GLYCOSYLATION OF ANTI-DSDNA IGG CORRELATES WITH ORGAN INVOLVEMENT IN TREATMENT-NAÏVE SYSTEMIC LUPUS ERYTHEMATOSUS PATIENTS

PV039 / #308 Poster Topic: AS04 - Biomarkers Background/Purpose Anti-double-stranded DNA (anti-dsDNA) antibodies are important antibodies in systemic lupus erythematosus (SLE). Glycosylation is one of the most commonly post-translational modifications of antibodies, and anti-dsDNA antibodies glycosylation is related with SLE disease activity. However, the association of anti-dsDNA antibodies glycosylation and SLE organ involvement is still unclear. Methods We enrolled 86 consecutive treatment-naïve SLE patients with positive anti-dsDNA antibodies from the Department of Rheumatology and Immunology at Ruijin Hospital, Shanghai, between 2017 and 2022. Serum samples were used in this study. We quantified and classified the organ involvement degree of SLE patients according to the number of organ systems involved in each patient. Then we analyzed each glycoform and combination of glycoforms based on the involvement degree. Random Forest Classifier and Artificial Neural Network were applied to evaluate the correlation between combinations of glycoforms and the organ involvement degree (Figure 1). Figure 1. Workflow of classifying and predicting the involvement degree of organ systems in SLE patients by leveraging glyco-pairs. Results Pearson correlation analysis presented a strong connection between involved organs compared with uninvolved organs in SLE patients. The bisection(Bis) of IgG3/4, galactosylation (Gal) of IgG1, fucosylation (Fuc) of IgG1, and sialylation (Sia) of IgG2 displayed high Area under Curve (AUC) values when combined with other glycoforms for classifying the involvement degree. The result of Random Forest showed that the combination of IgG1Gal&IgG3/4Bis had the highest accuracy (0.7692) and AUC value (0.8187). In terms of predicting the involvement rate using Artificial Neural Network, IgG3/4Bis&IgG1Gal had the lowest MSE (0.0244) (Figure 2). Figure 2. The results of the RF classification model on glyco-pairs. In a-f, the dots on the figure represented the original samples and the grid-like background colors represented the output of the model. Dots sharing the same color with the background color were correctly classified by the RF model. The accuracy might seem smaller than the intuition as it was derived solely from the test set’s samples. In g-l, ROC of the RF models were plotted. AUC, 95% CI and p-value all reflected the performance of a an RF model. A larger AUC value and a wider gap between the 95% CI and 0.5 suggested a better classification ability of the glyco-pair. In both evaluation metrics, glyco-pair IgG1Gal&IgG3/4Bis performed the best. Abbreviations: AUC: Area under Curve; ROC: Receiver Operating Curves; CI: confidence interval. Conclusions Our study showcased the effectiveness of combining glycotypes to classify and predict SLE organ involvement degree. Different glycotypes were correlated with the involvement degree to different extents, and the combination of IgG3/4Bis&IgG1Gal had best correlation with SLE organ involvement.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.