Abstract: BACKGROUND: Fibromyalgia syndrome, as a common rheumatic disease, is related to central sensitization and immune abnormalities. However, the specific mechanism has not been elucidated, and there is a lack of specific diagnostic markers. Exploring the possible pathogenesis of this disease has important clinical significance. 
OBJECTIVE: To screen the potential diagnostic marker genes of fibromyalgia syndrome and analyze the possible immune infiltration characteristics based on bioinformatics methods, such as weighted gene co-expression network analysis (WGCNA), and machine learning. 
METHODS: Gene expression profiles in peripheral serum of fibromyalgia syndrome patients and healthy controls were obtained from the gene expression omnibus (GEO) database. The differentially co-expressed genes were screened in the expression profile by differential analysis and WGCNA analysis. Least absolute shrinkage and selection operator (LASSO) and support vector machine -recursive feature elimination (SVM-RFE) machine learning algorithm were further used to identify hub biomarkers, and draw receiver operating characteristic curve (ROC) to evaluate the accuracy of diagnosing fibromyalgia syndrome. Finally, single sample gene set enrichment analysis (ssGSEA) and gene set enrichment analysis (GSEA) were used to evaluate the immune cell infiltration and pathway enrichment in patients with fibromyalgia syndrome. 
RESULTS AND CONCLUSION: Eight down-regulated differentially expressed genes (DEGs) were obtained after differential analysis of the GSE67311 dataset according to the conditions of log2|(FC)| > 0 and P < 0.05. After WGCNA analysis, 497 genes were included in the module (MEdarkviolet) with the highest positive correlation (r=0.22, P=0.04), and 19 genes were included in the module (MEsalmon2) with the highest negative correlation (r=-0.41, P=6×10-5). After intersecting DEGs and the module genes of WGCNA, seven genes were obtained. Four genes were screened out by LASSO regression algorithm and five genes were screened out by SVM-RFE machine learning algorithm. After the intersection of the two, three core genes were identified, which were germinal center associated signaling and motility like, integrin beta-8, and carboxypeptidase A3. The areas under the ROC curve of the three core genes were 0.744, 0.739, and 0.734, respectively, indicating that they have good diagnostic value and can be used as biomarkers for fibromyalgia syndrome. The results of immune infiltration analysis showed that memory B cells, CD56 bright NK cells, and mast cells were significantly down-regulated in patients with fibromyalgia syndrome compared with the control group (P < 0.05), and were significantly positively correlated with the above three biomarkers (P < 0.05). The enrichment analysis suggested that there were nine fibromyalgia syndrome enrichment pathways, mainly related to olfactory transduction pathway, neuroactive ligand-receptor interaction, and infection pathway. The above results showed that the occurrence and development of fibromyalgia syndrome are related to the involvement of multiple genes, abnormal immune regulation, and multiple pathways imbalance. However, the interactions between these genes and immune cells, as well as their relationships with various pathways need to be further investigated.

Key words:  fibromyalgia syndrome, bioinformatics, machine learning, immune infiltration, weighted gene co-expression network analysis, Least absolute shrinkage and selection operator, support vector machine -recursive feature elimination, single sample gene set enrichment analysis, gene set enrichment analysis