Immune Single-Cell Annotation by Molecular Signature-Based Deconvolution with MIXTURE

Abstract

Accurate cell type identification in single-cell RNA sequencing (scRNA-seq) datasets is essential for dissecting the tumor immune microenvironment (TIME) and optimizing immunotherapy strategies. Here, we present an annotation protocol based on MIXTURE, our v-SVR deconvolution algorithm, which addresses the limitations of unsupervised annotation methods. We applied MIXTURE to an annotated melanoma scRNA-seq dataset, demonstrating its effectiveness in enhancing detailed cell type annotation and composition analysis of unannotated clusters. Our protocol successfully explores the composition of clusters, by providing detailed insights into cluster heterogeneity, enabling more accurate and granular identification of distinct cell types within complex single-cell RNA sequencing datasets.