Principal component analysis verifying each dataset independence
To identify key genes involved in osteoarthritis pathogenesis, principal component analysis (PCA) of thecollected dataset was made to distinguish the significant difference between normal and osteoarthritis samples. Results from PCA showed that our collected sub-datasets (GSE43923, GSE113825, GSE129147 and GSE169077) displayed a significant difference after removing some samples in partly datasets(Fig. 1), indicating that these sub-datasets can be used to identify key genes involved in osteoarthritis pathogenesis.
Differentially Expressed Genes Of Osteoarthritis
Duo to platform difference and batch effect, the collected datasets were independently analyzed. Following the criteria: p-value < 0.05 and |log2FC| > 1, a number of differentially expressed genes (DEGs) were identified and sixty-six overlapped DEGs were demonstrated in four collected datasets thought Venn diagram (Fig. 2A). In addition,the two-way clustering heatmaps of DEGs were shown in Fig. 2B, in which there was fifty-six up-regulated and ten down-regulated DEGs in osteoarthritis compared with normal samples. Chromosome mapping showed chromosome distribution of the overlapped DEGs, with chromosomes 3 and 5 containing the greatest number of dysregulated genes in osteoarthritis (Fig. 2C). Interestingly, two genes (MXRA5 and TMSB4X) on the X chromosome showed the dysregulation in osteoarthritis .
Functional Enrichment Analysis And Protein-protein Interaction
GO annotation and KEGG pathways were performed to reveal the potential biological functions of DEGs. Results from GO terms showed that the biological processes (BP), cell component (CC) and molecular functions (MF) of 66 DEGs were primarily related with extracellular matrix and collagen (Fig. 3A, B and C). KEGG pathway analysis showed that DEGs were significantly enriched in focal adhesion, PI3K-Akt signaling pathway and ECM-receptor interaction pathway, in which autophagy were participated (Fig. 3D). Protein-protein interaction enrichment analysis with STRING showed that there were broader interaction in most of DEGs and the higher interaction nodes were associated with ECM-receptor interaction pathway (Fig. 4A). Moreover, three significant sub-network clusters were aggregated based on molecular Complex Detection (MCODE) algorithm with the Cytoscape software (Fig. 4B).
Construction Of A ceRNA Network For Osteoarthritis
To release whether these DEGs in canregulate microRNAs and long non-coding RNAs in osteoarthritis. osteoarthritis-related miRNAs were firstly collected with HMDD database and then compared with predicated differentially expressed miRNAs from literatures in osteoarthritis34. We identified 8 miRNAs (hsa-mir-132, hsa-mir-140, hsa-mir-146a, hsa-mir-146b, hsa-mir-155, hsa-mir-16, hsa-mir-181a and hsa-mir-98) involved in osteoarthritis. Moreover,miRNA-lncRNA and miRNA-mRNA interaction analysis based on ENCORI database prediction were integrated with osteoarthritis-related lncRNA from literature and the overlapped DEGs, respectively (Fig. 5). 8 miRNAs, 7 lncRNAs and 34 genes were constructed into ceRNA network and we found that XIST was enriched and may reuglate miRNA to control the expression of osteoarthritis-related genes in osteoarthritis.
Autophagy-related Genes Interacted With The Overlapped DEGs
Nine autophagy-related genes (CCL2, CDKN1A, CXCR4, DAPK1, DLC1, FAS, HSPA8, MYC and SERPINA1) were identified in our collected datasets of osteoarthritis (Fig. 6). In addition, results from protein-protein interaction showed that eight of these autophagy-related genes could interact with the overlapped DEGs that were associated with PI3K-Akt signaling pathway and ECM-receptor interaction pathway in osteoarthritis (Fig. 7).