Integrative and sparse singular value decomposition method for biclustering analysis in multi-sources dataset

• Published in: Chemometrics and intelligent laboratory systems Vol. 211; p. 104281
• Main Authors: Xu, Qing-Song, Li, Chuan-Quan, Wang, Xiaoyan, Li, Hongdong
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.04.2021
• Subjects: Biclustering analysis; Integrative analysis; Multi-sources dataset; Sparse singular value decomposition; Integrative analysis; Sparse singular value decomposition; Multi-sources dataset; Biclustering analysis
• ISSN: 0169-7439; 1873-3239
• DOI: 10.1016/j.chemolab.2021.104281

Multi-sources dataset analysis has been researched for a few years and it aims at improving the potential performance for discoveries by combining data from different sources. Meanwhile, biclustering analysis, widely applied for biology, chemistry, and so on, is a powerful tool that allows the clustering of rows and columns simultaneously. This paper proposes an integrative and sparse singular value decomposition (ISSVD) method for biclustering analysis in multi-sources datasets, which seeks the common left and individual right singular vectors, and finds the interpretable row-column associations. Simulation studies demonstrate that our model can handle ​with the different scales between the multi-source data, and obtain better performance than other popular biclustering methods. Two real-world data examples are also presented. •Our method is proposed for biclustering analysis in multi-sources dataset.•The alternative iterative algorithm is used to estimate the singular vectors.•Our method is insensitive to the scales between the data blocks, and can finds the correct row–column associations.•Two real datasets also show that our algorithm is able to obtain better performance than other biclustering analysis methods.