Heejoon Chae

10.5626/JOK.2025.52.2.125

Single-cell RNA sequencing Unsupervised domain adaptation semi-supervised learning Cell type classification

Single-cell RNA sequencing (scRNA-seq) techniques for measuring gene expression in individual cells have developed rapidly. Recently, deep learning has been employed to identify cell types in scRNA-seq analysis. Most methods utilize a dataset containing cell-type labels to train the model and then apply this model to other datasets. However, integrating multiple datasets can result in unexpected batch effects caused by variations in laboratories, experimenters, and sequencing techniques. Since batch effect can obscure the biological signals of interest, an effective batch correction method is essential. In this paper, we present a cell-type prediction model for scRNA-seq that utilizes unsupervised domain adaptation and semi-supervised learning to minimize distributional differences between datasets. First, we pre-train the proposed model using a source dataset that contains cell-type information. Subsequently, we train the model on the target dataset by leveraging adversarial training to align its distribution of the target dataset with that of the source dataset. Finally, we re-train the model to enhance performance through semi-supervised learning, utilizing both the source and target datasets with consistency regularization. The proposed model outperformed the other deep learning-based batch correction models by effectively removing batch effects.