Joint estimation of multiple related biological networks

Chris J. Oates, James Korkola, Joe Gray, Sach Mukherjee

Research output: Contribution to journalArticle

11 Citations (Scopus)

Abstract

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 de- signs 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.

Original languageEnglish (US)
Pages (from-to)1892-1919
Number of pages28
JournalAnnals of Applied Statistics
Volume8
Issue number3
DOIs
StatePublished - Sep 1 2014

Fingerprint

Biological Networks
Graphical Models
Molecular biology
Bayesian networks
Exchangeability
Dynamic Bayesian Networks
Formulation
Molecular Biology
Proteomics
Deterministic Algorithm
Exact Algorithms
Cells
Cancer
Likely
Simulation Study
Upper bound
Unit
Joint estimation
Line
Cell

Keywords

  • Bayesian network
  • Belief propagation
  • Hierarchical model
  • Information sharing

ASJC Scopus subject areas

  • Statistics, Probability and Uncertainty
  • Modeling and Simulation
  • Statistics and Probability

Cite this

Joint estimation of multiple related biological networks. / Oates, Chris J.; Korkola, James; Gray, Joe; Mukherjee, Sach.

In: Annals of Applied Statistics, Vol. 8, No. 3, 01.09.2014, p. 1892-1919.

Research output: Contribution to journalArticle

Oates, Chris J. ; Korkola, James ; Gray, Joe ; Mukherjee, Sach. / Joint estimation of multiple related biological networks. In: Annals of Applied Statistics. 2014 ; Vol. 8, No. 3. pp. 1892-1919.
@article{8f603f61efaf42d989f4a884718ca064,
title = "Joint estimation of multiple related biological networks",
abstract = "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 de- signs 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.",
keywords = "Bayesian network, Belief propagation, Hierarchical model, Information sharing",
author = "Oates, {Chris J.} and James Korkola and Joe Gray and Sach Mukherjee",
year = "2014",
month = "9",
day = "1",
doi = "10.1214/14-AOAS761",
language = "English (US)",
volume = "8",
pages = "1892--1919",
journal = "Annals of Applied Statistics",
issn = "1932-6157",
publisher = "Institute of Mathematical Statistics",
number = "3",

}

TY - JOUR

T1 - Joint estimation of multiple related biological networks

AU - Oates, Chris J.

AU - Korkola, James

AU - Gray, Joe

AU - Mukherjee, Sach

PY - 2014/9/1

Y1 - 2014/9/1

N2 - 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 de- signs 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.

AB - 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 de- signs 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.

KW - Bayesian network

KW - Belief propagation

KW - Hierarchical model

KW - Information sharing

UR - http://www.scopus.com/inward/record.url?scp=84907838128&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84907838128&partnerID=8YFLogxK

U2 - 10.1214/14-AOAS761

DO - 10.1214/14-AOAS761

M3 - Article

AN - SCOPUS:84907838128

VL - 8

SP - 1892

EP - 1919

JO - Annals of Applied Statistics

JF - Annals of Applied Statistics

SN - 1932-6157

IS - 3

ER -