arXiv subjects

Subharup Guha

Publications and source records attributed to Subharup Guha.

16 recordsLinked to original sources

arXiv2026

Supervised Low-Rank Structure Discovery for Developmental Epigenetic Aging in Ultra-High-Dimensional DNA Methylation Data

Ultra-high-dimensional array-based CpG methylation studies require statistical frameworks that simultaneously provide supervised structure discovery, interpretability, scalable latent-dimension identification, and computational feasibility. We propose SOLAR (Supervised Orthogonal Low-rank Adaptive Regression), a supervised low-rank latent-factor framework for identifying CpG-level methylation structure associated with residualized DNAm age. SOLAR combines orthogonal low-rank regression with a penalized maximum a posteriori formulation, dimension-adaptive BIC-type penalization, and a trans-dimensional simulated-annealing strategy for automatic latent-rank selection, together with theoretical guarantees including identifiability, fixed-rank recovery, and rank-selection consistency under suitable regularity conditions. The framework additionally incorporates computationally and memory-efficient optimization strategies demonstrating scalability up to $p=10^7$, while analyses at $p=10^6$ remain feasible on standard desktop computing environments. Simulation studies demonstrate stable rank recovery, competitive supervised signal recovery, and strong scalability across moderate-, high-, and ultra-high-dimensional regimes. Using longitudinal EPIC-array CpG methylation data from the GUSTO birth cohort, comprising $n=1051$ methylation profiles collected across infancy and early childhood with approximately 860,000 assayed CpGs per sample, SOLAR identifies heterogeneous supervised methylation structure associated with residualized DNAm age beyond chronological age alone, together with biologically coherent CpG signatures and enrichment patterns.

stat.ME↗

arXiv2026

Enhancing Inference for Small Cohorts via Transfer Learning and Weighted Integration of Multiple Datasets

Lung sepsis remains a significant concern in the Northeastern U.S., yet the national eICU Collaborative Database includes only a small number of patients from this region, highlighting underrepresentation. Understanding clinical variables such as FiO2, creatinine, platelets, and lactate, which reflect oxygenation, kidney function, coagulation, and metabolism, is crucial because these markers influence sepsis outcomes and may vary by sex. Transfer learning helps address small sample sizes by borrowing information from larger datasets, although differences in covariates and outcome-generating mechanisms between the target and external cohorts can complicate the process. We propose a novel weighting method, TRANSfer LeArning wiTh wEights (TRANSLATE), to integrate data from various sources by incorporating domain-specific characteristics through learned weights that align external data with the target cohort. These weights adjust for cohort differences, are proportional to each cohort's effective sample size, and downweight dissimilar cohorts. TRANSLATE offers theoretical guarantees for improved precision and applies to a wide range of estimands, including means, variances, and distribution functions. Simulations and a real-data application to sepsis outcomes in the Northeast cohort, using a much larger sample from other U.S. regions, show that the method enhances inference while accounting for regional heterogeneity.

stat.ME↗

arXiv2025

The R Package WMAP: Tools for Causal Meta-Analysis by Integrating Multiple Observational Studies

Integrating multiple observational studies for meta-analysis has sparked much interest. The presented R package WMAP (Weighted Meta-Analysis with Pseudo-Population) addresses a critical gap in the implementation of integrative weighting approaches for multiple observational studies and causal inferences about various groups of subjects, such as disease subtypes. The package features three weighting approaches, each representing a special case of the unified weighting framework introduced by Guha and Li (2024), which includes an extension of inverse probability weights for data integration settings. It performs meta-analysis on user-inputted datasets as follows: (i) it first estimates the propensity scores for study-group combinations, calculates subject balancing weights, and determines the effective sample size (ESS) for a user-specified weighting method; and (ii) it then estimates various features of multiple counterfactual group outcomes, such as group medians and differences in group means for the mRNA expression of eight genes. Additionally, bootstrap variability estimates are provided. Among the implemented weighting methods, we highlight the FLEXible, Optimized, and Realistic (FLEXOR) method, which is specifically designed to maximize the ESS within the unified framework. The use of the software is illustrated by simulations as well as a multi-site breast cancer study conducted in seven medical centers.

stat.ME↗

arXiv2024

Causal Meta-Analysis by Integrating Multiple Observational Studies with Multivariate Outcomes

Integrating multiple observational studies to make unconfounded causal or descriptive comparisons of group potential outcomes in a large natural population is challenging. Moreover, retrospective cohorts, being convenience samples, are usually unrepresentative of the natural population of interest and have groups with unbalanced covariates. We propose a general covariate-balancing framework based on pseudo-populations that extends established weighting methods to the meta-analysis of multiple retrospective cohorts with multiple groups. Additionally, by maximizing the effective sample sizes of the cohorts, we propose a FLEXible, Optimized, and Realistic (FLEXOR) weighting method appropriate for integrative analyses. We develop new weighted estimators for unconfounded inferences on wide-ranging population-level features and estimands relevant to group comparisons of quantitative, categorical, or multivariate outcomes. Asymptotic properties of these estimators are examined. Through simulation studies and meta-analyses of TCGA datasets, we demonstrate the versatility and reliability of the proposed weighting strategy, especially for the FLEXOR pseudo-population.

stat.ME↗

arXiv2024

Bayesian Estimation of Propensity Scores for Integrating Multiple Cohorts with High-Dimensional Covariates

Comparative meta-analyses of groups of subjects by integrating multiple observational studies rely on estimated propensity scores (PSs) to mitigate covariate imbalances. However, PS estimation grapples with the theoretical and practical challenges posed by high-dimensional covariates. Motivated by an integrative analysis of breast cancer patients across seven medical centers, this paper tackles the challenges associated with integrating multiple observational datasets. The proposed inferential technique, called Bayesian Motif Submatrices for Covariates (B-MSC), addresses the curse of dimensionality by a hybrid of Bayesian and frequentist approaches. B-MSC uses nonparametric Bayesian "Chinese restaurant" processes to eliminate redundancy in the high-dimensional covariates and discover latent motifs or lower-dimensional structure. With these motifs as potential predictors, standard regression techniques can be utilized to accurately infer the PSs and facilitate covariate-balanced group comparisons. Simulations and meta-analysis of the motivating cancer investigation demonstrate the efficacy of the B-MSC approach to accurately estimate the propensity scores and efficiently address covariate imbalance when integrating observational health studies with high-dimensional covariates.

stat.ME↗

arXiv2023

A Bayesian Nonparametric Approach for Identifying Differentially Abundant Taxa in Multigroup Microbiome Data with Covariates

Scientific studies in the last two decades have established the central role of the microbiome in disease and health. Differential abundance analysis seeks to identify microbial taxa associated with sample groups defined by a factor such as disease subtype, geographical region, or environmental condition. The results, in turn, help clinical practitioners and researchers diagnose disease and develop treatments more effectively. However, microbiome data analysis is uniquely challenging due to high-dimensionality, sparsity, compositionally, and collinearity. There is a critical need for unified statistical approaches for differential analysis in the presence of covariates. We develop a zero-inflated Bayesian nonparametric (ZIBNP) methodology that meets these multipronged challenges. The proposed technique flexibly adapts to the unique data characteristics, casts the high proportion of zeros in a censoring framework, and mitigates high-dimensionality and collinearity by utilizing the dimension-reducing property of the semiparametric Chinese restaurant process. Additionally, the ZIBNP approach relates the microbiome sampling depths to inferential precision while accommodating the compositional nature of microbiome data. Through simulation studies and analyses of the CAnine Microbiome during Parasitism (CAMP) and Global Gut microbiome datasets, we demonstrate the accuracy of ZIBNP compared to established methods for differential abundance analysis in the presence of covariates.

stat.ME↗

arXiv2023

A New Integrative Method for Multigroup Comparisons of Censored Survival Outcomes in Multiple Observational Studies

In observational studies, covariate imbalance generates confounding, resulting in biased comparisons. Although propensity score-based weighting approaches facilitate unconfounded group comparisons for implicit target populations, existing techniques may not directly or efficiently analyze multiple studies with multiple groups and provide results generalizable to larger populations. Moreover, few methods deliver precise inferences for various estimands with censored survival outcomes. We propose a new concordant target population approach, which constructs generalized balancing weights and realistic target populations. Our method can incorporate researcher-specified natural population attributes and synthesize information by appropriately compensating for over- or under-represented groups to achieve covariate balance. The constructed {concordant} weights are agnostic to specific estimators, estimands, and outcomes and maximize the effective sample size (ESS) for more precise inferences. Simulation studies and descriptive comparisons of glioblastoma outcomes of racial groups in multiple TCGA studies demonstrate the strategy's practical advantages. Unlike existing weighting techniques, the proposed concordant target population revealed a drastically different result: Blacks were more vulnerable and endured significantly worse prognoses; Asians had the best outcomes with a median overall survival of 1,024 (SE: 15.2) days, compared to 384 (SE: 1.2) and 329 (SE: 19.7) days for Whites and Blacks, respectively.

stat.ME↗

arXiv2023

Unconfounded Meta-analytical Frameworks for Multivariate Outcomes in Multigroup Observational Studies using Concordant Weights

While meta-analyzing retrospective cancer patient cohorts, an investigation of differences in the expressions of target oncogenes across cancer subtypes is of substantial interest because the results may uncover novel tumorigenesis mechanisms and improve screening and treatment strategies. Weighting methods facilitate unconfounded comparisons of multigroup potential outcomes in multiple observational studies. For example, Guha et al. (2022) introduced concordant weights, allowing integrative analyses of survival outcomes by maximizing the effective sample size. However, it remains unclear how to use this or other weighting approaches to analyze a variety of continuous, categorical, ordinal, or multivariate outcomes, especially when research interests prioritize uncommon or unplanned estimands suggested by post hoc analyses; examples include percentiles and moments of group potential outcomes and pairwise correlations of multivariate outcomes. This paper proposes a unified meta-analytical approach accommodating various types of endpoints and fosters new estimators compatible with most weighting frameworks. Asymptotic properties of the estimators are investigated under mild assumptions. For undersampled groups, we devise small-sample procedures for quantifying estimation uncertainty. We meta-analyze multi-site TCGA breast cancer data, shedding light on the differential mRNA expression patterns of eight targeted genes for the subtypes infiltrating ductal carcinoma and infiltrating lobular carcinoma.

stat.ME↗

arXiv2023

Nonparametric Bayes Differential Analysis of Multigroup DNA Methylation Data

DNA methylation datasets in cancer studies are comprised of measurements on a large number of genomic locations called cytosine-phosphate-guanine (CpG) sites with complex correlation structures. A fundamental goal of these studies is the development of statistical techniques that can identify disease genomic signatures across multiple patient groups defined by different experimental or biological conditions. We propose BayesDiff, a nonparametric Bayesian approach for differential analysis relying on a novel class of first order mixture models called the Sticky Pitman-Yor process or two-restaurant two-cuisine franchise (2R2CF). The BayesDiff methodology flexibly utilizes information from all CpG sites or probes, adaptively accommodates any serial dependence due to the widely varying inter-probe distances and performs simultaneous inferences about the differential genomic signature of the patient groups. Using simulation studies, we demonstrate the effectiveness of the BayesDiff procedure relative to existing statistical techniques for differential DNA methylation. The methodology is applied to analyze a gastrointestinal (GI) cancer dataset that displays both serial correlations and interaction patterns. The results support and complement known aspects of DNA methylation and gene association in upper GI cancers.

stat.ME↗

arXiv2022

Nonparametric Bayes Differential Analysis for Dependent Multigroup Data with Application to DNA Methylation Analyses in Cancer

Modern cancer genomics datasets involve widely varying sizes and scales, measurement variables, and correlation structures. A fundamental analytical goal in these high-throughput studies is the development of general statistical techniques that can cleanly sift the signal from noise in identifying disease-specific genomic signatures across a set of experimental or biological conditions. We propose BayesDiff, a nonparametric Bayesian approach based on a novel class of first order mixture models, called the Sticky Poisson-Dirichlet process or multicuisine restaurant franchise. The BayesDiff methodology flexibly utilizes information from all the measurements and adaptively accommodates any serial dependence in the data, accounting for the inter-probe distances, to perform simultaneous inferences on the variables. The technique is applied to analyze a DNA methylation gastrointestinal (GI) cancer dataset, which displays both serial correlations and complex interaction patterns. Our analyses and results both support and complement known aspects of DNA methylation and gene association in upper GI cancers. In simulation studies, we demonstrate the effectiveness of the BayesDiff procedure relative to existing techniques for differential DNA methylation.

stat.ME↗

arXiv2022

Predicting Phenotypes from Brain Connection Structure

This article focuses on the problem of predicting a response variable based on a network-valued predictor. Our motivation is the development of interpretable and accurate predictive models for cognitive traits and neuro-psychiatric disorders based on an individual's brain connection network (connectome). Current methods reduce the complex, high dimensional brain network into low-dimensional pre-specified features prior to applying standard predictive algorithms. These methods are sensitive to feature choice and inevitably discard important information. Instead, we propose a nonparametric Bayes class of models that utilize the entire adjacency matrix defining brain region connections to adaptively detect predictive algorithms, while maintaining interpretability. The Bayesian Connectomics (BaCon) model class utilizes Poisson-Dirichlet processes to find a lower-dimensional, bidirectional (covariate, subject) pattern in the adjacency matrix. The small n, large p problem is transformed into a "small n, small q" problem, facilitating an effective stochastic search of the predictors. A spike-and-slab prior for the cluster predictors strikes a balance between regression model parsimony and flexibility, resulting in improved inferences and test case predictions. We describe basic properties of the BaCon model and develop efficient algorithms for posterior computation. The resulting methods are found to outperform existing approaches and applied to a creative reasoning data set.

stat.ME↗

arXiv2022

A Clustering Approach to Integrative Analysis of Multiomic Cancer Data

Rapid technological advances have allowed for molecular profiling across multiple omics domains from a single sample for clinical decision making in many diseases, especially cancer. As tumor development and progression are dynamic biological processes involving composite genomic aberrations, key challenges are to effectively assimilate information from these domains to identify genomic signatures and biological entities that are druggable, develop accurate risk prediction profiles for future patients, and identify novel patient subgroups for tailored therapy and monitoring. We propose integrative probabilistic frameworks for high-dimensional multiple-domain cancer data that coherently incorporate dependence within and between domains to accurately detect tumor subtypes, thus providing a catalogue of genomic aberrations associated with cancer taxonomy. We propose an innovative, flexible and scalable Bayesian nonparametric framework for simultaneous clustering of both tumor samples and genomic probes. We describe an efficient variable selection procedure to identify relevant genomic aberrations that can potentially reveal underlying drivers of a disease. Although the work is motivated by several investigations related to lung cancer, the proposed methods are broadly applicable in a variety of contexts involving high-dimensional data. The success of the methodology is demonstrated using artificial data and lung cancer omics profiles publicly available from The Cancer Genome Atlas.

stat.ME↗

arXiv2020

Probabilistic Detection and Estimation of Conic Sections from Noisy Data

Inferring unknown conic sections on the basis of noisy data is a challenging problem with applications in computer vision. A major limitation of the currently available methods for conic sections is that estimation methods rely on the underlying shape of the conics (being known to be ellipse, parabola or hyperbola). A general purpose Bayesian hierarchical model is proposed for conic sections and corresponding estimation method based on noisy data is shown to work even when the specific nature of the conic section is unknown. The model, thus, provides probabilistic detection of the underlying conic section and inference about the associated parameters of the conic section. Through extensive simulation studies where the true conics may not be known, the methodology is demonstrated to have practical and methodological advantages relative to many existing techniques. In addition, the proposed method provides probabilistic measures of uncertainty of the estimated parameters. Furthermore, we observe high fidelity to the true conics even in challenging situations, such as data arising from partial conics in arbitrarily rotated and non-standard form, and where a visual inspection is unable to correctly identify the type of conic section underlying the data.

stat.ME↗

arXiv2016

Nonparametric Variable Selection, Clustering and Prediction for High-Dimensional Regression

The development of parsimonious models for reliable inference and prediction of responses in high-dimensional regression settings is often challenging due to relatively small sample sizes and the presence of complex interaction patterns between a large number of covariates. We propose an efficient, nonparametric framework for simultaneous variable selection, clustering and prediction in high-throughput regression settings with continuous or discrete outcomes, called VariScan. The VariScan model utilizes the sparsity induced by Poisson-Dirichlet processes (PDPs) to group the covariates into lower-dimensional latent clusters consisting of covariates with similar patterns among the samples. The data are permitted to direct the choice of a suitable cluster allocation scheme, choosing between PDPs and their special case, a Dirichlet process. Subsequently, the latent clusters are used to build a nonlinear prediction model for the responses using an adaptive mixture of linear and nonlinear elements, thus achieving a balance between model parsimony and flexibility. We investigate theoretical properties of the VariScan procedure that differentiate the allocations patterns of PDPs and Dirichlet processes both in terms of the number and relative sizes of their clusters. Additional theoretical results guarantee the high accuracy of the model-based clustering procedure, and establish model selection and prediction consistency. Through simulation studies and analyses of benchmark data sets, we demonstrate the reliability of VariScan's clustering mechanism and show that the technique compares favorably to, and often outperforms, existing methodologies in terms of the prediction accuracies of the subject-specific responses.

stat.ME↗

arXiv2016

A Nonparametric Bayesian Technique for High-Dimensional Regression

This paper proposes a nonparametric Bayesian framework called VariScan for simultaneous clustering, variable selection, and prediction in high-throughput regression settings. Poisson-Dirichlet processes are utilized to detect lower-dimensional latent clusters of covariates. An adaptive nonlinear prediction model is constructed for the response, achieving a balance between model parsimony and flexibility. Contrary to conventional belief, cluster detection is shown to be aposteriori consistent for a general class of models as the number of covariates and subjects grows. Simulation studies and data analyses demonstrate that VariScan often outperforms several well-known statistical methods.

stat.ME↗

arXiv2015

hmmSeq: A hidden Markov model for detecting differentially expressed genes from RNA-seq data

We introduce hmmSeq, a model-based hierarchical Bayesian technique for detecting differentially expressed genes from RNA-seq data. Our novel hmmSeq methodology uses hidden Markov models to account for potential co-expression of neighboring genes. In addition, hmmSeq employs an integrated approach to studies with technical or biological replicates, automatically adjusting for any extra-Poisson variability. Moreover, for cases when paired data are available, hmmSeq includes a paired structure between treatments that incoporates subject-specific effects. To perform parameter estimation for the hmmSeq model, we develop an efficient Markov chain Monte Carlo algorithm. Further, we develop a procedure for detection of differentially expressed genes that automatically controls false discovery rate. A simulation study shows that the hmmSeq methodology performs better than competitors in terms of receiver operating characteristic curves. Finally, the analyses of three publicly available RNA-seq data sets demonstrate the power and flexibility of the hmmSeq methodology. An R package implementing the hmmSeq framework will be submitted to CRAN upon publication of the manuscript.

stat.AP↗