High-performance cartesian genetic programming on GPU for the inference of gene regulatory networks using scRNA-seq time-series data.

The inference of Gene Regulatory Networks (GRNs) is an important topic with biotechnological and health applications, as comprehending patterns of gene interactions can lead to findings regarding living organisms. Evolutionary computation techniques, such as genetic programming, have been successfully applied to infer GRNs. However, evaluating candidate GRNs during the evolutionary search is computationally expensive and the adoption of such a type of method becomes an impediment when a large number of genes are involved. In this paper, we propose a strategy to infer GRN models from gene expression data using Cartesian Genetic Programming (CGP) and high-performance computing for reducing its processing time. In this regard, Graphics Processing Units (GPUs) are used through OpenCL to parallelize and accelerate the evaluation of the models. The results obtained in the computational experiments show that the high-performance CGP on GPU is suitable for inferring GRNs due to its significant reduction in the processing time when compared to the standard sequential CGP. Also, the proposal obtained better or competitive results when compared to state-of-art algorithms. Furthermore, the proposal outputs an interpretable solution that can help the domain experts in the field of Systems Biology.