Downloads

To download the original, full 10M cell cytokine perturbation dataset, see the Parse Biosciences website:

The Parse Biosciences dataset and downloads derived by the Allen Institute are licensed under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. Non-commercial users are free to share and adapt the data by providing appropriate credit. See the Citation & Contributors page for citation information. Contact Parse Biosciences (support@parsebiosciences.com), for commercial use terms related to the Parse Biosciences dataset.

Differentially expressed genes

We utilized DESeq2 to compare pseudobulk scRNA-seq population data between cytokine treatments and the PBS control condition for the 12 subjects in this study. These comparisons were performed with all available genes, as well as with a subset of available genes after removing mitochondrial, ribosomal, and pseudogenes (see note in the descriptors below). The results of these pairwise comparisons are provided below, and descriptions of the columns are available by clicking on the header below:

Gene set enrichment analysis results

After performing DEG analysis, we ranked DEGs for each cytokine by nominal p-value and direction of change relative to the PBS control, then performed Gene Set Enrichment Analysis (GSEA) using the fgsea package for R against gene sets from MSigDB (Hallmark pathways) and Reactome. These results are used to display Pathway enrichment in our DEG Explorer visualization tool.

Gene sets for cytokine response enrichment analysis

After identifying differentially expressed genes, we ranked DEGs for each cytokine based on the direction of differential expression relative to PBS (up- or down-regulated), and selected the top 500 genes in each direction for each treatment condition. We provide these gene sets for use in Gene Set Enrichment Analysis tools, such as fgsea, to assist in identifying cytokine signatures in differential expression results.

CellTypist models for Parse PBMCs

To apply cell type labels that matched the Human Immune Health Atlas, we carefully labeled the PBS control cells in the Parse GigaLab dataset to match the cell types available in the atlas. We then used the labeled data to build CellTypist labeling models for AIFI_L1 (low resolution cell classes) and AIFI_L2 (intermediate resolution cell types), and applied these models to the cytokine perturbation datasets.

To enable other researchers to label their own Parse Biosciences PBMC scRNA-seq data, we provide the model files below, suitable for annotation using the celltypist package in Python.