scMAE is a novel masked autoencoder designed for single-cell RNA sequencing (scRNA-seq) data clustering. It improves clustering performance by introducing gene expression perturbations and using a masked autoencoder to reconstruct the original data, learning robust and informative cell representations. The masked autoencoder includes a masking predictor that identifies masked gene expression values, enabling the model to capture gene correlations and latent structures. scMAE outperforms existing state-of-the-art methods on 15 scRNA-seq datasets, accurately identifying rare cell types and demonstrating superior clustering performance. It achieves high accuracy in clustering and effectively captures biological significance of cell subpopulations. The method is evaluated using metrics such as Adjusted Rand Index (ARI), Normalized Mutual Information (NMI), and average silhouette width (ASW). scMAE's performance is validated through extensive experiments on various datasets, including those with batch effects. The method is implemented in Python and available at https://zenodo.org/records/10465991. scMAE's ability to accurately identify rare cell types and its robustness in clustering diverse scRNA-seq data highlight its effectiveness in single-cell analysis. The method's performance is further supported by biological analysis, which confirms the significance of identified cell subpopulations. scMAE's integration of masked autoencoder and masking predictor enables it to capture gene correlations and improve clustering accuracy. The method's efficiency and scalability make it a promising tool for scRNA-seq data analysis.scMAE is a novel masked autoencoder designed for single-cell RNA sequencing (scRNA-seq) data clustering. It improves clustering performance by introducing gene expression perturbations and using a masked autoencoder to reconstruct the original data, learning robust and informative cell representations. The masked autoencoder includes a masking predictor that identifies masked gene expression values, enabling the model to capture gene correlations and latent structures. scMAE outperforms existing state-of-the-art methods on 15 scRNA-seq datasets, accurately identifying rare cell types and demonstrating superior clustering performance. It achieves high accuracy in clustering and effectively captures biological significance of cell subpopulations. The method is evaluated using metrics such as Adjusted Rand Index (ARI), Normalized Mutual Information (NMI), and average silhouette width (ASW). scMAE's performance is validated through extensive experiments on various datasets, including those with batch effects. The method is implemented in Python and available at https://zenodo.org/records/10465991. scMAE's ability to accurately identify rare cell types and its robustness in clustering diverse scRNA-seq data highlight its effectiveness in single-cell analysis. The method's performance is further supported by biological analysis, which confirms the significance of identified cell subpopulations. scMAE's integration of masked autoencoder and masking predictor enables it to capture gene correlations and improve clustering accuracy. The method's efficiency and scalability make it a promising tool for scRNA-seq data analysis.