Artificial intelligence in cell annotation for high-resolution RNA sequencing data

Characterization of cell heterogeneity in gene expression is a "hot" but complex issue in the fields of biology and medicine. Therefore, high-resolution RNA sequencing technologies, involving single-cell RNA sequencing and spatial transcriptomics technologies, have been constantly evolving and widely applied. As the cornerstone of high-resolution RNA sequencing data analysis, cell annotation encounters challenges, including the high sparsity and noise levels in datasets, as well as batch effects. Artificial intelligence technologies have developed rapidly in recent years and have penetrated the field of cell annotation algorithm development, demonstrating excellent performance. While several algorithm reviews in this field exist, they primarily focus on evaluating algorithmic performance rather than introducing computational solutions. To fill the gap, we reviewed the cutting-edge 27 single-cell and 22 spatial transcriptomics cell annotation methods and categorized them according to machine learning strategies. We anticipate this article will provide readers with a comprehensive picture of cell annotation strategies.