Investigating changes and similarities in brain connectivity networks across task conditions is a central topic in neuroscience. We propose a novel framework for jointly estimating multiple graphical models using a hybrid Bayesian integration technique that can handle high-dimensional or large-scale data sets. This framework accommodates multiple graphical models and performs a series of conditional independence tests to infer the underlying network structures. Theoretical justification for consistency is established, and synthetic experiments demonstrate that our approach outperforms existing methods in both accuracy and robustness. We further apply this framework to an functional magnetic resonance imaging study of dynamic functional connectivity among regions of interest during an emotion-processing task. Results reveal that inter- and intra-module interactions involving the subcortical–cerebellum module are reduced during emotion processing compared to shape processing, highlighting this module’s key role in emotional processing.
OgawaSLeeTMKayAR, et al.Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA1990; 87: 9868–9872.
2.
LukemireJKunduSPagnoniG, et al.Bayesian joint modeling of multiple brain functional networks. J Am Stat Assoc2021; 116: 518–530.
3.
BullmoreESpornsO. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci2009; 10: 186–198.
4.
FristonKJ. Functional and effective connectivity in neuroimaging: a synthesis. Hum Brain Mapp1994; 2: 56–78.
5.
HorwitzB. The elusive concept of brain connectivity. Neuroimage2003; 19: 466–470.
6.
FristonKJ. Functional and effective connectivity: a review. Brain Connect2011; 1: 13–36.
7.
BiswalBBYetkinFZHaughtonV, et al.Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med1995; 34: 537–541.
8.
CerlianiLMennesMThomasRM, et al.Increased functional connectivity between subcortical and cortical resting-state networks in autism spectrum disorder. JAMA Psychiatry2015; 72: 767–777.
9.
ZhangAFangJLiangF, et al.Aberrant brain connectivity in schizophrenia detected via a fast gaussian graphical model. IEEE J Biomed Health Inform2018; 23: 1479–1489.
10.
FriedmanJHastieTTibshiraniR. Sparse inverse covariance estimation with the graphical lasso. Biostatistics2008; 9: 432–441.
11.
VaroquauxGGramfortAPolineJB, et al.Brain covariance selection: better individual functional connectivity models using population prior. NIPS2010; 23: 1–9.
12.
HinneMAmbrogioniLJanssenRJ, et al.Structurally-informed Bayesian functional connectivity analysis. NeuroImage2014; 86: 294–305.
13.
ChangCGloverGH. Time–frequency dynamics of resting-state brain connectivity measured with fMRI. NeuroImage2010; 50: 81–98.
14.
HutchisonRMWomelsdorfTAllenEA, et al.Dynamic functional connectivity: promise, issues, and interpretations. Neuroimage2013; 80: 360–378.
15.
ZhangJLiXLiC, et al.Inferring functional interaction and transition patterns via dynamic Bayesian variable partition models. Hum Brain Mapp2014; 35: 3314–3331.
16.
ChiangSCasseseAGuindaniM, et al.Time-dependence of graph theory metrics in functional connectivity analysis. NeuroImage2016; 125: 601–615.
17.
WarnickRGuindaniMErhardtEB, et al.A Bayesian approach for estimating dynamic functional network connectivity in fMRI data. J Am Stat Assoc2018; 113: 134–151.
18.
JiangFJinHGaoY, et al.Time-varying dynamic network model for dynamic resting state functional connectivity in fMRI and MEG imaging. NeuroImage2022; 254: 119131.
ColcloughGLWoolrichMWHarrisonSJ, et al.Multi-subject hierarchical inverse covariance modelling improves estimation of functional brain networks. NeuroImage2018; 178: 370–384.
23.
LiangFSongQQiuP. An equivalent measure of partial correlation coefficients for high dimensional gaussian graphical models. J Am Stat Assoc2015; 110: 1248–1265.
24.
MeinshausenNBühlmannP. High-dimensional graphs and variable selection with the lasso. Ann Stat2006; 34: 1436–1462.
25.
YuanMLinY. Model selection and estimation in the gaussian graphical model. Biometrika2007; 94: 19–35.
26.
MazumderRHastieT. The graphical lasso: new insights and alternatives. Electron J Stat2012; 6: 2125.
27.
GuglielminiSClaeskensG. Asymptotic post-selection inference for regularized graphical models. Stat Comput2025; 35: 36.
WangH. Scaling it up: stochastic search structure learning in graphical models. Bayesian Anal2015; 10: 351–377.
30.
RoveratoA. Hyper inverse Wishart distribution for non-decomposable graphs and its application to Bayesian inference for gaussian graphical models. Scand J Stat2002; 29: 391–411.
31.
WangHLiSZ. Efficient gaussian graphical model determination under G-Wishart prior distributions. Electron J Stat2012; 6: 168–198.
RočkováVGeorgeEI. The spike-and-slab lasso. J Am Stat Assoc2018; 113: 431–444.
36.
GanLNarisettyNNLiangF. Bayesian regularization for graphical models with unequal shrinkage. J Am Stat Assoc2018; 114: 1218–1231.
37.
AtchadéYF. Quasi-Bayesian estimation of large gaussian graphical models. J Multivar Anal2019; 173: 656–671.
38.
YangXGanLNarisettyNN, et al.GemBag: group estimation of multiple Bayesian graphical models. J Mach Learn Res2021; 22: 1–48.
39.
GuoJLevinaEMichailidisG, et al.Joint estimation of multiple graphical models. Biometrika2011; 98: 1–15.
40.
DanaherPWangPWittenDM. The joint graphical Lasso for inverse covariance estimation across multiple classes. J R Stat Soc Ser B: Stat Methodol2014; 76: 373–397.
ChungFRKLuL. The average distances in random graphs with given expected degrees. Proc Natl Acad Sci2002; 99: 15879–15882.
43.
PetersonCBStingoFCVannucciM. Bayesian inference of multiple gaussian graphical models. J Am Stat Assoc2015; 110: 159–174.
44.
TanLSJasraADe IorioM, et al.Bayesian inference for multiple gaussian graphical models with application to metabolic association networks. Ann Appl Stat2017; 11: 2222–2251.
45.
GanLYangXNarisettyNN, et al.Bayesian joint estimation of multiple graphical models. In NeurIPS. Vol. 32. pp.1–11. Vancouver, Canada: Neural Information Processing Systems Foundation, 2019.
46.
LingjærdeCFairfaxBPRichardsonS, et al.Scalable multiple network inference with the joint graphical horseshoe. Ann Appl Stat2024; 18: 1899–1923.
47.
ZhangYYangY. Joint estimation for multisource gaussian graphical models based on transfer learning. Pattern Recognit2025; 158: 110964.
FanJLvJ. Sure independence screening for ultrahigh dimensional feature space. J R Stat Soc B (Stat Methodol)2008; 70: 849–911.
50.
LiangFJiaBXueJ, et al.An imputation–regularized optimization algorithm for high dimensional missing data problems and beyond. J R Stat Soc B (Stat Methodol)2018; 80: 899–926.
51.
LinTILeeJCYenSY. Finite mixture modelling using the skew normal distribution. Stat Sin2007; 17: 909–927.
52.
JiaBLiangF and TEDDY Study Group. Fast hybrid Bayesian integrative learning of multiple gene regulatory networks for type 1 diabetes. Biostatistics2021; 22: 233–249.
53.
JiaBLiangF. Joint estimation of multiple mixed graphical models for pan-cancer network analysis. Stat2020; 9: e271.
54.
LiuHLaffertyJWassermanL. The nonparanormal: semiparametric estimation of high dimensional undirected graphs. J Mach Learn Res2009; 10: 2295–2328.
55.
YuZPradoRQuinlanEB, et al.Understanding the impact of stroke on brain motor function: a hierarchical Bayesian approach. J Am Stat Assoc2016; 111: 549–563.
56.
ZhangLGuindaniMVannucciM. Bayesian models for functional magnetic resonance imaging data analysis. Wiley Interdiscip Rev Comput Stat2015; 7: 21–41.
57.
RyaliSSupekarKChenT, et al.Multivariate dynamical systems models for estimating causal interactions in fMRI. NeuroImage2010; 54: 807–823.
58.
CastruccioSOmbaoHGentonMG. A scalable multi-resolution spatio-temporal model for brain activation and connectivity in fMRI data. Biometrics2018; 74: 823–833.
59.
MarkettSAJawinskiPKirschP, et al.Specific and segregated changes to the functional connectome evoked by the processing of emotional faces: a task-based connectome study. Sci Rep2020; 10: 4822.
60.
FoxMDRaichleME. Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nat Rev Neurosci2007; 8: 700–711.
61.
ChenJERubinovMChangC. Methods and considerations for dynamic analysis of functional mr imaging data. Neuroimaging Clin N Am2017; 27: 547–560.
62.
FanJSongR. Sure independence screening in generalized linear models with NP-dimensionality. Ann Stat2010; 38: 3567–3604.
63.
CeleuxGChauveauDDieboltJ. Stochastic versions of the EM algorithm: an experimental study in the mixture case. J Stat Comput Simul1996; 55: 287–314.
64.
NielsenS. The stochastic EM algorithm: estimation and asymptotic results. Bernoulli2000; 6: 457–489.
65.
LiangFZhangJ. Estimating the false discovery rate using the stochastic approximation algorithm. Biometrika2008; 95: 961–977.
66.
PelletJPElisseeffA. Using Markov blankets for causal structure learning. J Mach Learn Res2008; 9: 1295–1342.
67.
XuSJiaBLiangF. Learning moral graphs in construction of high-dimensional Bayesian networks for mixed data. Neural Comput2019; 31: 1183–1214.
68.
SunLZhangALiangF. Time-varying dynamic Bayesian network learning for an fMRI study of emotion processing. Stat Med2024; 43: 2713–2733.
69.
ZaykinDV. Optimally weighted z-test is a powerful method for combining probabilities in meta-analysis. J Evol Biol2011; 24: 1836–1841.
70.
StoufferSSuchmanEDeVinneyLC, et al.The American soldier: adjustment during army life. Vol. 1. Princeton: Princeton University Press, 1949.
71.
OwenAB. Karl Pearson’s meta analysis revisited. Qual Eng2009; 55: 493–494.
72.
PoundsSChengC. Robust estimation of the false discovery rate. Bioinformatics2006; 22: 1979–1987.
73.
EfronB. Large-scale simultaneous hypothesis testing. J Am Stat Assoc2004; 99: 104–196.
74.
EfronBTibshiraniR. Empirical Bayes methods and false discovery rates for microarrays. Genet Epidemiol2002; 23: 70–86.
75.
EfronBTibshiraniRStoreyJD, et al.Empirical Bayes analysis of a microarray experiment. J Am Stat Assoc2001; 96: 1151–1160.
76.
FanJLiR. Variable selection via nonconcave penalized likelihood and its oracle properties. J Am Stat Assoc2001; 96: 1348–1360.
77.
Van EssenDCSmithSMBarchDM, et al.The WU-Minn human connectome project: an overview. Neuroimage2013; 80: 62–79.
78.
HaririARBrownSMWilliamsonDE, et al.Preference for immediate over delayed rewards is associated with magnitude of ventral striatal activity. J Neurosci2006; 26: 13213–13217.
79.
GlasserMFSotiropoulosSNWilsonJA, et al.The minimal preprocessing pipelines for the human connectome project. Neuroimage2013; 80: 105–124.
80.
FinnESShenXScheinostD, et al.Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity. Nat Neurosci2015; 18: 1664–1671.
81.
CaiBZhangGZhangA, et al.Functional connectome fingerprinting: identifying individuals and predicting cognitive functions via autoencoder. Hum Brain Mapp2021; 42: 2691–2705.
82.
ZigmondMJBloomFELandisSC, et al.Fundamental neuroscience. Vol. 207. San Diego: Academic Press, 1999.
83.
RajmohanVMohandasE. The limbic system. Indian J Psychiatry2007; 49: 132.
84.
FerrucciRGiannicolaGRosaM, et al.Cerebellum and processing of negative facial emotions: cerebellar transcranial dc stimulation specifically enhances the emotional recognition of facial anger and sadness. Cogn Emot2012; 26: 786–799.
85.
BaumannOMattingleyJB. Functional topography of primary emotion processing in the human cerebellum. NeuroImage2012; 61: 805–811.
86.
AdamaszekMD’AgataFFerrucciR, et al.Consensus paper: cerebellum and emotion. Cerebellum2017; 16: 552–576.
87.
GoldAKToomeyR. The role of cerebellar impairment in emotion processing: a case study. Cerebellum Ataxias2018; 5: 1–9.
88.
MeinshausenNBuhlmannP. High-dimensional graphs and variable selection with the lasso. Ann Statist2006; 34: 1436–1462.
