Learning Gaussian Bayesian Network from Censored Data Subject to Limit of Detection by the Structural EM Algorithm
Abstract
:1. Introduction
1.1. Structure Learning for Bayesian Networks
1.2. Related Work on Censored Data Subject to Limit of Detection
1.3. Our Contributions
2. Censored Gaussian Bayesian Network
3. Structural Learning of Censored GBN
3.1. EBIC for Censored GBN
3.2. Censored Structural EM Algorithm
Algorithm 1: The structural EM (SEM) algorithm for structural learning |
|
3.2.1. E-Step
3.2.2. M-Step
3.2.3. Optional Parameter-Learning Step
3.2.4. Initial Values of cSEM
4. Simulation
4.1. Evaluating MC Sampling in E-Step
4.2. Comparison of Methods on Data Simulated from Censored GBN
4.3. Additional Simulations on Artificially Censored Biological Data
5. Structural Learning of Single Cell Data
6. Conclusions
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
Appendix A. Detailed Derivations in M-Step
Appendix B. Average CPU Time
c = 0.25 | c = 0.5 | com | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
EM.update | E.step | M.step | Total | Iter | EM.update | E.step | M.step | Total | Iter | Total | |||
p = 50 | n = 200 | csem50.ges.emup | 11.38 | 10.19 | 131.10 | 153.03 | 17.08 | 26.61 | 22.86 | 153.96 | 204.02 | 20.08 | |
csem50.imcmc.emup | 12.19 | 11.05 | 141.25 | 167.18 | 19.30 | 23.57 | 21.31 | 138.31 | 186.13 | 19.98 | |||
csem50.ges | 0.00 | 13.60 | 178.98 | 192.94 | 22.78 | 0.00 | 32.38 | 213.27 | 246.24 | 29.54 | |||
csem50.imcmc | 0.00 | 13.08 | 173.22 | 189.04 | 22.32 | 0.00 | 32.74 | 213.42 | 249.11 | 30.66 | |||
ges | 0.10 | 0.08 | 0.16 | ||||||||||
pc | 0.08 | 0.06 | 0.09 | ||||||||||
mmhc | 0.10 | 0.10 | 0.10 | ||||||||||
iMCMC | 7.76 | 7.57 | 7.53 | ||||||||||
n = 500 | csem50.ges.emup | 26.98 | 24.12 | 126.75 | 178.43 | 15.60 | 64.30 | 52.64 | 146.22 | 264.27 | 18.52 | ||
csem50.imcmc.emup | 27.45 | 24.65 | 132.85 | 187.85 | 17.00 | 62.33 | 54.44 | 152.43 | 272.63 | 19.82 | |||
csem50.ges | 0.00 | 29.68 | 157.77 | 188.06 | 19.74 | 0.00 | 86.22 | 245.03 | 332.40 | 30.44 | |||
csem50.imcmc | 0.00 | 31.82 | 171.69 | 206.50 | 21.60 | 0.00 | 80.91 | 232.69 | 317.17 | 29.04 | |||
ges | 0.12 | 0.12 | 0.17 | ||||||||||
pc | 0.15 | 0.16 | 0.14 | ||||||||||
mmhc | 0.14 | 0.14 | 0.15 | ||||||||||
iMCMC | 9.09 | 10.49 | 7.87 | ||||||||||
p = 100 | n = 200 | csem50.ges.emup | 22.89 | 20.56 | 658.80 | 715.40 | 17.34 | 126.91 | 103.92 | 695.10 | 929.27 | 17.78 | |
csem50.imcmc.emup | 22.72 | 20.40 | 600.17 | 655.79 | 17.34 | 105.08 | 88.42 | 577.90 | 785.19 | 16.10 | |||
csem50.ges | 0.00 | 22.58 | 694.62 | 718.74 | 18.74 | 0.00 | 163.38 | 1107.90 | 1274.56 | 28.90 | |||
csem50.imcmc | 0.00 | 25.05 | 765.48 | 803.26 | 21.44 | 0.00 | 145.89 | 984.50 | 1144.85 | 26.42 | |||
ges | 0.43 | 0.40 | 0.50 | ||||||||||
pc | 0.19 | 0.15 | 0.21 | ||||||||||
mmhc | 0.33 | 0.33 | 0.33 | ||||||||||
iMCMC | 36.60 | 38.54 | 35.43 | ||||||||||
n = 500 | csem50.ges.emup | 45.15 | 39.43 | 494.57 | 581.10 | 12.88 | 126.91 | 103.92 | 695.10 | 929.27 | 17.78 | ||
csem50.imcmc.emup | 43.27 | 37.66 | 473.67 | 567.26 | 13.10 | 105.08 | 88.42 | 577.90 | 785.19 | 16.10 | |||
csem50.ges | 0.00 | 54.44 | 700.19 | 756.62 | 18.26 | 0.00 | 163.38 | 1107.90 | 1274.56 | 28.90 | |||
csem50.imcmc | 0.00 | 57.86 | 754.20 | 825.35 | 19.96 | 0.00 | 145.89 | 984.50 | 1144.85 | 26.42 | |||
ges | 0.51 | 0.56 | 0.53 | ||||||||||
pc | 0.39 | 0.38 | 0.38 | ||||||||||
mmhc | 0.45 | 0.44 | 0.45 | ||||||||||
iMCMC | 41.80 | 53.31 | 36.43 |
References
- Needham, C.J.; Bradford, J.R.; Bulpitt, A.J.; Westhead, D.R. BA primer on learning in Bayesian networks for computational biology. PLoS Comput. Biol. 2007, 3, e129. [Google Scholar] [CrossRef] [PubMed]
- Agrahari, R.; Foroushani, A.; Docking, T.R.; Chang, L.; Duns, G.; Hudoba, M.; Karsan, A.; Zare, H. Applications of Bayesian network models in predicting types of hematological malignancies. Sci. Rep. 2018, 8, 6951. [Google Scholar] [CrossRef]
- Pearl, J. Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference, 2nd ed.; Morgan Kaufmann: San Francisco, CA, USA, 1988. [Google Scholar]
- Zhang, H.; Jiang, L.; Zhang, W.; Li, C. Multi-view attribute weighted naive Bayes. IEEE Trans. Knowl. Data Eng. 2023, 35, 7291–7302. [Google Scholar] [CrossRef]
- Zhang, H.; Jiang, L.; Yu, L. Attribute and instance weighted naive Bayes. Pattern Recognit. 2021, 111, 107674. [Google Scholar] [CrossRef]
- Liu, F.; Zhang, S.W.; Guo, W.F.; Wei, Z.G.; Chen, L. Inference of gene regulatory network based on local Bayesian networks. PLoS Comput. Biol. 2016, 12, e1005024. [Google Scholar] [CrossRef] [PubMed]
- Han, S.W.; Chen, G.; Cheon, M.S.; Zhong, H. Estimation of directed acyclic graphs through two-stage adaptive lasso for gene network inference. J. Am. Stat. Assoc. 2016, 111, 1004–1019. [Google Scholar] [CrossRef]
- Kitson, N.K.; Constantinou, A.C.; Guo, Z.; Liu, Y.; Chobtham, K. A survey of Bayesian network structure learning. Artif. Intell. Rev. 2023, 56, 8721–8814. [Google Scholar] [CrossRef]
- Spirtes, P.; Glymour, C. An algorithm for fast recovery of sparse causal graphs. Soc. Sci. Comput. Rev. 1991, 9, 62–72. [Google Scholar] [CrossRef]
- Colombo, D.; Maathuis, M.H. Order-independent constraint-based causal structure learning. J. Mach. Learn. Res. 2014, 15, 3921–3962. [Google Scholar]
- Colombo, D.; Maathuis, M.H.; Kalisch, M.; Richardson, T.S. Learning high-dimensional directed acyclic graphs with latent and selection variables. Ann. Stat. 2012, 40, 294–321. [Google Scholar] [CrossRef]
- Marella, D.; Vicard, P. Bayesian network structural learning from complex survey data: A resampling based approach. Stat. Methods Appl. 2022, 31, 981–1013. [Google Scholar] [CrossRef]
- Chickering, D.M. Optimal structure identification with greedy search. J. Mach. Learn. Res. 2002, 3, 507–554. [Google Scholar]
- Moffa, G. Partition MCMC for inference on acyclic digraphs. J. Am. Stat. Assoc. 2017, 112, 282–299. [Google Scholar]
- Tsamardinos, I.; Brown, L.E.; Aliferis, C.F. The max-min hill-climbing Bayesian network structure learning algorithm. Mach. Learn. 2006, 65, 31–78. [Google Scholar]
- Gasse, M.; Aussem, A.; Elghazel, H. A hybrid algorithm for Bayesian network structure learning with application to multi-label learning. Expert Syst. Appl. 2014, 41, 6755–6772. [Google Scholar] [CrossRef]
- Kuipers, J.; Suter, P.; Moffa, G. Efficient sampling and structure learning of Bayesian networks. J. Comput. Graph. Stat. 2022, 31, 639–650. [Google Scholar] [CrossRef]
- Suter, P.; Kuipers, J.; Moffa, G.; Beerenwinkel, N. Bayesian structure learning and sampling of Bayesian networks with the R Package BiDAG. J. Stat. Softw. 2023, 105, 1–31. [Google Scholar] [CrossRef]
- Song, J.; Barnhart, H.X.; Lyles, R.H. A GEE approach for estimating correlation coefficients involving left-censored variables. J. Data Sci. 2004, 2, 245–257. [Google Scholar] [CrossRef]
- Lee, G.; Scott, C. EM algorithms for multivariate Gaussian mixture models with truncated and censored data. Comput. Stat. Data Anal. 2012, 56, 2816–2829. [Google Scholar]
- Perkins, N.J.; Schisterman, E.F.; Vexler, A. Multivariate normally distributed biomarkers subject to limits of detection and receiver operating characteristic curve inference. Acad. Radiol. 2013, 20, 838–846. [Google Scholar] [CrossRef]
- Hoffman, H.J.; Johnson, R.E. Pseudo-likelihood estimation of multivariate normal parameters in the presence of left-censored data. J. Agric. Biol. Environ. Stat. 2015, 20, 156–171. [Google Scholar]
- Jones, M.P.; Perry, S.S.; Thorne, P.S. Maximum pairwise pseudo-likelihood estimation of the covariance matrix from left-censored data. J. Agric. Biol. Environ. Stat. 2015, 20, 83–99. [Google Scholar] [PubMed]
- Pesonen, M.; Pesonen, H.; Nevalainen, J. Covariance matrix estimation for left-censored data. Comput. Stat. Data Anal. 2015, 92, 13–25. [Google Scholar]
- Augugliaro, L.; Abbruzzo, A.; Vinciotti, V. ℓ1-penalized censored Gaussian graphical model. Biostatistics 2020, 21, e1–e16. [Google Scholar]
- Augugliaro, L.; Sottile, G.; Vinciotti, V. The conditional censored graphical lasso estimator. Stat. Comput. 2020, 30, 1273–1289. [Google Scholar]
- Sottile, G.; Augugliaro, L.; Vinciotti, V. The joint censored Gaussian graphical lasso model. In SIS 2022 Book of the Short Papers; Università degli Studi di Palermo: Palermo, Italy, 2022; pp. 1829–1834. [Google Scholar]
- Koller, D.; Friedman, N. Probabilistic Graphical Models: Principles and Techniques; The MIT Press: London, UK, 2009. [Google Scholar]
- Geiger, D.; Heckerman, D. Parameter priors for directed acyclic graphical models and the characterization of several probability distributions. Ann. Stat. 2002, 30, 1412–1440. [Google Scholar] [CrossRef]
- Moffa, G.; Heckerman, D. Addendum on the scoring of Gaussian directed acyclic graphical models. Ann. Stat. 2014, 42, 1689–1691. [Google Scholar]
- Friedman, N. Learning belief networks in the presence of missing values and hidden variables. In Proceedings of the 14th International Confeerence on Machine Learning, Nashville, TN, USA, 8–12 July 1997; pp. 125–133. [Google Scholar]
- Friedman, N. The Bayesian structural EM algorithm. In Proceedings of the 14th Conference on Uncertainty in Artifical Intelligence, Madison, WI, USA, 24–26 July 1998; pp. 129–138. [Google Scholar]
- Lauritzen, S.L. Graphical Models; Oxford Statistical Sience Series; Clarendon Press: Oxford, UK, 1996. [Google Scholar]
- Shojaie, A.; Michailidis, G. Penalized likelihood methods for estimation of sparse high-dimensional directed acyclic graphs. Biometrika 2010, 97, 519–538. [Google Scholar] [CrossRef]
- Little, R.J.A.; Rubin, D.B. Statistical Analysis with Missing Data, 3rd ed.; John Wiley & Sons: Hoboken, NJ, USA, 2019. [Google Scholar]
- Chen, J.; Chen, Z. Extended Bayesian information criteria for model selection with large model spaces. Biometrika 2008, 95, 759–771. [Google Scholar] [CrossRef]
- Foygel, R.; Drton, M. Extended Bayesian information criteria for Gaussian graphical models. In Proceedings of the 24th International Conference on Neural Information Processing Systems, Vancouver, Canada, 6–9 December 2010; pp. 604–612. [Google Scholar]
- Scutari, M.; Graafland, C.E.; Gutiérrez, J.M. Who learns better Bayesian network structures: Accuracy and speed of structure learning algorithms. Int. J. Approx. Reason. 2019, 115, 235–253. [Google Scholar] [CrossRef]
- Fu, F.; Zhou, Q. Learning sparse causal Gaussian networks with experimental intervention: Regularization and coordinate descent. J. Am. Stat. Assoc. 2013, 108, 288–300. [Google Scholar]
- Aragam, B.; Zhou, Q. Concave penalized estimation of sparse Gaussian Bayesian networks. J. Mach. Learn. Res. 2015, 16, 2273–2328. [Google Scholar]
- Dempster, A.P.; Laird, N.M.; Rubin, D.B. Maximum likelihood from incomplete data via the EM algorithm. J. R. Stat. Soc. Ser. B (Methodol.) 1977, 39, 1–38. [Google Scholar]
- Genz, A.; Bretz, F. Computation of Multivariate Normal and t Probabilities; Springer: Berlin, Germany, 2009. [Google Scholar]
- Wei, G.G.; Tanner, M.A. A Monte Carlo implementation of the EM algorithm and the poor man’s data augmentation algorithms. J. Am. Stat. Assoc. 1990, 85, 699–704. [Google Scholar]
- Wilhelm, S.; Manjunath, B.G. tmvtnorm: Truncated Multivariate Normal and Student t Distribution, version 1.5. 2022. Available online: https://cran.r-project.org/web/packages/tmvtnorm/index.html (accessed on 10 February 2024).
- Luo, X.G.; Moffa, G.; Kuipers, J. Learning Bayesian networks from ordinal data. J. Mach. Learn. Res. 2021, 22, 1–44. [Google Scholar]
- Kalisch, M.; Mächler, M.; Colombo, D.; Maathuis, M.H.; Bühlmann, P. Causal inference using graphical models with the R package pcalg. J. Stat. Softw. 2012, 47, 1–26. [Google Scholar] [CrossRef]
- Kalisch, M.; Bühlman, P. Estimating high-dimensional directed acyclic graphs with the PC-algorithm. J. Mach. Learn. Res. 2007, 8, 613–636. [Google Scholar]
- Scutari, M. Learning Bayesian networks with the bnlearn R Package. J. Stat. Softw. 2010, 35, 1–22. [Google Scholar] [CrossRef]
- Ma, J.; Shojaie, A.; Michailidis, G. Network-based pathway enrichment analysis with incomplete network information. Bioinformatics 2016, 32, 3165–3174. [Google Scholar] [CrossRef]
- Psaila, B.; Barkas, N.; Iskander, D.; Roy, A.; Anderson, S.; Ashley, N.; Caputo, V.S.; Lichtenberg, J.; Loaiza, S.; Bodine, D.M.; et al. Single-cell profiling of human megakaryocyte-erythroid progenitors identifies distinct megakaryocyte and erythroid differentiation pathways. Genome Biol. 2016, 17, 1–19. [Google Scholar] [CrossRef]
- Vowels, M.J.; Camgoz, N.C.; Bowden, R. D’ya like dags? A survey on structure learning and causal discovery. ACM Comput. Surv. 2022, 55, 1–36. [Google Scholar] [CrossRef]
SHD | 0 | 1 | 4 | 5 | 13 | 14 | 15 | 16 |
---|---|---|---|---|---|---|---|---|
K = 50 | 3 | 3 | 10 | 27 | 1 | 2 | 2 | 2 |
K = 200 | 9 | 39 | 2 | |||||
K = 1000 | 1 | 49 | ||||||
K = 2000 | 2 | 48 |
SHD | me | ee | md | ed | rd | MSE | ||
---|---|---|---|---|---|---|---|---|
c = 0.1 | csem.50.ges | 36 | 10 | 12 | 4 | 2 | 8 | 0.003 |
csem.200.ges | 34 | 9 | 9 | 2 | 4 | 10 | 0.002 | |
csem.1k.ges | 35 | 8 | 9 | 5 | 3 | 10 | 0.002 | |
csem.2k.ges | 45 | 10 | 13 | 10 | 2 | 10 | 0.003 | |
csem.50.imcmc | 48 | 10 | 10 | 11 | 3 | 14 | 0.003 | |
csem.200.imcmc | 29 | 8 | 8 | 5 | 2 | 6 | 0.002 | |
csem.1k.imcmc | 34 | 10 | 11 | 5 | 2 | 6 | 0.003 | |
csem.2k.imcmc | 34 | 8 | 10 | 5 | 3 | 8 | 0.002 | |
missForest | 74 | 24 | 23 | 3 | 2 | 22 | 0.016 | |
missknn | 78 | 23 | 24 | 7 | 2 | 22 | 0.013 | |
c = 0.2 | csem.50.ges | 42 | 15 | 15 | 4 | 2 | 6 | 0.006 |
csem.200.ges | 47 | 13 | 15 | 5 | 2 | 12 | 0.006 | |
csem.1k.ges | 38 | 11 | 11 | 6 | 2 | 8 | 0.005 | |
csem.2k.ges | 39 | 13 | 12 | 6 | 2 | 6 | 0.004 | |
csem.50.imcmc | 44 | 13 | 14 | 5 | 2 | 10 | 0.005 | |
csem.200.imcmc | 51 | 17 | 15 | 9 | 2 | 8 | 0.005 | |
csem.1k.imcmc | 45 | 12 | 12 | 11 | 2 | 8 | 0.007 | |
csem.2k.imcmc | 39 | 15 | 14 | 1 | 1 | 8 | 0.008 | |
missForest | 89 | 33 | 31 | 8 | 3 | 14 | 0.040 | |
missknn | 79 | 33 | 26 | 10 | 2 | 8 | 0.035 | |
c = 0.3 | csem.50.ges | 62 | 21 | 18 | 11 | 2 | 10 | 0.010 |
csem.200.ges | 64 | 20 | 18 | 14 | 2 | 10 | 0.009 | |
csem.1k.ges | 75 | 22 | 22 | 11 | 2 | 18 | 0.011 | |
csem.2k.ges | 61 | 19 | 18 | 12 | 2 | 10 | 0.009 | |
csem.50.imcmc | 55 | 25 | 16 | 7 | 1 | 6 | 0.009 | |
csem.200.imcmc | 60 | 28 | 15 | 14 | 1 | 2 | 0.009 | |
csem.1k.imcmc | 62 | 26 | 17 | 14 | 1 | 4 | 0.010 | |
csem.2k.imcmc | 56 | 26 | 17 | 6 | 1 | 6 | 0.009 | |
missForest | 90 | 34 | 35 | 7 | 4 | 10 | 0.059 | |
missknn | 87 | 34 | 30 | 8 | 3 | 12 | 0.056 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Xu, P.-F.; Lin, S.; Zheng, Q.-Z.; Tang, M.-L. Learning Gaussian Bayesian Network from Censored Data Subject to Limit of Detection by the Structural EM Algorithm. Mathematics 2025, 13, 1482. https://doi.org/10.3390/math13091482
Xu P-F, Lin S, Zheng Q-Z, Tang M-L. Learning Gaussian Bayesian Network from Censored Data Subject to Limit of Detection by the Structural EM Algorithm. Mathematics. 2025; 13(9):1482. https://doi.org/10.3390/math13091482
Chicago/Turabian StyleXu, Ping-Feng, Shanyi Lin, Qian-Zhen Zheng, and Man-Lai Tang. 2025. "Learning Gaussian Bayesian Network from Censored Data Subject to Limit of Detection by the Structural EM Algorithm" Mathematics 13, no. 9: 1482. https://doi.org/10.3390/math13091482
APA StyleXu, P.-F., Lin, S., Zheng, Q.-Z., & Tang, M.-L. (2025). Learning Gaussian Bayesian Network from Censored Data Subject to Limit of Detection by the Structural EM Algorithm. Mathematics, 13(9), 1482. https://doi.org/10.3390/math13091482