A predictive model based on the gut microbiota improves the diagnostic effect in patients with rheumatoid arthritis

Abstract Objectives Rheumatoid arthritis (RA) is a chronic, destructive autoimmune disorder predominantly targeting the joints, with gut microbiota dysbiosis being intricately associated with its progression. The aim of the present study is to develop effective early diagnostic methods for early RA based on gut microbiota. Methods A cohort comprising 262 RA patients and 475 healthy controls (HCs) was recruited. Faecal samples were collected from all participants, and microbial DNA was subsequently extracted. The V3-V4 region of the 16S rRNA gene was amplified via polymerase chain reaction (PCR) and subjected to high-throughput sequencing using the Illumina MiSeq platform. Additionally, a dataset with the accession number PRJNA450340 from the European Nucleotide Archive (ENA) was incorporated into the study. The sequencing data underwent processing and analysis utilizing QIIME2. To construct microbiome-based diagnostic models, Random Forest (RF), Support Vector Machine (SVM) and Generalized Linear Model (GLM) methodologies were employed, with the self-test data functioning as the training set and the PRJNA450340 dataset serving as the validation set. Result The results indicated that patients with RA exhibited a significantly reduced gut microbial α-diversity compared with the HCs group. The β-diversity analysis demonstrated notable distinctions in the gut microbiota structure between RA patients and HCs. Variations in the gut microbiome composition between RA patients and HCs were evident at both the phylum and genus levels. LEfSe analysis revealed a substantial number of significantly different microbiota between RA patients and HC, and seven key genera were obtained by intersection of the different flora in the two data sets: Ruminococcus_gnavus_group, Fusicatenibacter, Butyricicoccus, Subdoligranulum, Erysipelotrichaceae_UCG-003, Romboutsia and Dorea. Utilizing these seven core genera, RA diagnostic models were developed employing RF, SVM and GLM methodologies. The GLM model exhibited consistent performance, achieving an area under the curve (AUC) of 71.03% in the training set and 74.71% in the validation set. Conclusion Notable differences in gut microbiota exist between RA patients and healthy individuals. Diagnostic models based on key microbial genera hold potential for aiding in the early identification of individuals at risk for developing RA, thereby suggesting new avenues for its diagnosis.