Primary Sjögren’s syndrome (pSS) is a chronic and systemic autoimmune disorder comprised of xerostomia and xerophthalmia – symptoms caused by a lymphocytic infiltration of salivary and lacrimal glands. pSS can also affect other areas of the body, including joints, lungs, kidneys, liver, nervous system, and musculoskeletal system. This condition is more common in women, with the female-to-male ratio approximately at 9:1. To diagnose pSS, clinicians rely on a combination of clinical signs and symptoms, such as serological tests for autoantibody biomarkers and salivary gland histopathology. Since the pathogenesis of pSS is poorly understood, the need of uncovering genuine biomarkers and constructing diagnostic models is of great importance.
In order to develop a predictive model for pSS, four gene profiling datasets were obtained from the Gene Expression Omnibus (GEO) database. Using the ‘limma’ software, differentially expressed genes (DEGs) associated with pSS were identified. Moreover, a randomized forest-supervised classification algorithm was utilized to screen disease-specific genes. After that, the ANN, RF, and SVM machine learning algorithms were used to build a diagnostic model for pSS. The performance of the model was tested using the area under the receiver operating characteristic (AUC) curve. To view the immune cell infiltration involved in pSS, CIBERSORT was employed. In summary, a total of 96 DEGs were identified, out of which 14 signature genes were pinpointed using RF classification for key transcription regulation and disease progression in pSS.
Subsequent to the development of training and testing datasets, ANN, RF, and SVM diagnostic models for pSS yielded AUCs of 0.972, 1.00, and 0.9742, respectively. For the validation set, the AUC rates stood at 0.766, 0.8321, and 0.8223, respectively. Among all these three, the RF model demonstrated the highest prediction performance. The success of developing an early predictive model for pSS with a high diagnostic ability perfects provides a valuable resource for the
