Vaibhav A. Narayan

MULTI-SOURCE FEATURE LEARNING FOR JOINT ANALYSIS OF INCOMPLETE MULTIPLE HETEROGENEOUS NEUROIMAGING DATA

Analysis of incomplete data is a big challenge when integrating large-scale brain imaging datasets from different imaging modalities. In the Alzheimers Disease Neuroimaging Initiative (ADNI), for example, over half of the subjects lack cerebrospinal fluid (CSF) measurements; an independent half of the subjects do not have fluorodeoxyglucose positron emission tomography (FDG-PET) scans; many lack proteomics measurements. Traditionally, subjects with missing measures are discarded, resulting in a severe loss of available information. In this paper, we address this problem by proposing an incomplete Multi-Source Feature (iMSF) learning method where all the samples (with at least one available data source) can be used. To illustrate the proposed approach, we classify patients from the ADNI study into groups with Alzheimers disease (AD), mild cognitive impairment (MCI) and normal controls, based on the multi-modality data. At baseline, ADNIs 780 participants (172AD, 397 MCI, 211 NC), have at least one of four data types: magnetic resonance imaging (MRI), FDG-PET, CSF and proteomics. These data are used to test our algorithm. Depending on the problem being solved, we divide our samples according to the availability of data sources, and we learn shared sets of features with state-of-the-art sparse learning methods. To build a practical and robust system, we construct a classifier ensemble by combining our method with four other methods for missing value estimation. Comprehensive experiments with various parameters show that our proposed iMSF method and the ensemble model yield stable and promising results.

Modeling disease progression via multi-task learning

Alzheimers disease (AD), the most common type of dementia, is a severe neurodegenerative disorder. Identifying biomarkers that can track the progress of the disease has recently received increasing attentions in AD research. An accurate prediction of disease progression would facilitate optimal decision-making for clinicians and patients. A definitive diagnosis of AD requires autopsy confirmation, thus many clinical/cognitive measures including Mini Mental State Examination (MMSE) and Alzheimers Disease Assessment Scale cognitive subscale (ADAS-Cog) have been designed to evaluate the cognitive status of the patients and used as important criteria for clinical diagnosis of probable AD. In this paper, we consider the problem of predicting disease progression measured by the cognitive scores and selecting biomarkers predictive of the progression. Specifically, we formulate the prediction problem as a multi-task regression problem by considering the prediction at each time point as a task and propose two novel multi-task learning formulations. We have performed extensive experiments using data from the Alzheimers Disease Neuroimaging Initiative (ADNI). Specifically, we use the baseline MRI features to predict MMSE/ADAS-Cog scores in the next 4 years. Results demonstrate the effectiveness of the proposed multi-task learning formulations for disease progression in comparison with single-task learning algorithms including ridge regression and Lasso. We also perform longitudinal stability selection to identify and analyze the temporal patterns of biomarkers in disease progression. We observe that cortical thickness average of left middle temporal, cortical thickness average of left and right Entorhinal, and white matter volume of left Hippocampus play significant roles in predicting ADAS-Cog at all time points. We also observe that several MRI biomarkers provide significant information for predicting MMSE scores for the first 2 years, however very few are shown to be significant in predicting MMSE score at later stages. The lack of predictable MRI biomarkers in later stages may contribute to the lower prediction performance of MMSE than that of ADAS-Cog in our study and other related studies.

The ADAS-Cog revisited: novel composite scales based on ADAS-Cog to improve efficiency in MCI and early AD trials.

The Alzheimers Disease Assessment Scale‐Cognitive (ADAS‐Cog) has been used widely as a cognitive end point in Alzheimers Disease (AD) clinical trials. Efforts to treat AD pathology at earlier stages have also used ADAS‐Cog, but failure in these trials can be difficult to interpret because the scale has well‐known ceiling effects that limit its use in mild cognitive impairment (MCI) and early AD. A wealth of data exists in ADAS‐Cog from both historical trials and contemporary longitudinal natural history studies that can provide insights about parts of the scale that may be better suited for MCI and early AD trials.

Quantifying the Pathophysiological Timeline of Alzheimer's Disease

Hypothetical models of AD progression typically relate clinical stages of AD to sequential changes in CSF biomarkers, imaging, and cognition. However, quantifying the continuous trajectories proposed by these models over time is difficult because of the difficulty in relating the dynamics of different biomarkers during a clinical trial that is significantly shorter than the duration of the disease. We seek to show that through proper synchronization, it is possible to de-convolve these trends and quantify the periods of time associated with different pathophysiological changes associated with Alzheimers disease (AD). We developed a model that replicated the observed progression of ADAS-Cog 13 scores and used this as a more precise estimate of disease-duration and thus pathologic stage. We then synchronized cerebrospinal fluid (CSF) and imaging biomarkers according to our new disease timeline. By de-convolving disease progression via ADAS-Cog 13, we were able to confirm the predictions of previous hypothetical models of disease progression as well as establish concrete timelines for different pathobiological events. Specifically, our work supports a sequential pattern of biomarker changes in AD in which reduction in CSF Aβ(42) and brain atrophy precede the increases in CSF tau and phospho-tau.

Multi-source learning for joint analysis of incomplete multi-modality neuroimaging data

Incomplete data present serious problems when integrating large-scale brain imaging data sets from different imaging modalities. In the Alzheimers Disease Neuroimaging Initiative (ADNI), for example, over half of the subjects lack cerebrospinal fluid (CSF) measurements; an independent half of the subjects do not have fluorodeoxyglucose positron emission tomography (FDG-PET) scans; many lack proteomics measurements. Traditionally, subjects with missing measures are discarded, resulting in a severe loss of available information. We address this problem by proposing two novel learning methods where all the samples (with at least one available data source) can be used. In the first method, we divide our samples according to the availability of data sources, and we learn shared sets of features with state-of-the-art sparse learning methods. Our second method learns a base classifier for each data source independently, based on which we represent each source using a single column of prediction scores; we then estimate the missing prediction scores, which, combined with the existing prediction scores, are used to build a multi-source fusion model. To illustrate the proposed approaches, we classify patients from the ADNI study into groups with Alzheimers disease (AD), mild cognitive impairment (MCI) and normal controls, based on the multi-modality data. At baseline, ADNIs 780 participants (172 AD, 397 MCI, 211 Normal), have at least one of four data types: magnetic resonance imaging (MRI), FDG-PET, CSF and proteomics. These data are used to test our algorithms. Comprehensive experiments show that our proposed methods yield stable and promising results.

Sparse generalized functional linear model for predicting remission status of depression patients.

Complex diseases such as major depression affect people over time in complicated patterns. Longitudinal data analysis is thus crucial for understanding and prognosis of such diseases and has received considerable attention in the biomedical research community. Traditional classification and regression methods have been commonly applied in a simple (controlled) clinical setting with a small number of time points. However, these methods cannot be easily extended to the more general setting for longitudinal analysis, as they are not inherently built for time-dependent data. Functional regression, in contrast, is capable of identifying the relationship between features and outcomes along with time information by assuming features and/or outcomes as random functions over time rather than independent random variables. In this paper, we propose a novel sparse generalized functional linear model for the prediction of treatment remission status of the depression participants with longitudinal features. Compared to traditional functional regression models, our model enables high-dimensional learning, smoothness of functional coefficients, longitudinal feature selection and interpretable estimation of functional coefficients. Extensive experiments have been conducted on the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) data set and the results show that the proposed sparse functional regression method achieves significantly higher prediction power than existing approaches.

Side effect profile similarities shared between antidepressants and immune-modulators reveal potential novel targets for treating major depressive disorders

BackgroundSide effects, or the adverse effects of drugs, contain important clinical phenotypic information that may be useful in predicting novel or unknown targets of a drug. It has been suggested that drugs with similar side-effect profiles may share common targets. The diagnostic class, Major Depressive Disorder, is increasingly viewed as being comprised of multiple depression subtypes with different biological root causes. One ‘type’ of depression generating substantial interest today focuses on patients with high levels of inflammatory burden, indicated by elevated levels of C-reactive proteins (CRP) and pro-inflammatory cytokines such as interleukin 6 (IL-6). It has been suggested that drugs targeting the immune system may have beneficial effect on this subtype of depressed patients, and several studies are underway to test this hypothesis directly. However, patients have been treated with both anti-inflammatory and antidepressant compounds for decades. It may be possible to exploit similarities in clinical readouts to better understand the antidepressant effects of immune-related drugs.MethodsHere we explore the space of approved drugs by comparing the drug side effect profiles of known antidepressants and drugs targeting the immune system, and further examine the findings by comparing the human cell line expression profiles induced by them with those induced by antidepressants.ResultsWe found 7 immune-modulators and 14 anti-inflammatory drugs sharing significant side effect profile similarities with antidepressants. Five of the 7 immune modulators share most similar side effect profiles with antidepressants that modulate dopamine release and/or uptake. In addition, the immunosuppressant rapamycin and the glucocorticoid alclometasone induces transcriptional changes similar to multiple antidepressants.ConclusionsThese findings suggest that some antidepressants and some immune-related drugs may affect common molecular pathways. Our findings support the idea that certain medications aimed at the immune system may be helpful in relieving depressive symptoms, and suggest that it may be of value to test immune-modulators for antidepressant-like activity in future proof-of-concept studies.

Melancholic depression prediction by identifying representative features in metabolic and microarray profiles with missing values.

Recent studies have revealed that melancholic depression, one major subtype of depression, is closely associated with the concentration of some metabolites and biological functions of certain genes and pathways. Meanwhile, recent advances in biotechnologies have allowed us to collect a large amount of genomic data, e.g., metabolites and microarray gene expression. With such a huge amount of information available, one approach that can give us new insights into the understanding of the fundamental biology underlying melancholic depression is to build disease status prediction models using classification or regression methods. However, the existence of strong empirical correlations, e.g., those exhibited by genes sharing the same biological pathway in microarray profiles, tremendously limits the performance of these methods. Furthermore, the occurrence of missing values which are ubiquitous in biomedical applications further complicates the problem. In this paper, we hypothesize that the problem of missing values might in some way benefit from the correlation between the variables and propose a method to learn a compressed set of representative features through an adapted version of sparse coding which is capable of identifying correlated variables and addressing the issue of missing values simultaneously. An efficient algorithm is also developed to solve the proposed formulation. We apply the proposed method on metabolic and microarray profiles collected from a group of subjects consisting of both patients with melancholic depression and healthy controls. Results show that the proposed method can not only produce meaningful clusters of variables but also generate a set of representative features that achieve superior classification performance over those generated by traditional clustering and data imputation techniques. In particular, on both datasets, we found that in comparison with the competing algorithms, the representative features learned by the proposed method give rise to significantly improved sensitivity scores, suggesting that the learned features allow prediction with high accuracy of disease status in those who are diagnosed with melancholic depression. To our best knowledge, this is the first work that applies sparse coding to deal with high feature correlations and missing values, which are common challenges in many biomedical applications. The proposed method can be readily adapted to other biomedical applications involving incomplete and high-dimensional data.

Optimizing Regions-of-Interest Composites for Capturing Treatment Effects on Brain Amyloid in Clinical Trials

BACKGROUNDnPittsburgh Compound B (PiB) positron emission tomography (PET) neuroimaging is a powerful research tool to characterize amyloid evolution in the brain. Quantification of amyloid load critically depends on (i) the choice of a reference region (RR) and (ii) on the selection of regions of interest (ROIs) to derive the standard uptake value ratios (SUVRs).nnnOBJECTIVEnTo evaluate the stability, i.e., negligible amyloid accumulation over time, of different RRs, and the performance of different PiB summary measures defined by selected ROIs and RRs for their sensitivity to detecting longitudinal change in amyloid burden.nnnMETHODSnTo evaluate RRs, cross-sectional and longitudinal analyses of focal regional and composite measures of amyloid accumulation were carried out on the standardized PiB-PET regional data for cerebellar grey matter (CER), subcortical white matter (SWM), and pons (PON). RRs and candidate composite SUVR measures were further evaluated to select regions and develop novel composites, using standardized 2-year change from baseline.nnnRESULTSnLongitudinal trajectories of PiB4-average of anterior cingulate (ACG), frontal cortex (FRC), parietal cortex, and precuneus-demonstrated marked variability and small change from baseline when normalized to CER, larger changes and less variability when normalized to SWM, which was further enhanced for the composite in PON-normalized settings. Novel composite PiB3, comprised of the average SUVRs of lateral temporal cortex, ACG, and FRC was created.nnnCONCLUSIONnPON and SWM appeared to be more stable RRs than the CER. PiB3 showed compelling sample size reduction and gains in power calculations for clinical trials over conventional PiB4 composite.

A Novel Subject Synchronization Clinical Trial Design for Alzheimer's Disease

One of the challenges in developing a viable therapy for Alzheimers disease has been demonstrating efficacy within a clinical trial. Using this as motivation, we sought to re-examine conventional clinical trial practices in order to determine whether efficacy can be better shown through alternative trial designs and novel analysis methods. In this work, we hypothesize that the confounding factors which hamper the ability to discern a treatment signal are the variability in observations as well as the insidious nature of the disease. We demonstrate that a two-phase trial design in which drug dosing is administered after a certain level of disease severity has been reached, coupled with a method to account more accurately for the progression of the disease, may allow us to compensate for these factors, and thus enable us to make treatment effects more apparent. Utilizing data from two previously failed trials which involved the evaluation of galantamine for indication in mild cognitive impairment, we were able to demonstrate that a clear treatment effect can be realized through both visual and statistical means, and propose that future trials may be more likely to show success if similar methods are utilized.