Fast, sensitive and accurate integration of single-cell data with Harmony

Abstract

The emerging diversity of single cell RNAseq datasets allows for the full transcriptional characterization of cell types across a wide variety of biological and clinical conditions. However, it is challenging to analyze them together, particularly when datasets are assayed with different technologies. Here, real biological differences are interspersed with technical differences. We present Harmony, an algorithm that projects cells into a shared embedding in which cells group by cell type rather than dataset-specific conditions. Harmony simultaneously accounts for multiple experimental and biological factors. In six analyses, we demonstrate the superior performance of Harmony to previously published algorithms. We show that Harmony requires dramatically fewer computational resources. It is the only currently available algorithm that makes the integration of ~10 6 cells feasible on a personal computer. We apply Harmony to PBMCs from datasets with large experimental differences, 5 studies of pancreatic islet cells, mouse embryogenesis datasets, and cross-modality spatial integration. Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:

Single-cell Analysis of ACE2 Expression in Human Kidneys and Bladders Reveals a Potential Route of 2019-nCoV Infection

“…The single cell suspension was proceeded to the single cell library preparation and sequencing, as previously described 15 . To be consistent with the public data used above, we applied a similar downstream analysis workflow 14 . We modified the number of principal components to 20 and used resolution 0.6 to obtain comparable cell type clustering results with the public kidney data.…”

“…By analyzing the public single-cell transcriptome dataset of normal human kidney cells from three donors 14 , we found ACE2 expression distributed across multiple cell types. Notably, ACE2 was mostly enriched in proximal tubule cells, including both convoluted tubule and straight tubule ( Fig 1A-D).…”

Benchmarking atlas-level data integration in single-cell genomics

“…Harmony 18 ], a bulk data integration tool (ComBat 19 ), and a perturbation modeling tool [transformer variational autoencoder (trVAE) 20 ]. Moreover, we use 14 metrics to evaluate the integration methods on their ability to remove batch effects while conserving biological variation.…”

Drosophila PDGF/VEGF signaling from muscles to hepatocyte-like cells protects against obesity

“…Because tissue dissociations for the single nuclei preparations are often associated with the risk of ambient RNA contamination, our quality control pipeline included SoupX (Young and Behjati, 2018) to eliminate potential ambient RNA from our analysis. Further, we used Harmony (Korsunsky et al, 2019) that is integrated into Seurat (Stuart et al, 2019) to correct for batch effects in the two replicates to finally retain 15,280 nuclei with a median of 192 genes per nucleus for downstream analysis ( Figure S2-B,D; Table S1). Our clustering analysis revealed 10 unique clusters, where three major clusters were assigned to adipose tissue, oenocytes, and muscle based on known markers for each of these tissue types including apolpp, fasn3, and mhc, respectively ( Figure 2B; S2E; Table S2).…”

“…We filtered the cells beyond UMI counts ± 2-fold Standard Deviation of the average total sample counts (log10) after SoupX, which were regarded as doublets or dead cells in droplet. The quality filtered datasets were combined into a single Seurat (version 3.1.2) object (Stuart et al, 2019) and integrated using Harmony (version 1.0) (Korsunsky et al, 2019) with default analysis workflow and parameters. A resolution of 0.1 was chosen as clustering parameter.…”