Supplementary Figure 7: Bulk RNA-seq analysis for flow sorted subpopulations of synovial fibroblasts, monocytes, and B cells to validate identified scRNA-seq clusters.

For each cell type, we trained a linear discriminant analysis (LDA) model on the scRNA-seq clusters. Next, we applied this LDA model to classify each bulk RNA-seq sample (Supplementary Fig. 6). After discovering scRNA-seq cluster markers (top 500 genes sorted by AUC for each cluster), we wanted to test if we could sort new cells from independent samples and see the same gene expression profiles in the new bulk samples as the original scRNA-seq samples. a, LDA projection of training data on single-cell fibroblasts (SC-F1-4). b, LDA projection of bulk RNA-seq samples that include sorted THY1^-DR^- populations from 4 OA, THY1⁺DR^- population from 4 OA and 6 RA, and THY1⁺DR⁺ population from 6 RA samples. c, Posterior probabilities showing confidence of assigning each sorted fibroblast bulk sample to fibroblast scRNA-seq clusters. d, Genes (top 10 by Z-score) that are differentially expressed between two scRNA-seq clusters (SC-F2 and SC-F4) are also differentially expressed in the sorted bulk RNA-seq. e, LDA projection of training data on single-cell monocytes (SC-M1-4). f, LDA projection of bulk RNA-seq samples that include sorted CD14⁺CD11c⁺⁺⁺CD38⁺⁺⁺ population from 2 RA and CD14⁺CD11c⁺CD38^– population from 2 OA. g, Posterior probabilities showing confidence of assigning each sorted monocyte bulk sample to monocyte scRNA-seq clusters. h. Genes (top 10 by Z-score) that are differentially expressed between two scRNA-seq clusters (SC-M1 and SC-M2) are also differentially expressed in the sorted bulk RNA-seq. i, LDA projection of training data on single-cell B cells (SC-B1-4). j, LDA projection of bulk RNA-seq samples that include sorted CD11c^–IgD^–CD27⁺ population from 6 RA, CD11c^–IgD⁺CD27^– population from 3 RA, CD19⁺CD11c⁺ population from 3 RA, and plasma cells from 3 RA. k, Posterior probabilities showing confidence of assigning each sorted B cell bulk sample to B cell scRNA-seq clusters. l, Genes (top ten by z score) that are differentially expressed between two scRNA-seq clusters (SC-B3 and SC-B4) are also differentially expressed in the sorted bulk RNA-seq.