ProjectSVR: Mapping single-cell RNA-seq data to reference atlases by supported vector regression

The rapid accumulation of public single-cell RNA-seq (scRNA-seq) data and advances in integration algorithms make it possible to build the organ and body-scale cell atlases after overcoming the batch effect originating from different studies, donors, and sequence platforms. However, projecting the query cells onto the carefully constructed reference atlas to rapidly interpret cell states remains challenging. Here, we present ProjectSVR (), a machine learning-based algorithm for mapping the query cells onto well-constructed reference embeddings. We demonstrate the effectiveness of ProjectSVR in multiple case studies of different tissue and experimental design, including (1) interpreting the decidual immune microenvironment from recurrent pregnancy loss (RPL) patients, (2) resolving the tumor-infiltrated T cell states from patients receiving anti-PD1 treatment, (3) mapping the genetic-perturbed mouse germ cells to integrated mouse testicular cell atlas (mTCA) for inferring their arrested developmental stage, and (4) evaluating in vitro induced meiosis by via mapping the induced germ cells onto mTCA. Taken together, our results show that ProjectSVR is a useful tool for reference mapping and make the scRNA-seq data analysis more convenient when well-annotated atlases emerge. ### Competing Interest Statement The authors have declared no competing interest.