*2.1. Initial Evaluation of the Synovial and Cartilage RNA-seq Data Sets*

By using the keywords "osteoarthritis" in the https://www.ncbi.nlm.nih.gov/gds with the selection of "*Homo sapiens*" under the column of "Top Organisms" and "Expression profiling by high throughput sequencing" under the column of "study type", 499 items were identified which containing 4 datasets, 72 series, and 423 samples. After reviewing all the items, one dataset (GSE114007) containing transcriptome data of human knee cartilage from 18 healthy (5 females, 13 males) and 20 OA (11 females, 9 males) samples, and one dataset (GSE89408) containing transcriptome data of human synovium from 28 healthy (14 females, 14 males) and 22 OA (13 females, 9 males) samples that cover the U.S. population were included in the current study. After removing the samples with an overall alignment rate ≤75%, there were 11 human healthy cartilage samples (4 females, 9 males), 14 human OA cartilage samples (8 females, 6 males), 22 human healthy synovium samples (10 females, 12 males) and 20 human OA synovium (12 females, 8 males) samples underwent further analysis (Table S1).

By distinguishing the gene expression pattern between cartilage and synovium, multidimensional scaling (MDS) and principal component analysis (PCA) on the transcriptome profiling validated the quality of the included samples with minimum tissue contamination as expected (Figure 1).

**Figure 1.** Multidimensional scaling (MDS) (**a**) and principal component analysis (PCA) (**b**) distinguish transcriptome of synovium and cartilage samples included in the current study.
