Gene selection in a single cell gene decision space based on class-consistent technology and fuzzy rough iterative computation model

• Published in: Applied intelligence (Dordrecht, Netherlands) Vol. 53; no. 24; pp. 30113 - 30132
• Main Authors: Zhang, Jie, Yu, Guangji, Huang, Dan, Wang, Yuxian
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.12.2023; Springer Nature B.V
• Subjects: Algorithms; Approximation; Artificial Intelligence; Clustering; Computation; Computer Science; Gene expression; Iterative methods; Machines; Manufacturing; Mechanical Engineering; Processes; Single cell gene decision space; Single cell RNA-seq data; Gene selection; Fuzzy rough set; FRIC-model; Class-consistent technology
• ISSN: 0924-669X; 1573-7497
• DOI: 10.1007/s10489-023-05115-0

This study explores gene selection in a single cell gene decision space ( scgd -space) based on class-consistent technology and fuzzy rough iterative computation model (FRIC-model). Gene expression data ( ge -data) exhibit characteristics such as limited sample size, high dimensionality, and noise. Due to their high dimensionality, gene selection must be carried out before clustering and classifying them. The existing gene selection methods based on equivalence relation are not effective for ge -data owing to the strictness of the equality between gene expression values. In order to overcome this weakness, class-consistent technology of replacing equality with approximate equality between gene expression values is first proposed. Then, “the class consistency between gene expression values is fed back to the gene set” is considered with the help of class-consistent technology, and fuzzy symmetric relations on the cell set of a scgd -space are induced. In addition, fuzzy rough approximations in a scgd -space are defined. Next, FRIC-model is given. This model employs the iterative computation strategy to define fuzzy rough approximations and dependency functions. A gene selection algorithm based on this model is designed. Finally, the designed algorithm is testified in several publicly open ge -data sets to estimate its performance. The experimental results show that the designed algorithm is more effective than some existing algorithms.