Landscape of molecular crosstalk between fibromyalgia and COVID-19: New insights into diagnostic biomarkers based on six machine learning algorithms

Purpose To explore the crosstalk in molecular perturbation between fibromyalgia and COVID-19, and identify diagnostic biomarkers based on machine learning algorithms. Methods We obtained differentially expressed gene data for fibromyalgia (FM) and COVID-19 from the Gene Expression Omnibus database. The underlying mechanisms and potential pathways associated with the occurrence of fibromyalgia and COVID-19 were explored. Subsequently, six machine learning algorithms (LASSO, RandomForest, Xgboost, Gradient Boosting, Decision tree, and SVM ) were employed to conduct feature selection and a diagnostic nomogram model was constructed. Furthermore, molecular clusters were generated through consensus clustering analysis. The clinical features, and immune infiltration were compared further between clusters. Results A total of 30 COVID-FM differentially expressed genes were identified, contributing to fibromyalgia progression by participating in immune and related pathways. A downregulation in the fraction of CD56 bright natural killer cells and Memory B cells in FM and COVID samples compared with control. Then 6 genes (CPA3, FCER1A, EPSTI1, ZNF737, JCHAIN, and KIAA1324) were screened out by overlapping features from six machine learning algorithms. The diagnostic potential of two nomograms based on 6 biomarkers was confirmed in fibromyalgia (AUC: 0.813) and COVID (AUC: 0.944) datasets, and evaluated by decision curve, calibration curve and clinical impact curve analysis. Two molecular cluster were identified based on 6 biomarkers in fibromyalgia and COVID datasets. Similar analysis results between two clusters in FM and COVID dataset were obtained: CPA3 and FCER1A were significantly upregulated in cluster B, while ICHAIN was enriched in cluster A; Memory B cells, and CD56dim natural killer cells also exhibited upregulation in cluster B, whereas activated CD4 T cells showed activation in cluster A; OXIDATIVE_PHOSPHORYLATION, and PARKINSONS_DISEASE were enriched in both FM cluster B and COVID cluster B. Conclusions This study revealed molecular crosstalk between fibromyalgia and COVID-19 and identified diagnostic biomarkers utilizing machine learning algorithms. Further research is recommended to elucidate the underlying molecular drivers between two diseases.