Searcharxiv⌕ Search

arXiv subjects

Sach Mukherjee

Publications and source records attributed to Sach Mukherjee.

34 records · Page 2Linked to original sources
arXiv2019

Ancestral causal learning in high dimensions with a human genome-wide application

We consider learning ancestral causal relationships in high dimensions. Our approach is driven by a supervised learning perspective, with discrete indicators of causal relationships treated as labels to be learned from available data. We focus on the setting in which some causal (ancestral) relationships are known (via background knowledge or experimental data) and put forward a general approach that scales to large problems. This is motivated by problems in human biology which are characterized by high dimensionality and potentially many latent variables. We present a case study involving interventional data from human cells with total dimension $p \! \sim \! 19{,}000$. Performance is assessed empirically by testing model output against previously unseen interventional data. The proposed approach is highly effective and demonstrably scalable to the human genome-wide setting. We consider sensitivity to background knowledge and find that results are robust to nontrivial perturbations of the input information. We consider also the case, relevant to some applications, where the only prior information available concerns a small number of known ancestral relationships.

stat.ML↗
arXiv2019

Model-based clustering in very high dimensions via adaptive projections

Mixture models are a standard approach to dealing with heterogeneous data with non-i.i.d. structure. However, when the dimension $p$ is large relative to sample size $n$ and where either or both of means and covariances/graphical models may differ between the latent groups, mixture models face statistical and computational difficulties and currently available methods cannot realistically go beyond $p \! \sim \! 10^4$ or so. We propose an approach called Model-based Clustering via Adaptive Projections (MCAP). Instead of estimating mixtures in the original space, we work with a low-dimensional representation obtained by linear projection. The projection dimension itself plays an important role and governs a type of bias-variance tradeoff with respect to recovery of the relevant signals. MCAP sets the projection dimension automatically in a data-adaptive manner, using a proxy for the assignment risk. Combining a full covariance formulation with the adaptive projection allows detection of both mean and covariance signals in very high dimensional problems. We show real-data examples in which covariance signals are reliably detected in problems with $p \! \sim \! 10^4$ or more, and simulations going up to $p = 10^6$. In some examples, MCAP performs well even when the mean signal is entirely removed, leaving differential covariance structure in the high-dimensional space as the only signal. Across a number of regimes, MCAP performs as well or better than a range of existing methods, including a recently-proposed $\ell_1$-penalized approach; and performance remains broadly stable with increasing dimension. MCAP can be run "out of the box" and is fast enough for interactive use on large-$p$ problems using standard desktop computing resources.

stat.ML↗
arXiv2016

High-dimensional regression over disease subgroups

We consider high-dimensional regression over subgroups of observations. Our work is motivated by biomedical problems, where disease subtypes, for example, may differ with respect to underlying regression models, but sample sizes at the subgroup-level may be limited. We focus on the case in which subgroup-specific models may be expected to be similar but not necessarily identical. Our approach is to treat subgroups as related problem instances and jointly estimate subgroup-specific regression coefficients. This is done in a penalized framework, combining an $\ell_1$ term with an additional term that penalizes differences between subgroup-specific coefficients. This gives solutions that are globally sparse but that allow information-sharing between the subgroups. We present algorithms for estimation and empirical results on simulated data and using Alzheimer's disease, amyotrophic lateral sclerosis and cancer datasets. These examples demonstrate the gains our approach can offer in terms of prediction and the ability to estimate subgroup-specific sparsity patterns.

stat.AP↗
arXiv2015

Inferring network structure from interventional time-course experiments

Graphical models are widely used to study biological networks. Interventions on network nodes are an important feature of many experimental designs for the study of biological networks. In this paper we put forward a causal variant of dynamic Bayesian networks (DBNs) for the purpose of modeling time-course data with interventions. The models inherit the simplicity and computational efficiency of DBNs but allow interventional data to be integrated into network inference. We show empirical results, on both simulated and experimental data, that demonstrate the need to appropriately handle interventions when interventions form part of the design.

stat.ME↗
arXiv2014

Joint estimation of multiple related biological networks

Graphical models are widely used to make inferences concerning interplay in multivariate systems. In many applications, data are collected from multiple related but nonidentical units whose underlying networks may differ but are likely to share features. Here we present a hierarchical Bayesian formulation for joint estimation of multiple networks in this nonidentically distributed setting. The approach is general: given a suitable class of graphical models, it uses an exchangeability assumption on networks to provide a corresponding joint formulation. Motivated by emerging experimental designs in molecular biology, we focus on time-course data with interventions, using dynamic Bayesian networks as the graphical models. We introduce a computationally efficient, deterministic algorithm for exact joint inference in this setting. We provide an upper bound on the gains that joint estimation offers relative to separate estimation for each network and empirical results that support and extend the theory, including an extensive simulation study and an application to proteomic data from human cancer cell lines. Finally, we describe approximations that are still more computationally efficient than the exact algorithm and that also demonstrate good empirical performance.

stat.ME↗
arXiv2014

Exact Estimation of Multiple Directed Acyclic Graphs

This paper considers the problem of estimating the structure of multiple related directed acyclic graph (DAG) models. Building on recent developments in exact estimation of DAGs using integer linear programming (ILP), we present an ILP approach for joint estimation over multiple DAGs, that does not require that the vertices in each DAG share a common ordering. Furthermore, we allow also for (potentially unknown) dependency structure between the DAGs. Results are presented on both simulated data and fMRI data obtained from multiple subjects.

stat.ML↗
arXiv2014

Estimating causal structure using conditional DAG models

This paper considers inference of causal structure in a class of graphical models called "conditional DAGs". These are directed acyclic graph (DAG) models with two kinds of variables, primary and secondary. The secondary variables are used to aid in estimation of causal relationships between the primary variables. We give causal semantics for this model class and prove that, under certain assumptions, the direction of causal influence is identifiable from the joint observational distribution of the primary and secondary variables. A score-based approach is developed for estimation of causal structure using these models and consistency results are established. Empirical results demonstrate gains compared with formulations that treat all variables on an equal footing, or that ignore secondary variables. The methodology is motivated by applications in molecular biology and is illustrated here using simulated data and in an analysis of proteomic data from the Cancer Genome Atlas.

stat.ME↗
arXiv2014

Joint Structure Learning of Multiple Non-Exchangeable Networks

Several methods have recently been developed for joint structure learning of multiple (related) graphical models or networks. These methods treat individual networks as exchangeable, such that each pair of networks are equally encouraged to have similar structures. However, in many practical applications, exchangeability in this sense may not hold, as some pairs of networks may be more closely related than others, for example due to group and sub-group structure in the data. Here we present a novel Bayesian formulation that generalises joint structure learning beyond the exchangeable case. In addition to a general framework for joint learning, we (i) provide a novel default prior over the joint structure space that requires no user input; (ii) allow for latent networks; (iii) give an efficient, exact algorithm for the case of time series data and dynamic Bayesian networks. We present empirical results on non-exchangeable populations, including a real data example from biology, where cell-line-specific networks are related according to genomic features.

stat.ME↗
arXiv2014

Penalized estimation in high-dimensional hidden Markov models with state-specific graphical models

We consider penalized estimation in hidden Markov models (HMMs) with multivariate Normal observations. In the moderate-to-large dimensional setting, estimation for HMMs remains challenging in practice, due to several concerns arising from the hidden nature of the states. We address these concerns by $\ell_1$-penalization of state-specific inverse covariance matrices. Penalized estimation leads to sparse inverse covariance matrices which can be interpreted as state-specific conditional independence graphs. Penalization is nontrivial in this latent variable setting; we propose a penalty that automatically adapts to the number of states $K$ and the state-specific sample sizes and can cope with scaling issues arising from the unknown states. The methodology is adaptive and very general, applying in particular to both low- and high-dimensional settings without requiring hand tuning. Furthermore, our approach facilitates exploration of the number of states $K$ by coupling estimation for successive candidate values $K$. Empirical results on simulated examples demonstrate the effectiveness of the proposed approach. In a challenging real data example from genome biology, we demonstrate the ability of our approach to yield gains in predictive power and to deliver richer estimates than existing methods.

stat.ME↗
arXiv2013

Network-based multivariate gene-set testing

The identification of predefined groups of genes ("gene-sets") which are differentially expressed between two conditions ("gene-set analysis", or GSA) is a very popular analysis in bioinformatics. GSA incorporates biological knowledge by aggregating over genes that are believed to be functionally related. This can enhance statistical power over analyses that consider only one gene at a time. However, currently available GSA approaches are all based on univariate two-sample comparison of single genes. This means that they cannot test for differences in covariance structure between the two conditions. Yet interplay between genes is a central aspect of biological investigation and it is likely that such interplay may differ between conditions. This paper proposes a novel approach for gene-set analysis that allows for truly multivariate hypotheses, in particular differences in gene-gene networks between conditions. Testing hypotheses concerning networks is challenging due the nature of the underlying estimation problem. Our starting point is a recent, general approach for high-dimensional two-sample testing. We refine the approach and show how it can be used to perform multivariate, network-based gene-set testing. We validate the approach in simulated examples and show results using high-throughput data from several studies in cancer biology.

stat.ME↗
arXiv2013

Network Inference Using Steady State Data and Goldbeter-Koshland Kinetics

Network inference approaches are widely used to shed light on regulatory interplay between molecular players such as genes and proteins. Biochemical processes underlying networks of interest (e.g. gene regulatory or protein signalling networks) are generally nonlinear. In many settings, knowledge is available concerning relevant chemical kinetics. However, existing network inference methods for continuous, steady-state data are typically rooted in statistical formulations, which do not exploit chemical kinetics to guide inference. Herein, we present an approach to network inference for steady-state data that is rooted in non-linear descriptions of biochemical mechanism. We use equilibrium analysis of chemical kinetics to obtain functional forms that are in turn used to infer networks using steady-state data. The approach we propose is directly applicable to conventional steady-state gene expression or proteomic data and does not require knowledge of either network topology or any kinetic parameters. We illustrate the approach in the context of protein phosphorylation networks, using data simulated from a recent mechanistic model and proteomic data from cancer cell lines. In the former, the true network is known and used for assessment, whereas in the latter, results are compared against known biochemistry. We find that the proposed methodology is more effective at estimating network topology than methods based on linear models.

stat.AP↗
arXiv2013

Two-Sample Testing in High-Dimensional Models

We propose novel methodology for testing equality of model parameters between two high-dimensional populations. The technique is very general and applicable to a wide range of models. The method is based on sample splitting: the data is split into two parts; on the first part we reduce the dimensionality of the model to a manageable size; on the second part we perform significance testing (p-value calculation) based on a restricted likelihood ratio statistic. Assuming that both populations arise from the same distribution, we show that the restricted likelihood ratio statistic is asymptotically distributed as a weighted sum of chi-squares with weights which can be efficiently estimated from the data. In high-dimensional problems, a single data split can result in a "p-value lottery". To ameliorate this effect, we iterate the splitting process and aggregate the resulting p-values. This multi-split approach provides improved p-values. We illustrate the use of our general approach in two-sample comparisons of high-dimensional regression models ("differential regression") and graphical models ("differential network"). In both cases we show results on simulated data as well as real data from recent, high-throughput cancer studies.

stat.ME↗
arXiv2013

Network-based clustering with mixtures of L1-penalized Gaussian graphical models: an empirical investigation

In many applications, multivariate samples may harbor previously unrecognized heterogeneity at the level of conditional independence or network structure. For example, in cancer biology, disease subtypes may differ with respect to subtype-specific interplay between molecular components. Then, both subtype discovery and estimation of subtype-specific networks present important and related challenges. To enable such analyses, we put forward a mixture model whose components are sparse Gaussian graphical models. This brings together model-based clustering and graphical modeling to permit simultaneous estimation of cluster assignments and cluster-specific networks. We carry out estimation within an L1-penalized framework, and investigate several specific penalization regimes. We present empirical results on simulated data and provide general recommendations for the formulation and use of mixtures of L1-penalized Gaussian graphical models.

stat.ML↗
arXiv2012

Network inference and biological dynamics

Network inference approaches are now widely used in biological applications to probe regulatory relationships between molecular components such as genes or proteins. Many methods have been proposed for this setting, but the connections and differences between their statistical formulations have received less attention. In this paper, we show how a broad class of statistical network inference methods, including a number of existing approaches, can be described in terms of variable selection for the linear model. This reveals some subtle but important differences between the methods, including the treatment of time intervals in discretely observed data. In developing a general formulation, we also explore the relationship between single-cell stochastic dynamics and network inference on averages over cells. This clarifies the link between biochemical networks as they operate at the cellular level and network inference as carried out on data that are averages over populations of cells. We present empirical results, comparing thirty-two network inference methods that are instances of the general formulation we describe, using two published dynamical models. Our investigation sheds light on the applicability and limitations of network inference and provides guidance for practitioners and suggestions for experimental design.

stat.AP↗
arXiv2012

On the relationship between ODEs and DBNs

Recently, Li et al. (Bioinformatics 27(19), 2686-91, 2011) proposed a method, called Differential Equation-based Local Dynamic Bayesian Network (DELDBN), for reverse engineering gene regulatory networks from time-course data. We commend the authors for an interesting paper that draws attention to the close relationship between dynamic Bayesian networks (DBNs) and differential equations (DEs). Their central claim is that modifying a DBN to model Euler approximations to the gradient rather than expression levels themselves is beneficial for network inference. The empirical evidence provided is based on time-course data with equally-spaced observations. However, as we discuss below, in the particular case of equally-spaced observations, Euler approximations and conventional DBNs lead to equivalent statistical models that, absent artefacts due to the estimation procedure, yield networks with identical inter-gene edge sets. Here, we discuss further the relationship between DEs and conventional DBNs and present new empirical results on unequally spaced data which demonstrate that modelling Euler approximations in a DBN can lead to improved network reconstruction.

stat.AP↗