Performance Research of Clustering Methods for Detecting State Transition Trajectories in Hemoglobin

Abstract

The time-series clustering method is one of unsupervised machine learning techniques that classify time-series data. In this article, we applied three methods to the clustering analysis for 200 molecular dynamics (MD) trajectories of human adult hemoglobin (HbA), and have reported their clustering performances for detecting the T-R state transition trajectories (Traj_T-R). By compared with their silhouette indices, we have discussed the proper clustering conditions.

Figures

Figure 1.

 ED and DTW. (Top) Two time series x and y. (Bottom) The distances between x and y in ED and those in DTW. The terminal points are each linked by a line segment.

Figure 2.

 The silhouette index ${\overline{s}}_k$ depending on the number of clusters partitioning k (k=2 to 5) in three time-series clustering methods, ED, DBA and soft-DTW.

Table

Table 1. The silhouette indices and the numbers of Traj_T-R in three methods, ED, DBA and soft-DTW.

Methods	$\overline{s}$ka	# of TrajT-Rb
This work:
ED	0.38	34 (17%)
DBA	0.37	54 (27%)
soft-DTW	0.78	24 (12%)
Previous studies:
Empirical criterion	-	8%[2], 16%[3]

^a$\overline{s}$_{kis an average of}$\overline{s}$_k(i_α) over all the number of Traj_T-R.

^b The percentage in parentheses are the ratio of Traj_T-R in all the 200 MD trajectories.

References

• [1]   J. F. Storz, (2019). "HEMOGLOBIN: Insights into Protein Structure, Function, and Evolution", Oxford, Oxford University Press. ISBN; 978–0-19–881068–1.
• [2]   M. Tanakayagi, I. Kurisaki, M. Nagaoka, Sci. Rep., 4, 4601 (2014). doi:10.1038/srep04601 PMID:24710521
• [3]   K. Suzuki, (2019) "Theoretical Research of Allosteric T R State Transition of Human Hemoglobin: Application of Principal Component Analysis and Motion Tree Method", Master thesis, Nagoya University, Japan.
• [4]   J. Paparrizos, L. Gravano, ACM Trans. Database Syst., 42, 1 (2017). doi:10.1145/3044711
• [5]   H. Sakoe, S. Chiba, IEEE Trans. Acoust. Speech Signal Process., 26, 43 (1978). doi:10.1109/TASSP.1978.1163055
• [6]   E. Keogh, C. A. Ratanamahatana, Knowl. Inf. Syst., 7, 358 (2005). doi:10.1007/s10115-004-0154-9
• [7]   M. Cuturi, M. Blondel, "Soft-DTW: a Differentiable Loss Function for Time-Series," ICML 2017.
• [8]   P. J. Rousseeuw, J. Comput. Appl. Math., 20, 53 (1987). doi:10.1016/0377-0427(87)90125-7
• [9] s ¯ k was called the overall average silhouette width in [8].
• [10]   D. A. Case, R. M. Betz, D. S. Cerutti, T. E. Cheatham, III, T. A. Darden, R. E. Duke, T. J. Giese, H. Gohlke, A. W. Goetz, N. Homeyer, S. Izadi, P. Janowski, J. Kaus, A. Kovalenko, T. S. Lee, S. LeGrand, P. Li, C. Lin, T. Luchko, R. Luo, B. Madej, D. Mermelstein, K. M. Merz, G. Monard, H. Nguyen, H. T. Nguyen, I. Omelyan, A. Onufriev, D. R. Roe, A. Roitberg, C. Sagui, C. L. Simmerling, W. M. Botello-Smith, J. Swails, R. C. Walker, J. Wang, R. M. Wolf, X. Wu, L. Xiao, P. A. Kollman, (2016), AMBER 2016, University of California, San Francisco.
• [11]   S. Y. Park, T. Yokoyama, N. Shibayama, Y. Shiro, J. R. H. Tame, J. Mol. Biol., 360, 690 (2006). doi:10.1016/j.jmb.2006.05.036 PMID:16765986
• [12]   R. Tavenard, J. Faouzi, G. Vandewiele, F. Divo, G. Androz, C. Holtz, M. Payne, R. Yurchak, M. Rußwurm, K. Kolar, E. Woods, J. Mach. Learn. Res., 21, 1 (2020).