Gene selection for single cell RNA-seq data via fuzzy rough iterative computation model

Single cell RNA-seq data have the characteristics of small samples, high dimension and noise. Due to these characteristics, gene selection must be carried out before clustering and classifying. This study explores gene selection in a single cell gene decision space (scgd-space) via fuzzy rough iterative computation model (FRIC-model). First, in order to overcome the strictness of the equality between gene expression values, the equality between gene expression values is replaced by the distance between gene expression values, and the fuzzy symmetric relation on the cell set of a scgd-space is defined. In this fuzzy symmetric relation, two variable parameters are introduced: one controls gene subsets, the other dominates the distance between gene expression values. Then, FRIC-model in a scgd-space is established, which overcomes the deficiencies of classical rough set model and fuzzy rough set model. This model applies the iterative computation strategy to define some evaluation functions. These functions include fuzzy rough approximations and dependency functions. Next, a gene selection algorithm based on FRIC-model is designed. At last, the designed algorithm is testified in several publicly open single cell RNA-seq datasets to estimate its performance. The experimental results show that the designed algorithm is more effective than some existing algorithms, is fast and does not occupy too much memory.