Abstract
A differential network is an important tool for capturing the changes in conditional correlations under two sample cases. In this paper, we introduce a fast iterative algorithm to recover the differential network for high-dimensional data. The computational complexity of our algorithm is linear in the sample size and the number of parameters, which is optimal in that it is of the same order as computing two sample covariance matrices. The proposed method is appealing for high-dimensional data with a small sample size. The experiments on simulated and real datasets show that the proposed algorithm outperforms other existing methods.
Similar content being viewed by others
References
Anderson T (2003) An introduction to multivariate statistical analysis. Wiley series in probability and statistics. Wiley, New York
Bandyopadhyay S, Mehta M, Kuo D, Sung MK, Chuang R, Jaehnig EJ, Bodenmiller B, Licon K, Copeland W, Shales M et al (2010) Rewiring of genetic networks in response to DNA damage. Science 330(6009):1385–1389
Barabási AL, Oltvai ZN (2004) Network biology: understanding the cell’s functional organization. Nat Rev Genet 5(2):101
Barabási AL, Gulbahce N, Loscalzo J (2011) Network medicine: a network-based approach to human disease. Nat Rev Genet 12(1):56
Beck A, Teboulle M (2009) A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM J Imaging Sci 2(1):183–202
Bickel PJ, Levina E (2008) Regularized estimation of large covariance matrices. Ann Stat 36(1):199–227
Boyd S, Parikh N, Chu E, Peleato B, Eckstein J (2011) Distributed optimization and statistical learning via the alternating direction method of multipliers. Found Trends Mach Learn 3(1):1–122
Cai T, Liu W (2011) Adaptive thresholding for sparse covariance matrix estimation. J Am Stat Assoc 106(494):672–684
Cai T, Zhang L (2018) A convex optimization approach to high-dimensional sparse quadratic discriminant analysis. Ann Stat (submitted)
Cai T, Liu W, Luo X (2011) A constrained \(\ell _1\) minimization approach to sparse precision matrix estimation. J Am Stat Assoc 106(494):594–607
Ding X, Yang Y, Han B, Du C, Xu N, Huang H, Cai T, Zhang A, Han ZG, Zhou W, Chen L (2014) Transcriptomic characterization of hepatocellular carcinoma with ctnnb1 mutation. PLoS ONE 9(5):e95307
Fan J, Liao Y, Liu H (2016) An overview of the estimation of large covariance and precision matrices. Econom J 19(1):C1–C32
Gambardella G, Moretti MN, De Cegli R, Cardone L, Peron A, Di Bernardo D (2013) Differential network analysis for the identification of condition-specific pathway activity and regulation. Bioinformatics 29(14):1776–1785
Gambardella G, Peluso I, Montefusco S, Bansal M, Medina DL, Lawrence N, Bernardo DD (2015) A reverse-engineering approach to dissect post-translational modulators of transcription factors activity from transcriptional data. BMC Bioinform 16(1):279
Gambardella G, Carissimo A, Chen A et al (2017) The impact of micrornas on transcriptional heterogeneity and gene co-expression across single embryonic stem cells. Nat Commun 8:14126
Hsiao TH, Chiu YC, Hsu PY, Lu TP, Lai LC, Tsai MH, Huang THM, Chuang EY, Chen Y (2016) Differential network analysis reveals the genome-wide landscape of estrogen receptor modulation in hormonal cancers. Sci Rep 6:23035
Ideker T, Krogan NJ (2014) Differential network biology. Mol Syst Biol 8(1):565
Jerome F, Trevor H, Robert T (2008) Sparse inverse covariance estimation with the graphical lasso. Biostatistics 9(3):432–441
Jian G, Elizaveta L, George M, Ji Z (2011) Joint estimation of multiple graphical models. Biometrika 98(1):1–15
Jiang B, Wang X, Leng C (2018) A direct approach for sparse quadratic discriminant analysis. J Mach Learn Res 19(31):1–37
Julien C, Yves G, Christophe A (2011) Inferring multiple graphical structure. Stat Comput 21(4):537–553
Kaushik A, Ali S, Gupta D (2017) Altered pathway analyzer: a gene expression dataset analysis tool for identification and prioritization of differentially regulated and network rewired pathways. Sci Rep 7:40450
Li Q, Shao J (2015) Sparse quadratic discriminant analysis for high dimensional data. Stat Sin 25:457–473
Liu H, Lafferty J, Wasserman L (2009) The nonparanormal: semiparametric estimation of high dimensional undirected graphs. J Mach Learn Res 10:2295–2328
Meinshausen N, Bühlmann P (2006) High-dimensional graphs and variable selection with the lasso. Ann Stat 34(3):1436–1462
Nesterov Y (1983) A method for solving the convex programming problem with convergence rate \(O(k^2)\). Soviet Math Dokl 27:372–376
Rothman AJ, Levina E, Zhu J (2009) Generalized thresholding of large covariance matrices. J Am Stat Assoc 104(485):177–186
Tibshirani R (1996) Regression shrinkage and selection via the lasso. J R Stat Soc Ser B 58(1):267–288
Tong T, Wang C, Wang Y (2014) Estimation of variances and covariances for high-dimensional data: a selective review. Comput Stat 6(4):255–264
Wu MY, Zhang XF, Dai DQ, Le OY, Zhu Y, Yan H (2016) Regularized logistic regression with network-based pairwise interaction for biomarker identification in breast cancer. BMC Bioinform 17(1):108
Xue L, Zou H (2012) Regularized rank-based estimation of high-dimensional nonparanormal graphical models. Ann Stat 40(5):2541–2571
Yuan H, Xi R, Chen C, Deng M (2017) Differential network analysis via the lasso penalized D-trace loss. Biometrika 104(4):755–770
Zhang T, Zou H (2014) Sparse precision matrix estimation via lasso penalized D-trace loss. Biometrika 101(1):103–120
Zhao SD, Cai TT, Li H (2014) Direct estimation of differential networks. Biometrika 101(2):253–268
Zhu Y, Li L (2018) Multiple matrix gaussian graphs estimation. J R Stat Soc Ser B 80:927–950
Zou H, Li R (2008) One-step sparse estimates in nonconcave penalized likelihood models. Ann Stat 36(4):1509–1533
Acknowledgements
We thank two reviewers, an associate editor, and the editor for their most helpful comments. Yu was supported in part by National Natural Science Foundation of China 11671256, 2016YFC0902403 of the Chinese Ministry of Science and Technology and Neil Shen’s SJTU Medical Research Fund. Wang was partially supported by the Shanghai Sailing Program 16YF1405700 and the National Natural Science Foundation of China 11701367 and 11825104.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
According to the main results of Beck and Teboulle (2009), to complete the proof of Theorem 1, we need to show only that the loss function \(L_1(\varDelta )\) is convex, which is the result of the following lemma.
Lemma 1.1 The loss function (2.3) is a smooth convex function, and its gradient is Lipschitz continuous with Lipschitz constant \(L=\lambda _{\max }(S_1) \lambda _{\max }(S_2)\), that is
where \(\lambda _{\max }(S_i)\) is the largest eigenvalue of the sample covariance matrix \(S_i\) for \(i=1,2\).
Proof
Because the loss function (2.3) is defined by
We can calculate the gradient of \(L_1(\varDelta )\)
and the Hessian matrix is \(S_2 \otimes S_1\). Because both covariance matrices \(S_1\) and \(S_2\) are definite positive matrices, the Hessian matrix is a definite positive matrix. Hence, the loss function \(L_1(\varDelta )\) is a smooth convex function.
Moreover, for any \(\varDelta _1, \varDelta _2 \in \text {dom}(\nabla L_1)\), we have
The proof is now complete. \(\square \)
Rights and permissions
About this article
Cite this article
Tang, Z., Yu, Z. & Wang, C. A fast iterative algorithm for high-dimensional differential network. Comput Stat 35, 95–109 (2020). https://doi.org/10.1007/s00180-019-00915-w
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00180-019-00915-w