Cluster similarity spectrum integration of single-cell genomics data
Technologies to sequence the transcriptome, genome or epigenome from thousands of single cells in an experiment provide extraordinary resolution into the molecular states present within a complex biological system at any given moment. However, it is a major challenge to integrate single-cell sequencing data across experiments, conditions, batches, timepoints and other technical considerations. New computational methods are required that can integrate samples while simultaneously preserving biological information. Here, we propose an unsupervised reference-free data representation, Cluster Similarity Spectrum (CSS), where each cell is represented by its similarities to clusters independently identified across samples. We show that CSS can be used to assess cellular heterogeneity and enable reconstruction of differentiation trajectories from cerebral organoid single-cell transcriptomic data, and to integrate data across experimental conditions and human individuals. We compare CSS to other integration algorithms and show that it can outperform other methods in certain integration scenarios. We also show that CSS allows projection of single-cell genomic data of different modalities to the CSS-represented reference atlas for visualization and cell type identity prediction. In summary, CSS provides a straightforward and powerful approach to understand and integrate challenging single-cell multi-omic data.