scParadise: Tunable highly accurate multi-task cell type annotation and surface protein abundance prediction

Abstract

scRNA-seq is revolutionizing biomedical research by revealing tissue architecture, cellular composition, and functional interactions. However, accurate cell type annotation remains a challenge, particularly for rare cell types, with existing automated methods often falling short. Multimodal data, combining mRNA expression and protein markers, improves deep cellular analysis and make functional characterization of complex tissues more accurate. However, it is costly and complex to obtain. We present scParadise, a cutting-edge Python framework featuring three tools: scAdam for multi-level cell annotation, scEve for surface protein prediction, and scNoah for benchmarking. scAdam surpasses current methods in annotating rare cell types and ensures consistent results across diverse datasets, while scEve enhances clustering and cell type separation. With scNoahs advanced metrics, scParadise offers a powerful, fast, and reliable solution for single-cell analysis, setting a new standard in scRNA-seq data processing.