselves. The director of the ADA Core is Dr. Calhoun, who has over 20 years of experience in developing tools and approaches for working with unimodal and multimodal imaging and genetics data. Codirector Dr. Cheryl Aine has extensive experience in unimodal and multimodal imaging with MEG/EEG and codirector Dr. Julia Stephen, a graduate of the phase I COBRE, is currently the director of the MEG facility at MRN and has considerable experience in combining MEG and fMRI data, as well as EEG and MEG data in clinical groups.

DESCRIPTION (provided by applicant): This Phase II COBRE project is a natural extension of our Phase I COBRE on multimodal neuroimaging in schizophrenia. In the next evolution we will build on our success in Phase I to include a wider range of disease categories making the overarching theme of the Phase II COBRE the use of multimodal neuroimaging to better understand the neural mechanisms of psychosis and mood disorders. The Mind Research Network (MRN) houses an Elekta MEG System, a high density EEG lab, and a 3T Siemens Trio MRl scanner. Additional resources include a centralized neuroinformatics system, a strong IT management plan, and state-of-the-art image analysis tools. The Phase II COBRE will provide support to five outstanding junior investigators through the assistance of strong senior mentors. The five projects each focus on distinct, but related, aspects of psychosis and mood disorders. Project 1 will utilize advanced data fusion methods to evaluate the ability of multimodal brain imaging data to differentiate patient groups and to push beyond discrete diagnostic categories by identifying individuals in intermediate positions on the continuum. Project 2 is an expansion of the pilot genetic program from the Phase I to evaluate the shared and unique aspects of genetic influence on brain structural networks using advanced multivariate methods. Project 3 will focus on the lens of social cognition and evaluate functional networks in patients while perceiving facial and vocal emotions. The ability of both structural and functional networks to differentiate groups and predict outcomes will be evaluated. Project 4 will focus on auditory hallucinations using MEG and fMRI. Evaluation of the ability to predict hallucinations from the imaging data as well as the impact of transcranial direct current stimulation (tDCS) on the identified brain networks will be investigated. And finally, Project 5 will use a longitudinal desin to study brain networks related to major depression and relapse after treatment with electro-convulsive therapy (ECT). We will continue with the cores established during the Phase I project including administration, clinical assessment, and mentoring (ACAM), multimodal data acquisition (MDA), algorithm and data analysis (ADA), and biostatistics and neuro-informatics (BNI). These cores have begun to serve MRN and the greater community, as well as other institutions including extension collaborations with IDeA funded projects in New Mexico and other states. A highly successful pilot project program will be continued. We believe this Phase II COBRE is extremely well-positioned to establish New Mexico as one of the premier brain imaging sites. We include an extensive educational, mentoring, and faculty development program to carefully mentor and establish junior investigators as independently funded investigators, thus fulfilling the ultimate goals of the COBRE program.

The administrative core (ACAM) will function in an overall supervisory capacity for the COBRE. It is comprised of both internal and external advisory boards, as well as the mentors and core directors. The ACAM core will be directed by Dr. Vince Calhoun, the PI for the current proposal. It will also operate under the supervision of two well-respected psychiatric clinicians: Drs. Juan Bustillo and Jose Canive. Dr. Perrone-Bizzozero, an established neuroscientist and educator, will coordinate the educational plan. Drs. Jeff Lieberman, MD and Charles Bowden, MD will serve as consultants. The core director will coordinate the duties and goals of the ACAM members, focused on three aims: 1) to coordinate all the specific budgetary, regulatory and personnel issues involved in the four cores and five individual projects; 2) to recruit subjects for projects 3 to 5 and manage the acquisition of all clinical, neuropsychological, genetic and demographic data; and 3) to implement a faculty development plan for the junior Pis, facilitate their interactions with internal and external mentors, and monitor the specific milestones of the mentoring plans. A long-term aim of this core is to expand the capability of our facilities and further develop a diversified neuroimaging research environment that will continue to be competitive nationally and internationally. This will strengthen the mentorship opportunities for junior investigators, post-doctoral fellows and graduate students who work in the New Mexico environment. We believe that through the support of the COBRE phase II, our cohesive and collaborative program of interdisciplinary and translational research can be leveraged into a P50-level center grant which will be submitted approximately 3 years after the start of the Phase II award. Hence, this Phase II core is essential for the overall success of our New Mexico program on multimodal imaging of psychotic and mood disorders. PHS

DESCRIPTION (provided by applicant): The PREDICT-HD study has collected an impressive sample of healthy subjects with the marker for Huntington's Disease (HD), and a sample of controls without that marker. The analyses of cognitive, psychiatric, and motor function over time in this sample has provided evidence for a prodromal phase preceding clinical diagnosis of HD: In many of the domains being measured, subjects who are more than 15 years away from their predicted age of onset show little or no difference from controls, while subjects in the 9- 1 year window are already showing significant if subtle declines, and within 9 years of the predicted age of onset are showing large losses. While this is an impactful clarification of the prodromal phase of HD, it needs to be further clarified; the predicted age of onset as calculated by the number of CAG repeats in the Huntington gene is very precise when the number of repeats is high, but can lead to a very large window of several decades when the number is low. In this ancillary study, we apply multivariate techniques such as parallel independent components analysis (pICA) to the combined structural and genetic imaging data from the PREDICT sample. In Aim 1, using a cross-sectional technique we will identify the genetic profiles which covary with disease-related patterns of gray matter loss. In Aim 2, using a longitudinal sample we will identify the brain structure and genetic profiles which correlate with loss of motor and cognitive function. In Aim 3, we will ensure that the PREDICT team is trained on these techniques and can apply them to its ongoing data collection, and that the results are incorporated into their data management system. The conclusion of this proposal places the effect of the CAG repeats within an initial context of genetic influences from the larger genome. We leverage the brain imaging measures to identify relevant profiles of genotypes within the HTT genetic network which accelerate or provide resilience to disease onset.

DESCRIPTION (provided by applicant): The goal of this project is to develop integrative approaches for the detection of biomarkers from multiscale genomic and imaging data, so that multiple mental illnesses such as schizophrenia (SC), Unipolar (UD) and bipolar (BI) disorder can be better identified. Imaging genetics is an emerging technique, which integrates imaging and genomic approaches to explore the association between genetic variations and brain functions and behaviors. Although it promises a better and more powerful approach for disease diagnosis and prognosis, the field is facing several major challenges: 1) First, most of current imaging genetics studies focus on pair-wise data correlation and integration; other important genetic factors such as epigenomics and genetic interactions (epistasis) have not been incorporated. 2) Second, multiscale imaging genetics data often exhibit specific characteristics such as inter- correlations, but this prior knowledge has not been incorporated into existing integrative models. 3) Finally, there is a high dimensionality problem with the analysis of imaging genetic data the number of sample is always significantly less than that of features. The solution of these problems necessitates a paradigm shift in computational models by considering the specific characteristics of these multiscale and multimodal data. Our multidisciplinary research team consisting of imaging scientist (Dr. Calhoun), statistical geneticist (Dr. Deng), biomedical engineer and bioimaging informatician (Dr. Wang), and psychiatrist (Dr. Pearson) has worked productively and creatively over the past few years in developing a number of data integration methods for fusion of imaging and genomic data. Building on our initial success, we will accomplish the following specific aims: 1) to study the correlation between multiple imaging and genomic data for the detection of epistasis factors or interaction networks; 2) to integrate multiscale imaging and genomic data, especially incorporating epistasis factors, for the identification of biomarkers, from which risk genes can be better detected; 3) to apply the detected biomarkers for the classification of multiple mental illnesses that are currently based on symptoms and are often misdiagnosed; and 4) to develop and disseminate an open source sparse model based data integration toolbox to the broad research community. The project will make significant impact on more accurate classification of clinically cryptic subgroups (e.g., SC, UD, BI) with an innovative and integrative paradigm by taking into account specific features of multiscale imaging genomic data and incorporation of prior knowledge. This will bring transformative changes on the current diagnosis of these mental illnesses (e.g., primarily based on imaging symptoms, which are often inaccurate), promising for personalized and optimal treatments. The developed methodology and tools are also applicable to many other neurological and psychiatric disorders. By the dissemination of the developed software tools to the research community, the project will have a broad and sustained impact.

Project Summary/Abstract The brain imaging community is greatly benefiting from extensive data sharing efforts currently underway5,10. However, there is a significant gap in existing strategies which focus on anonymized, post-hoc sharing of either 1) full raw or preprocessed data [in the case of open studies] or 2) manually computed summary measures [such as hippocampal volume11, in the case of closed (or not yet shared) studies] which we propose to address. Current approaches to data sharing often include significant logistical hurdles both for the investigator sharing the data as well as for the individual requesting the data (e.g. often times multiple data sharing agreements and approvals are required from US and international institutions). This needs to change, so that the scientific community be- comes a venue where data can be collected, managed, widely shared and analyzed while also opening up access to the (many) data sets which are not currently available (see recent overview on this from our group2). The large amount of existing data requires an approach that can analyze data in a distributed way while also leaving control of the source data with the individual investigator; this motivates a dynamic, decentralized way of approaching large scale analyses. We are proposing a peer-to-peer system called the Collaborative Informat- ics and Neuroimaging Suite Toolkit for Anonymous Computation (COINSTAC). The system will provide an inde- pendent, open, no-strings-attached tool that performs analysis on datasets distributed across different locations. Thus, the step of actually aggregating data can be avoided, while the strength of large-scale analyses can be retained. To achieve this, in Aim 1, the uniform data interfaces that we propose will make it easy to share and cooperate. Robust and novel quality assurance and replicability tools will also be incorporated. Collaboration and data sharing will be done through forming temporary (need and project-based) virtual clusters of studies performing automatically generated local computation on their respective data and aggregating statistics in global inference procedures. The communal organization will provide a continuous stream of large scale projects that can be formed and completed without the need of creating new rigid organizations or project-oriented stor- age vaults. In Aim 2, we develop, evaluate, and incorporate privacy-preserving algorithms to ensure that the data used are not re-identifiable even with multiple re-uses. We also will develop advanced distributed and pri- vacy preserving approaches for several key multivariate families of algorithms (general linear model, matrix fac- torization [e.g. independent component analysis], classification) to estimate intrinsic networks and perform data fusion. Finally, in Aim 3, we will demonstrate the utility of this approach in a proof of concept study through distributed analyses of substance abuse datasets across national and international venues with multiple imaging modalities. 4

? DESCRIPTION (provided by applicant): There has been great progress in the use of functional connectivity measures to study the healthy and dis- eased brain. The fMRI community has now realized that assessment of functional connectivity has been limited by an implicit assumption of spatial and temporal stationarity throughout the measurement period. Dynamics are potentially even more prominent in the resting-state, during which mental activity is unconstrained. There is a need for new methods to both estimate and quantify these changes. We propose to develop and compare a diverse but unified family of multivariate methods to address important aspects of dynamic connectivity that are not presently captured with existing approaches. Pilot data with initial approaches show robust changes in mental illness. Using a powerful framework that builds on the well-structured framework of joint blind source separation, we will make use of all available prior and statistical information-higher-order-statistics, sparsity, smoothness, sample and dataset dependence to derive a class of novel and effective dynamic models for full characterization of static and dynamic brain connectivity. We will validate these new methods while determining their properties and robustness to noise and other factors. We show preliminary work suggesting that there are important changes in dynamic properties that are not detectable in the static results and vice versa. Thus, we also propose models that can simultaneously capture stationary and non-stationary activity. We will apply our new set of methods to evaluate the common and distinct aspects of two patient groups (schizophrenia and bipolar disorder) as well as comorbid conditions (smoking and drinking). We will provide open source tools and release data throughout the duration of the project via a web portal and the NITRC repository, hence enabling other investigators to compare their own methods with our own as well as to apply them to a large variety of brain disorders. Our tools have wide application to the study of the healthy brain as well as many other diseases such as Alzheimer's and attention deficit hyperactivity disorder. 37

Project Summary. Current studies are using genome wide scan (GWS) approaches to identify the numerous genes which might play a role in increased risk for psychosis. Structural neuroimaging measures implicate gray matter loss in schizophrenia; a pattern of regional loss in the medial frontal, temporal and insular gyri have been identified by us and others in schizophrenia. However, the recognition that case/control approaches are not perhaps the most useful, has led to an emphasis on cognitive constructs such as attention, memory, language, as in the Research Domain Criteria (RDoC) matrix, to identify cross-diagnostic mechanisms. This has left the psychotic symptoms per se without a clear connection to the neuroanatomical circuits and genetic mechanisms. Identifying the relationships among patterns of gray matter reduction, symptom co-occurrence patterns, and genetic profiles which exist across schizophrenia and bipolar disorder is the goal of this project. We propose a multivariate method for analyzing already existing GWS data, voxelwise measures of gray matter density, and symptom measures from an aggregated dataset of over 4000 individuals with diagnoses from the schizophrenia and bipolar spectrum. We will apply three way parallel ICA, with reference; this technique identifies patterns of spatial variation in the brain structure, symptom profiles, and patterns of genotypes which are linked. We begin with over 7,000 structural imaging, symptom scores, and GWS samples from cases and controls, from aggregated legacy data. We constrain the imaging and genetic analyses with reference vectors to incorporate a priori information. In Aims 1 and 2 we will develop initial a priori spatial patterns, structural networks using source- based morphometry methods, both alone and in conjunction with symptom measures; in Aim 3 we will determine the heritability and quantitative trait loci for these networks in independent family samples; in Aim 4 we use the quantitative trait loci as a priori constraints on the genetic data, and the heritable structural networks as constraints on the imaging data on our three-way parallel ICA analysis. We include a split-half analysis for replication and a follow-up high-density genotyping plan. Using these methods, we will determine the spatial patterns and genetic profiles that covary within our sample, and which show relationships with symptom profiles across schizophrenia and bipolar disorder; this forms the basis for linking the symptoms to the brain circuits and genetics ?units of analysis?. Using higher-order clustering on the identified patterns, we identify coherent sub- groupings of subjects using the genetics, brain structure, and symptom measures within the larger data matrix. The final results will be the combinations of genotypic networks which influence the patterns of structural brain effects in conjunction with variations in symptom clusters, refining the diagnostic categories based on biological evidence.

Project Summary/Abstract The brain is extremely complex as we know, involving a complicated interplay between functional information interacting with a structural (but not static) substrate. Brain imaging technology provides a way to sample various aspects of the brain albeit incompletely, providing a rich set of multitask and multimodal information. The field has advanced significantly in its approach to multimodal data, as there are more studies correlating, e.g. func- tional and structural measures. However the vast majority of studies still ignore the joint information among two or more modalities or tasks. Such information is critical to consider as each brain imaging modality reports on a different aspect of the brain (e.g. gray matter integrity, blood flow changes, white matter integrity). The field is still striving to understand how to diagnose and treat complex mental illness, such as schizophrenia, bipolar disorder, depression, and others, and ignoring the joint information among tasks and modalities is to miss a critical, but available, part of the puzzle. Combining multimodal imaging data is not easy since, among other reasons, the combination of multiple data sets consisting of thousands of voxels or timepoints yields a very high dimensional problem, requiring appropriate data reduction strategies. In the previous phase of the project we developed approaches based on multiset canonical correlation analysis (mCCA) and joint independent compo- nent analysis (jICA) that can capture high-dimensional, linear, relationships among 2 or more modalities, and which we showed can identify both modality-unique and modality-common features that are predictive of dis- ease. In this new phase of the project we will focus on two important areas. First, we will build on our previous success by extending our models to allow for incorporation of behavioral/cognitive constraints as well as devel- oping new approaches which leverage recent advances in deep learning enabling us to capture higher order relationships embedded in multimodal and multitask data. Secondly, we will address the key challenge of inte- grating possibly thousands of multimodal features by developing a new meta-modality framework which will enable us to bring together the existing and new features in an intuitive manner. This will also enable us to capture changes in multimodal information which might not be harmful separately but which together are jointly sufficient to convey risk of illness or to identify information flow through the meta-modal space for developing potential targets for treatment. We will apply these approaches to one of the largest multimodal imaging datasets of psychosis and mood disorders. Our proposed approach will be thoroughly evaluated using this large data set which includes multiple illnesses that have overlapping symptoms and which can sometimes be misdiagnosed and treated with the wrong medications for months or years (schizophrenia, bipolar disorder, and unipolar de- pression). As before, we will provide open source tools and release data throughout the duration of the project via a web portal and the NITRIC repository, hence enabling other investigators to compare their own methods with our own as well as to apply them to a large variety of brain disorders. 36

Abstract We propose to acquire a ?next-generation? Magnetom Prisma 3 Tesla Magnetic Resonance Imaging System in support of engineering, neuroscience, and clinical research in New Mexico. The new scanner will be housed at the Mind Research Network (MRN), an independent, 501c3 not-for-profit research organization. MRN is a member of the Lovelace Family of companies, and it is located on the University of New Mexico campus in Albuquerque. The Prisma affords a 64 channel head and neck coil and the strongest gradients available for a 3T machine. The equipment has the potential to directly support the engineering and neuroscience research activities of 30+ scientists across multiple institutions, including the Mind Research Network, The University of New Mexico (Departments of Electrical and Computer Engineering, Psychology, Neuroscience, Education, Psychiatry, Neurology, Neurosurgery, Pediatrics and Radiology), New Mexico State University, the Lovelace Biomedical and Environmental Research Institute, Lovelace Scientific Resources Inc., the Lovelace Respiratory Research Institute, Los Alamos National Laboratory, and Sandia National Laboratory. The new system configuration will be achieved through a near complete up-grade (everything but the core magnet) of MRN?s outdated (2006) 3T Magnetom Trio system. Direct access to a Prisma scanner will allow MRN and the State of New Mexico to continue to be driving forces in the elucidation of brain-behavioral relationships in health and disease. We anticipate that the proposed upgrade will support NIH, NSF, DOD, FDA, DARPA and IARPA funded research in a wide range of scientific endeavors including (1) the development of new engineering-based tools for processing and fusion of multimodal neuroscientific data, (2) assessment of brain structure and function across the lifespan, in health and disease. The proposed machine will be the only research available human 3T MRI system in the state of New Mexico, and it will align our scanning capabilities with those available at other major brain imaging research centers in the United States. Also, the new scanner will allow our data acquisition procedures to parallel those being used in the Human Connectome Project. Such alignment is critical for our ability to appropriately train students and post-doctoral fellows on state-of-the-art imaging methods, and to continue our development of shared data analytic tools.

Project Summary. The parent R01 (R01MH094524) is focused on identifying links between gray matter networks, genomics, and symptom data in a larger aggregated dataset of schizophrenia and bipolar patients. The aims of the R01 are: aim 1) to determine gray matter networks underlying psychosis, 2) to identify homogeneous subclusters with similar gray matter and symptom profiles, 3) to identify the genomic patterns underlying gray matter maps, and 4) to extract genetic and structurally similar individual with psychosis. In this supplement, we are proposing to extend the initial techniques for identifying gray matter patterns underlying symptom profiles, to test both their specificity and generalizability across diagnostic categories. We apply them to two neurodegenerative disorders, Alzheimer?s disease and Huntington?s disease, using the ADNI and PREDICT-HD datasets21-24. Building on the research domain criteria (RDoC) dimensional approach we will focus on similar cognitive constructs with related symptoms in an extended sample of individuals. This supplemental work serves the purpose of the parent grant by determining whether the measures we find are specific to a neuropsychiatric population or if they generalize to neurodegenerative disorders; and it extends the understanding of neuropsychiatric symptoms and brain changes in Alzheimer?s disease (AD), an area of growing need. As many as 80% of AD patients may be affected by neuropsychiatric symptoms including delusions, apathy, and depression, with hallucinations possible but less frequent2. Understanding why these arise in some individuals but not others; how they relate to disease progression or response to treatment; and the underlying physiology of these symptoms are questions that are commonly addressed in psychosis research?but do the same answers hold for AD?

Abstract The administrative core (ADM) will function in an overall supervisory capacity for our Phase III COBRE. It is comprised of a steering committee, external and internal advisory committees, and other stakeholders. The ADM core will be directed by Dr. Vince Calhoun, the PI for the current proposal. It will also operate under the supervision of a well-respected psychiatric clinician-researcher, Dr. Juan Bustillo, and Dr. Nora Perrone-Bizzozero, an established neuroscientist and educator. Drs. Jeff Lieberman, MD and Charles Bowden, MD will serve as external consultants. The core director will coordinate the duties and goals of the ADM core, focused on three aims: 1) to coordinate all the specific budgetary, regulatory, and personnel aspects required in the three technical cores and Pilot Project Program; 2) set up a strong community outreach program and support new investigator projects using our pilot project core; and 3) maintain and continue to develop vital core facilities in consultation with internal and external advisors. A long-term aim of this Phase III award is to expand the capability of our facilities and further develop a diversified neuroimaging research environment that will continue to be competitive nationally and internationally. We believe that through the support of a COBRE Phase III, our cohesive and collaborative program of interdisciplinary and translational research can be leveraged into a center-level grant (e.g. P41, P50) which will be submitted approximately 3 years after the start of the Phase III award. Hence, this Phase III award is essential for the overall success of our New Mexico program on multimodal imaging of neuropsychiatric illness.

Project Summary/Abstract This Phase III (P-III) COBRE project will extend the cores that have been successfully leveraged in our Phase I (P-I) and Phase II (P-II) COBRE projects and sustain these unique resources in New Mexico through the im- plementation of a business plan. Over the past eight years we have built up infrastructure and created a cutting edge brain imaging center, our P-II project is just over half-way through and is even more successful than our P- I was at this point in time. The Mind Research Network (MRN) houses an Elekta Neuromag 306-channel MEG System, a high density EEG lab, a 3T Siemens Trio MRI scanner, and a mobile 1.5T Siemens Avanto MRI scanner. Additional resources include a centralized neuroinformatics system, a strong IT management plan, and state-of-the-art image analysis expertise and tools. This P-III COBRE center will continue our momentum and move the cores we have developed into a position of long term sustainability. We will continue with the technical cores established during the P-II project including multimodal data acquisition (MDA), algorithm and data analy- sis (ADA), and biostatistics and neuro-informatics (BNI). These cores have begun to serve MRN and the greater community, as well as other institutions including extensive collaborations with IDeA funded projects in New Mexico and other states. We believe this P-III COBRE is extremely well-positioned to establish and sustain New Mexico as one of the premier brain imaging sites. We include an extensive pilot project program (PPP) that is built on the successful pilot programs implemented as part of the earlier COBRE phases. This includes an ex- tensive educational, mentoring, and faculty development program to carefully mentor and position faculty who use the cores to maximize their potential to successfully compete for external funding, thus fulfilling the ultimate goals of the COBRE program. 2

Current computing systems that support research on sensitive data, such as personally identifiable information, are frequently single-purpose and rely on ad-hoc approaches to data protection and management. This project develops system called SAMPRA: Scalable Analysis, Management, and Protection of Research Artifacts. SAMPRA's goal is to provide a compliant research computing platform that supports diverse, inter-disciplinary, collaborative research on protected data. SAMPRA leverages modern virtualization technology to enable the decentralized management of protected computing enclaves that can be customized to the needs of each specific research project. In addition, the project trains researchers and students on best practices for managing and analyzing protected data, and technical staff on how to customize environments to the needs of individual research groups.

SAMPRA investigates multiple techniques to meet these goals, with the overall technical goal of understanding the technical and administrative tradeoffs between isolating and sharing protected research infrastructure services. First, SAMPRA systematically virtualizes hardware, software, and network resources to provide a flexible system architecture that supports research computing with varying analysis, management, and protection needs. SAMPRA also provides virtual data transfer nodes to interface protected environments with external data acquisition systems, with the goal of supporting modern data-intensive research projects using central institutional resources. The project develops exemplar computing, data analysis, and data management virtual environments, and integrating these with institutional systems for managing protected data. These exemplar systems are also the examples used in workshops that train researchers on the use of SAMPRA to support research.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Project Summary/Abstract In the field of Alzheimer?s and related disorder, there has been very little work focusing on imaging genomics biomarker approaches, despite considerable promise. In part this is due to the fact that most studies have fo- cused on candidate gene approaches or those that do not capitalize on capturing (and amplifying) small effects spread across many sites. Even for genome wide studies, the vast majority of imaging genomic studies still rely on massive univariate analyses. The use of multivariate approaches provides a powerful tool for analyzing the data in the context of genomic and connectomic networks (i.e. weighted combinations of voxels and genetic variables). It is clear that imaging and genomic data are high dimensional and include complex relationships that are poorly understood. Multivariate data fusion models that have been proposed to date typically suffer from two key limitations: 1) they require the data dimensionality to match (i.e. 4D fMRI data has to be reduced to 1D to match with the 1D genomic data, and 2) models typically assume linear relationships despite evidence of non- linearity in brain imaging and genomic data. New methods are needed that can handle data that has mixed temporal dimensionality, e.g., single nucleotide polymorphisms (SNPs) do not change over time, brain structure changes slowly over time, while fMRI changes rapidly over time. Secondly, methods that can handle complex relationships, such as groups of networks that are tightly coupled or nonlinear relationships in the data. To ad- dress these challenges, we introduce a new framework called flexible subspace analysis (FSA) that can auto- matically identify subspaces (groupings of unimodal or multimodal components) in joint multimodal data. Our approach leverages the interpretability of source separation approaches and can include additional flexibility by allowing for a combination of shallow and ?deep? subspaces, thus leveraging the power of deep learning. We will apply the developed models to a large longitudinal dataset of individuals at various stages of cognitive impair- ment and dementia. Using follow-up outcomes data we will evaluate the predictive accuracy of a joint analysis compared to a unimodal analysis, as well as its ability to characterize various clinical subtypes including those driven by vascular effects including subcortical ischemic vascular dementia versus those that are more neuro- degenerative. We will evaluate the single subject predictive power of these profiles in independent data to max- imize generalization. All methods and results will be shared with the community. The combination of advanced algorithmic approach plus the large N data promises to advance our understanding of Alzheimer?s and related disorders in addition to providing new tools that can be widely applied to other studies of complex disease. 3

Project Summary/Abstract Substance use behavior (smoking/drinking) is well known to be the result of a complex interplay between genetics and environmental components. However macro- and micro-environmental effects on smoking and drinking behavior are still poorly understood and there is little work evaluating the relationship of such factors to neurobiological changes measured via brain imaging. The data are available though. There is a great amount of internationally collected imaging and genomics data, however much of this data is not being analyzed due to privacy issues or other factors which prevent sharing of the raw data. To address this barrier, we propose to leverage and extend a software platform which enables decentralized analysis of data (e.g. sharing without sharing). We will use this platform to perform an analysis which pools together data from the 10,000 participant longitudinal US-based adolescent brain cognitive development (ABCD) study with an international cohort of more than 20,000 participants from Europe, China, and India called the global imaging genetics of adolescents (GIGA). Our focus will be on assessing macro- and micro-environmental and heritability factors associated with sub- stance use behavior as well as their neuronal biomarkers in the context of factors such as cultural acceptance, urbanization, annual household income, and climate. Results will provide new insights into the factors contrib- uting to substance use behavior. We also propose to develop a persistent decentralized analysis nodes for the GIGA sites using our Collaborative Informatics and Neuroimaging Suite Toolkit for Anonymous Computation (COINSTAC) to allow for re-analyses of these large cohorts as well as pooling with one?s local data without requiring collocation of the data at the same site. 3

Project Summary The Research Domain Criteria (RDoC) matrix delineates general constructs, that reflect basic dimensions of human behavioral functioning that can range from normal to abnormal. The RDoC matrix organizes these constructs by domains (e.g., positive valence and social processing systems) and units of analysis (i.e., from genes, to molecules, cells, circuits, physiology, behavior, self-report, paradigms) such that they can be systematically studied at multiple levels of analysis. Most clinical research studies, to date, have employed standardized symptom assessments, which are often disorder specific and not directly linked to RDoC constructs. In schizophrenia (SZ), negative symptom domains, including avolition, anhedonia, asociality, alogia, and blunted affect (5 factor model), have been studied in some detail. Recently a theoretical mapping between negative symptom domains and RDoC constructs linked avolition, anhedonia, and avolition to positive valence system, and alogia and flat affect to the social processes system. However, the proposed mappings between behavior (negative symptom domains) and brain structures/circuitry have not been tested or validated; either in SZ, or in other neuropsychiatric illnesses such as bipolar disorder (BD) or major depressive disorder (MDD). Earlier work suggested a more parsimonious 2-factor model of negative symptoms, in which avolition, anhedonia, and asociality were linked to a motivation and pleasure (MAP) factor, and and blunted affect andalogia linked to an expressive (EXP) factor. Of note, with the exception of asociality, these factors appear to map onto positive valence and social processes systems in the RDoC matrix; lending additional support to the proposed RDoC matrix structure related to negative symptoms. Mappings between different interpretations of negative symptom domains (e.g., 5-factor and 2-factor models) and brain structures/circuitry have also not been conducted. Leveraging the worldwide collaborative ENIGMA (Enhancing Neuro Imaging Genetics through Meta-Analysis) consortium and the COINSTAC (Collaborative Informatics and Neuroimaging Suite Toolkit for Anonymous Computation) computational platform, this proposal will combine neuroimaging and clinical measures of negative symptoms across schizophrenia (SZ), bipolar disorder (BD), and major depressive disorder (MDD), to validate and extend the RDoC matrix representation of negative symptom domains in major mental illness. We extract joint multimodal features for each separable (sub)construct, evaluate them for their relationship with the behavior, and then use them in a subsequent cross-validation analysis. Subsequently, we evaluate their single subject prediction power. Through these powerful computational methods, we will map structural, diffusion tensor imaging, and resting state functional magnetic resonance imaging measures of brain structures/circuitry to negative symptom behavioral measures. Successful completion of this proposal?s aims will identify distinct and overlapping neural circuits associated with negative symptom domains, will test integrative models of functioning, and identify dysregulation in psychopathology-related mechanisms that cut across traditional diagnostic boundaries.

Project Summary/Abstract Disorders of mood and psychosis such as schizophrenia, bipolar disorder, and unipolar depression are incredibly complex, influenced by both genetic and environmental factors, and the clinical characterizations are primarily based on symptoms rather than biological information. Current diagnostic approaches are based on symptoms, which overlap extensively in some cases, and there is growing consensus that we should approach mental illness as a continuum, rather than as a categorical entity. Since both genetic and environmental factors play a large role in mental illness, the combination of brain imaging and genomic data are poised to play an important role is clarifying our understanding of mental illness. However, both imaging and genomic data are high dimensional and include complex relationships that are poorly understood. To characterize the available information, we are in need of approaches that can deal with high-dimensional data exhibiting interactions at multiple levels (i.e., data fusion), while providing interpretable solutions (i.e., a focus on brain and genomic networks). An additional challenge exists because the available data has mixed temporal dimensionality, e.g., single nucleotide polymorphisms (SNPs) do not change over time, brain structure changes slowly over time, while fMRI changes rapidly over time. To address these challenges, we introduce a new unified framework called flexible subspace analysis (FSA) that can automatically identify subspaces (groupings of unimodal or multimodal components) in joint multimodal data. Our approach leverages the interpretability of source separation approaches and can include additional flexibility by allowing for a combination of shallow and ?deep? subspaces, thus leveraging the power of deep learning. We will apply the developed models to a large (N>60,000) dataset of individuals along the mood and psychosis spectrum to evaluate the important question of disease categorization. We will compute fully cross-validated genomic-neuro-behavioral profiles of individuals including a comparison of the predictive accuracy of 1) standard categories from the diagnostic and statistical manual of mental disorders (DSM), 2) data-driven subgroups, and 3) dimensional relationships. We will also evaluate the single subject predictive power of these profiles in independent data to maximize generalization. All methods and results will be shared with the community. The combination of advanced algorithmic approach plus the large N data promises to advance our understanding of the nosology of mood and psychosis disorders in addition to providing new tools that can be widely applied to other studies of complex disease.

Abstract Wide-spread data sharing has started to permeate the brain imaging community from funders to researchers. However, in recent years there have also been some concerns raised regarding ethical issues related to privacy and data ownership among others. In the parent award we are leveraging and extending a privacy preserving decentralized data sharing platform called COINSTAC to perform a study of gene-by-environmental effects by pooling together data from across the world, some of which is unable to be openly shared. In this supplement we will study various bioethical issues related to different data sharing strategies. This will include calculating risk scores from existing data to evaluate the effectiveness of machine learning to potentially reidentify from similar or different data types, a detailed survey of various policy makers and stakeholders including researchers, federal employees, IRB members, and more, and finally the development of a forward looking white paper addressing both privacy, policy, and regulatory aspects which attempts to frame the various aspects that arise in the contact of the spectrum of data sharing approaches including fully open, ?trust? based via data usage agreements, privacy preserving via tools like COINSTAC, and more. The outcomes of this supplement will provide a useful guide for the field going forward and also provide initial data necessary to develop a larger scale project on these topics going forward.

Abstract In the supplement we propose to partner with the ENIGMA APOE2 working group, led by Paul Thompson, to uncover the structural and functional network changes associated with the protective factor of apolipoprotein variant ?2, which occurs in approximately 8% of the population. We will also evaluate the relationship of brain measure to apathy and anhedonia within each APOE variant (which has also shown to be protective against mood disorders) and identify common and unique aspects in our mental illness cohorts. To do this we will develop a decentralized statistical testing framework for decentralized multivariate analysis with group independent component analysis (ICA) and also an enhanced data fusion approach. This will enable us to evaluate, in thousands of individuals, the unique structural and functional signatures of individuals with the protective APOE variant and the profiles of negative symptoms and their relationship to profiles of these symptoms in mental illness assessed within the parent R01.

ABSTRACT: The focus of this supplement request is to leverage and reinforce our ongoing biomarker identification work with methods specifically focusing on Alzheimer's disease (AD) and related disorders (ADRD). Deep learning methods that we are developing in the parent grant can produce an optimal performance based on learning end-to-end directly from the data. Our goal is to leverage models trained to classify AD from the full brain fMRI dynamics for capturing novel dynamic biomarkers of AD via trained model introspection. However, it is notoriously difficult to train models to predict directly from full brain dynamics without prior dimensionality reduction. To overcome this difficulty, we will develop self-supervised approaches that would take advantage of unrelated datasets and provide a performance boost that would allow obtaining reliable classification improvements even on small data. This improved classification directly transfers into more reliable introspection of why the model classifies subjects to AD. We plan to improve the robustness of these predictive/introspective methods and study these full-brain fMRI dynamic measures in younger adults who have CSF risk markers assessed for AD in order to evaluate the potential for leveraging such models as biomarkers of AD.

Project Summary/Abstract The brain imaging community is greatly benefiting from extensive data sharing efforts currently underway. However, there is still a major gap in that much data is still not openly shareable, which we propose to address. In addition, current approaches to data sharing often include significant logistical hurdles both for the investigator sharing the data (e.g. often times multiple data sharing agreements and approvals are required from US and international institutions) as well as for the individual requesting the data (e.g. substantial computational re- sources and time is needed to pool data from large studies with local study data). This needs to change, so that the scientific community can create a venue where data can be collected, managed, widely shared and analyzed while also opening up access to the (many) data sets which are not currently available (see overview on this from our group7). The large amount of existing data requires an approach that can analyze data in a distributed way while (if required) leaving control of the source data with the individual investigator or the data host; this motivates a dynamic, decentralized way of approaching large scale analyses. During the previous funding period, we developed a peer-to-peer system called the Collaborative Informatics and Neuroimaging Suite Toolkit for Anonymous Computation (COINSTAC). Our system provides an independent, open, no-strings-attached tool that performs analysis on datasets distributed across different locations. Thus, the step of actually aggregating data is avoided, while the strength of large-scale analyses can be retained. During this new phase we respond to the need for advanced algorithms such as linear mixed effects models and deep learning, by proposing to develop decentralized models for these approaches and also implement a fully scalable cloud-based framework with enhanced security features. To achieve this, in Aim 1, we will incorporate the necessary functionality to scale up analyses via the ability to work with either local or commercial private cloud environments, together with advanced visualization, quality control, and privacy and security features. This suite of new functions will open the floodgates for the use of COINSTAC by the larger neuroscience community to enable new discovery and analysis of unprecedented amounts of brain imaging data located throughout the world. We will also improve usability, training materials, engage the community in contributing to the open source code base, and ultimately facilitate the use of COINSTAC's tools for additional science and discovery in a broad range of applications. In Aim 2 we will extend the framework to handle powerful algorithms such as linear mixed effects models and deep learning, and to perform meta-learning for leveraging and updating fit models. And finally, in Aim 3, we will test this new functionality through a partnership with the worldwide ENIGMA addiction group, which is currently not able to perform advanced machine learning analyses on data that cannot be centrally located. We will evaluate the impact of 6 main classes of substances of abuse (e.g. methamphetamines, cocaine, cannabis, nicotine, opiates, alcohol and their combinations) using the new developed functionality. 3

Project Summary/Abstract The development of biomarkers for identifying preclinical or prodromal Alzheimer?s disorder are of great in- terest. While some initial results based on resting fMRI have been presented, accuracy, robustness, and relia- bility are still relatively low. One highly promising direction is the development of dynamic functional activity and functional connectivity approaches. These approaches have been shown to be especially promising most likely due to the highly dynamic nature of the brain and the unconstrained nature of resting fMRI. Currently, there are no methods that can provide a full characterization of temporal, spatial, and spatio-temporal dynamics nor can most existing approaches characterize heterogenous subgroups or complex multiscale relationships. We will develop new methods that can effectively capture dynamic connectivity and provide summary metrics with a focus on individualized prediction of Alzheimer?s disease well prior to the onset of the illness. We propose a novel family of models that builds on the well-structured framework of joint blind source separation to capture a more complete characterization of (potentially nonlinear) spatio-temporal dynamics. Our models will also pro- duce a rich set of metrics to characterize the available dynamics and enable in depth comparison with currently available models. We show evidence that such measures are likely to be considerably more sensitive and more accurate in classifying individuals. We will extensively validate our approaches in a variety of ways including simulations, concurrent EEG/fMRI data, and evaluation on a large normative data set. We will apply the devel- oped methods to several large datasets including a large longitudinal sample of individuals who have been scanned at Emory University with resting fMRI who also have CSF amyloid and tau PET measures. We will use the developed markers to predict cognitive decline, amyloid, and tau levels in these data and include both a discovery data set as well as an independent replication data set. Successful completion of our aims will be an important first step towards providing an opportunity to develop and evaluate interventions early enough to have a positive impact on long-term prognosis. We will provide open source tools and release data throughout the duration of the project via GitHub, a web portal and the NITRC repository, hence enabling other investigators to compare their own methods with our own as well as to apply them to a large variety of brain disorders. Our tools also have wide application to the study of the healthy brain as well as many other diseases. 37

Project Summary/Abstract Existing approaches to estimate and characterize whole brain time-varying connectivity from fMRI data have shown considerable promise, with exponential growth in research in this field. We and others have developed a powerful set of tools that are now in wide use in the community. However, the impact of mental illness on brain connectivity is complex, and as we show, limitations in existing methods often result in missing important features associated with brain disorders (e.g. transient fractionation of the spatial structure of brain networks). Some of these important limitations include 1) the most widely-used approaches often require a number of prior and limiting assumptions that are not well studied, 2) methods often assume linear relationships either within or between networks over time, and 3) methods assume spatially fixed nodes and ignore the possibility of spatially fluid evolution of networks over time. We propose a novel family of models that builds on the well-structured framework of joint blind source separation to capture a more complete characterization of (potentially nonlinear) spatio-temporal dynamics while providing a way to relax other limiting assumptions. Our models will also produce a rich set of metrics to characterize the available dynamics and enable in depth comparison with currently avail- able models including those that are model based. We will extensively validate our approaches in a variety of ways including simulations and evaluation of rigor and robustness in large normative data sets. Finally, we will apply the developed tools to study the important area of dynamic properties in mental illnesses including schiz- ophrenia, bipolar disorder, and the autism spectrum. There is considerable evidence of disruption of dynamics in all three disorders, and as we show the use of static (or even exiting dynamic) approaches can miss important information about brain related differences associated with each. We will provide open source tools and release data throughout the duration of the project via a web portal and the NITRC repository, hence enabling other investigators to use our approaches and compare their own methods with our own. Our tools have wide appli- cation to the study of the healthy brain as well as many other diseases such as Alzheimer's disease and attention deficit hyperactivity disorder. 38

Project Summary The Research Domain Criteria (RDoC) matrix delineates general constructs, that reflect basic dimensions of human behavioral functioning that can range from normal to abnormal. The RDoC matrix organizes these constructs by domains (e.g., positive valence and social processing systems) and units of analysis (i.e., from genes, to molecules, cells, circuits, physiology, behavior, self-report, paradigms) such that they can be systematically studied at multiple levels of analysis. Most clinical research studies, to date, have employed standardized symptom assessments, which are often disorder specific and not directly linked to RDoC constructs. In schizophrenia (SZ), negative symptom domains, including avolition, anhedonia, asociality, alogia, and blunted affect (5 factor model), have been studied in some detail. Recently a theoretical mapping between negative symptom domains and RDoC constructs linked avolition, anhedonia, and avolition to positive valence system, and alogia and flat affect to the social processes system. However, the proposed mappings between behavior (negative symptom domains) and brain structures/circuitry have not been tested or validated; either in SZ, or in other neuropsychiatric illnesses such as bipolar disorder (BD) or major depressive disorder (MDD). Earlier work suggested a more parsimonious 2-factor model of negative symptoms, in which avolition, anhedonia, and asociality were linked to a motivation and pleasure (MAP) factor, and and blunted affect andalogia linked to an expressive (EXP) factor. Of note, with the exception of asociality, these factors appear to map onto positive valence and social processes systems in the RDoC matrix; lending additional support to the proposed RDoC matrix structure related to negative symptoms. Mappings between different interpretations of negative symptom domains (e.g., 5-factor and 2-factor models) and brain structures/circuitry have also not been conducted. Leveraging the worldwide collaborative ENIGMA (Enhancing Neuro Imaging Genetics through Meta-Analysis) consortium and the COINSTAC (Collaborative Informatics and Neuroimaging Suite Toolkit for Anonymous Computation) computational platform, this proposal will combine neuroimaging and clinical measures of negative symptoms across schizophrenia (SZ), bipolar disorder (BD), and major depressive disorder (MDD), to validate and extend the RDoC matrix representation of negative symptom domains in major mental illness. We extract joint multimodal features for each separable (sub)construct, evaluate them for their relationship with the behavior, and then use them in a subsequent cross-validation analysis. Subsequently, we evaluate their single subject prediction power. Through these powerful computational methods, we will map structural, diffusion tensor imaging, and resting state functional magnetic resonance imaging measures of brain structures/circuitry to negative symptom behavioral measures. Successful completion of this proposal?s aims will identify distinct and overlapping neural circuits associated with negative symptom domains, will test integrative models of functioning, and identify dysregulation in psychopathology-related mechanisms that cut across traditional diagnostic boundaries.