Machine Learning for the Identification of a Common Signature for Anti-SSA/Ro 60 Antibody Expression Across Autoimmune Diseases

Machine learning and artificial intelligence within pediatric autoimmune diseases: applications, challenges, future perspective

INTRODUCTION

Autoimmune disorders affect 4.5% to 9.4% of children, significantly reducing their quality of life. The diagnosis and prognosis of autoimmune diseases are uncertain because of the variety of onset and development. Machine learning can identify clinically relevant patterns from vast amounts of data. Hence, its introduction has been beneficial in the diagnosis and management of patients.

AREAS COVERED

This narrative review was conducted through searching various electronic databases, including PubMed, Scopus, and Web of Science. This study thoroughly explores the current knowledge and identifies the remaining gaps in the applications of machine learning specifically in the context of pediatric autoimmune and related diseases.

EXPERT OPINION

Machine learning algorithms have the potential to completely change how pediatric autoimmune disorders are identified, treated, and managed. Machine learning can assist physicians in making more precise and fast judgments, identifying new biomarkers and therapeutic targets, and personalizing treatment strategies for each patient by utilizing massive datasets and powerful analytics.

Myositis-Associated Autoantibodies in Patients With Juvenile Myositis Are Associated With Refractory Disease and Mortality

OBJECTIVE

Myositis-associated autoantibodies (MAAs) have been associated with overlap myositis, certain disease manifestations such as interstitial lung disease (ILD), and worse prognosis in the idiopathic inflammatory myopathies. MAAs overall remain largely uncharacterized in patients with juvenile-onset myositis. Moreover, it is unknown whether the number of MAAs is associated with disease severity.

METHODS

Patients with juvenile myositis in cross-sectional natural history studies who underwent testing for myositis autoantibodies were included. Demographics, myositis autoantibodies, clinical characteristics, medications received, and outcomes of those with and without MAAs were compared. Multivariable logistic regression was performed to determine whether the number of MAAs detected was associated with severe disease features.

RESULTS

Among 551 patients, 36% had an MAA and 13% had more than one MAA. Among those who were MAA positive, there was a higher frequency of overlap myositis (18% vs 5.9%, P < 0.001). MAA positivity was associated with certain clinical features, including Raynaud phenomenon (odds ratio [OR] 2.44, 95% confidence interval [CI] 1.41-4.28) and ILD (OR 3.43, 95% CI 1.75-6.96), as well as a chronic disease course (OR 1.72, 95% CI 1.10-2.72) and mortality (OR 3.76, 95% CI 1.72-8.43). The number of MAAs was also associated with mortality (OR 1.83, 95% CI 1.16-2.86).

CONCLUSION

MAAs were prevalent in a large cohort of patients with juvenile myositis. ILD, refractory disease, and mortality were associated with MAA positivity. Prospective studies are needed to determine whether early detection of MAAs may lead to improved outcomes for patients with juvenile myositis.

Association of Combined Anti-Ro52/TRIM21 and Anti-Ro60/SSA Antibodies With Increased Sjögren Disease Severity Through Interferon Pathway Activation

OBJECTIVE

The biologic diagnosis of primary Sjögren disease (SjD) mainly relies on anti-Ro60/SSA antibodies, whereas the significance of anti-Ro52/TRIM21 antibodies currently remains unclear. The aim of this study was to characterize the clinical, serological, biologic, transcriptomic, and interferon profiles of patients with SjD according to their anti-Ro52/TRIM21 antibody status.

METHODS

Patients with SjD from the European PRECISESADS (n = 376) and the Brittany Diagnostic Suspicion of primitive Sjögren's Syndrome (DIApSS); (n = 146) cohorts were divided into four groups: double negative (Ro52-/Ro60-), isolated anti-Ro52/TRIM21 positive (Ro52+), isolated anti-Ro60/SSA positive (Ro60+), and double-positive (Ro52+/Ro60+) patients. Clinical information; EULAR Sjögren Syndrome Disease Activity Index, a score representing systemic activity; and biologic markers associated with disease severity were evaluated. Transcriptome data obtained from whole blood by RNA sequencing and type I and II interferon signatures were analyzed for PRECISESADS patients.

RESULTS

In the DIApSS cohort, Ro52+/Ro60+ patients showed significantly more parotidomegaly (33.3% vs 0%-11%) along with higher β2-microglobulin (P = 0.0002), total immunoglobulin (P < 0.0001), and erythrocyte sedimentation rate levels (P = 0.002) as well as rheumatoid factor (RF) positivity (66.2% vs 20.8%-25%) compared to other groups. The PRECISESADS cohort corroborated these observations, with increased arthritis (P = 0.046), inflammation (P = 0.005), hypergammaglobulinemia (P < 0.0001), positive RF (P < 0.0001), leukopenia (P = 0.004), and lymphopenia (P = 0.009) in Ro52+/Ro60+ patients. Cumulative EULAR Sjögren Syndrome Disease Activity Index results further confirmed these disparities (P = 0.002). Transcriptome analysis linked anti-Ro52/TRIM21 antibody positivity to interferon pathway activation as an underlying cause for these clinical correlations.

CONCLUSION

These results suggest that the combination of anti-Ro52/TRIM21 and anti-Ro60/SSA antibodies is associated with a clinical, biologic, and transcriptional profile linked to greater disease severity in SjD through the potentiation of the interferon pathway activation by anti-Ro52/TRIM21 antibodies.

Systemic lupus in the era of machine learning medicine

Artificial intelligence and machine learning applications are emerging as transformative technologies in medicine. With greater access to a diverse range of big datasets, researchers are turning to these powerful techniques for data analysis. Machine learning can reveal patterns and interactions between variables in large and complex datasets more accurately and efficiently than traditional statistical methods. Machine learning approaches open new possibilities for studying SLE, a multifactorial, highly heterogeneous and complex disease. Here, we discuss how machine learning methods are rapidly being integrated into the field of SLE research. Recent reports have focused on building prediction models and/or identifying novel biomarkers using both supervised and unsupervised techniques for understanding disease pathogenesis, early diagnosis and prognosis of disease. In this review, we will provide an overview of machine learning techniques to discuss current gaps, challenges and opportunities for SLE studies. External validation of most prediction models is still needed before clinical adoption. Utilisation of deep learning models, access to alternative sources of health data and increased awareness of the ethics, governance and regulations surrounding the use of artificial intelligence in medicine will help propel this exciting field forward.

TAFRO syndrome is associated with anti-SSA/Ro60 antibodies, in contrast to idiopathic castleman disease

TAFRO syndrome is an acute systemic inflammatory disease characterized by thrombocytopenia, anasarca, fever, reticulin fibrosis/renal dysfunction, and organomegaly. There have been increasing reports that TAFRO is a disease distinct from idiopathic multicentric Castleman disease and that TAFRO patients may be positive for anti-SSA antibodies. To assess anti-SSA antibody positivity and the clinical characteristics of the two diseases, we retrospectively compared 7 TAFRO and 10 iMCD patients in our hospital. The mean age of onset of TAFRO and iMCD was 48.0 (interquartile range [IQR], 41-53) and 45.0 (IQR, 35-53) years, respectively. The TAFRO and iMCD groups had 6 (86%) and 4 (40%) male patients, respectively, and the following pretreatment laboratory values: platelet count, 3.8 (IQR, 2.2-6.4) and 35.5 (IQR, 22.2-42.8) × 104/μL, respectively; C-reactive protein, 10.2 (IQR, 6.8-21.4) and 9.5 (IQR, 6.2-13.6) mg/dL, respectively; IgG, 1431 (IQR, 1112-1815) and 4725 (IQR, 3755-5121) mg/dL, respectively. RNA immunoprecipitation (5 cases for anti-SSA) or protein array (5 cases for anti-SSA/Ro60) detected anti-SSA antibodies in six (86%) TAFRO patients but not in iMCD patients; it did not detect anti-SSB antibodies in any of the patients. None of the patients were diagnosed with Sjögren syndrome. All iMCD patients treated with tocilizumab (TCZ) responded well. Meanwhile, two of six TAFRO patients treated with TCZ showed inadequate responses; thus, both patients were switched to rituximab, following which they achieved remission. TAFRO and iMCD have different clinical features. TAFRO may be categorized as a severe phenotype of the anti-SSA antibody syndrome.

CXCL9 may serve as a potential biomarker for primary Sjögren's syndrome with extra-glandular manifestations

BACKGROUND

Primary Sjögren's syndrome (pSS) is an autoimmune condition that causes harm to exocrine glands and also has extra-glandular manifestations (EGM). pSS patients with EGM have a worse prognosis than those with only sicca symptoms. Previous studies have shown that the minor salivary glands (MSG) of pSS patients exhibit a unique profile of cytokines and chemokines compared to healthy controls. However, there is a lack of research comparing pSS with EGM (pSS-EGM) and pSS without EGM (pSS-non-EGM). This study aims to explore potential biomarkers associated with pSS, particularly pSS with EGM.

METHODS

By utilizing RNA sequencing, we conducted an analysis on the gene expression profiles of MSG in 63 patients diagnosed with pSS, as well as 12 non-pSS individuals. Furthermore, we also investigated the MSG of pSS patients, both with and without EGM. Through bioinformatics analysis, we identified genes with differential expression (DEGs) and determined the core hub genes using PPI network. We then analyzed the top 20 DEGs and their correlation with the patients' clinical characteristics, and validated our findings using peripheral blood plasma.

RESULTS

A total of 725 differentially expressed genes (DEGs) were identified in the comparison between pSS and non-pSS groups, and 727 DEGs were observed between pSS-EGM and pSS-non-EGM. It is noteworthy that the expression levels of CXCL9 were higher in both pSS patients and pSS-EGM when compared to the control group. Taking into consideration the significance of the top 20 DEGs in relation to clinical parameters and the central hub genes, we ultimately chose CXCL9. In comparison to the non-pSS group, pSS patients exhibited notably greater expression of the CXCL9 gene in the MSG, as well as higher levels of CXCL9 protein in their plasma (p < 0.001). Furthermore, the expression of the CXCL9 gene and levels of CXCL9 protein were notably higher in pSS patients accompanied by EGM and those with SSA antibodies. Additionally, a correlation was found between the expression of the CXCL9 gene and the EULAR Sjogren's Syndrome Disease Activity Index (ESSDAI), as well as with immunoglobulin G (IgG) levels and erythrocyte sedimentation rate (ESR). Meanwhile, the protein levels of CXCL9 were found to be correlated with IgG levels and ESSDAI.

CONCLUSION

CXCL9 proves to be a valuable biomarker in pSS, specifically due to its strong ability to differentiate between pSS patients with EGM and those without EGM. There is a significant correlation between CXCL9 and various clinical parameters both at the gene and protein level. Therefore, CXCL9 could be a potential target for future treatment of pSS.

Pervasive Sharing of Causal Genetic Risk Factors Contributes to Clinical and Molecular Overlap between Sjögren's Disease and Systemic Lupus Erythematosus

SjD (Sjögren's Disease) and SLE (Systemic Lupus Erythematosus) are similar diseases. There is extensive overlap between the two in terms of both clinical features and pathobiologic mechanisms. Shared genetic risk is a potential explanation of this overlap. In this study, we evaluated whether these diseases share causal genetic risk factors. We compared the causal genetic risk for SLE and SjD using three complementary approaches. First, we examined the published GWAS results for these two diseases by analyzing the predicted causal gene protein-protein interaction networks of both diseases. Since this method does not account for overlapping risk intervals, we examined whether such intervals also overlap. Third, we used two-sample Mendelian randomization (two sample MR) using GWAS summary statistics to determine whether risk variants for SLE are causal for SjD and vice versa. We found that both the putative causal genes and the genomic risk intervals for SLE and SjD overlap 28- and 130-times more than expected by chance (p < 1.1 × 10-24 and p < 1.1 × 10-41, respectively). Further, two sample MR analysis confirmed that alone or in aggregate, SLE is likely causal for SjD and vice versa. [SjD variants predicting SLE: OR = 2.56; 95% CI (1.98-3.30); p < 1.4 × 10-13, inverse-variance weighted; SLE variants predicting SjD: OR = 1.36; 95% CI (1.26-1.47); p < 1.6 × 10-11, inverse-variance weighted]. Notably, some variants have disparate impact in terms of effect size across disease states. Overlapping causal genetic risk factors were found for both diseases using complementary approaches. These observations support the hypothesis that shared genetic factors drive the clinical and pathobiologic overlap between these diseases. Our study has implications for both differential diagnosis and future genetic studies of these two conditions.

TLR7 agonism accelerates disease in a mouse model of primary Sjögren's syndrome and drives expansion of T-bet+ B cells

Primary Sjögren's syndrome (pSS) is a systemic autoimmune disease characterized by chronic inflammation of exocrine tissue, resulting in loss of tears and saliva. Patients also experience many extra-glandular disease manifestations. Treatment for pSS is palliative, and there are currently no treatments available that target disease etiology. Previous studies in our lab demonstrated that MyD88 is crucial for pSS pathogenesis in the NOD.B10Sn-H2^b (NOD.B10) pSS mouse model, although the way in which MyD88-dependent pathways become activated in disease remains unknown. Based on its importance in other autoimmune diseases, we hypothesized that TLR7 activation accelerates pSS pathogenesis. We administered the TLR7 agonist Imiquimod (Imq) or sham treatment to pre-disease NOD.B10 females for 6 weeks. Parallel experiments were performed in age and sex-matched C57BL/10 controls. Imq-treated pSS animals exhibited cervical lymphadenopathy, splenomegaly, and expansion of TLR7-expressing B cells. Robust lymphocytic infiltration of exocrine tissues, kidney and lung was observed in pSS mice following treatment with Imq. TLR7 agonism also induced salivary hypofunction in pSS mice, which is a hallmark of disease. Anti-nuclear autoantibodies, including Ro (SSA) and La (SSB) were increased in pSS mice following Imq administration. Cervical lymph nodes from Imq-treated NOD.B10 animals demonstrated an increase in the percentage of activated/memory CD4+ T cells. Finally, T-bet⁺ B cells were expanded in the spleens of Imq-treated pSS mice. Thus, activation of TLR7 accelerates local and systemic disease and promotes expansion of T-bet-expressing B cells in pSS.

Aberrant B Cell Signaling in Autoimmune Diseases

Aberrant B cell signaling plays a critical in role in various systemic and organ-specific autoimmune diseases. This is supported by genetic evidence by many functional studies in B cells from patients or specific animal models and by the observed efficacy of small-molecule inhibitors. In this review, we first discuss key signal transduction pathways downstream of the B cell receptor (BCR) that ensure that autoreactive B cells are removed from the repertoire or functionally silenced. We provide an overview of aberrant BCR signaling that is associated with inappropriate B cell repertoire selection and activation or survival of peripheral B cell populations and plasma cells, finally leading to autoantibody formation. Next to BCR signaling, abnormalities in other signal transduction pathways have been implicated in autoimmune disease. These include reduced activity of several phosphates that are downstream of co-inhibitory receptors on B cells and increased levels of BAFF and APRIL, which support survival of B cells and plasma cells. Importantly, pathogenic synergy of the BCR and Toll-like receptors (TLR), which can be activated by endogenous ligands, such as self-nucleic acids, has been shown to enhance autoimmunity. Finally, we will briefly discuss therapeutic strategies for autoimmune disease based on interfering with signal transduction in B cells.