Background/Purpose: There is an urgent need to develop objective biomarkers for early diagnosis and monitoring of disease activity in Rheumatoid arthritis (RA). Here we define a RA meta-profile using publicly available cross-tissue gene expression data and apply machine learning to identify putative biomarkers, which we further validate on independent datasets.

Methods: We carried out a comprehensive search for publicly available microarray gene expression data at NCBI Gene Expression Omnibus database for whole blood and synovial tissues in RA and healthy controls. The raw data from 13 synovium datasets with 284 samples and 14 blood datasets with 1,885 samples were downloaded and processed. The datasets were merged and batch corrected separately for each tissue. We further developed a robust feature selection pipeline to identify genes dysregulated in both tissues and highly associated with RA. Within training sets for each tissue, two sets of selected differentially expressed genes were identified and overlapped followed by the condition of co-directionality. Each gene in the resulting set was evaluated on the testing sets using AUROC. The process was repeated 100 times. 2 synovium and 3 blood independent datasets were used to validate the feature selected (FS) genes. We used the averaged AUC 0.8 threshold for the final filtering to define the RAScore, composed of a geometric mean of the selected genes.

Results: The result of the feature selection pipeline was a set of 25 upregulated and 28 downregulated genes. To confirm the robustness of the feature selected genes, we trained a Random Forest machine learning model with this set of 53 genes and the set of 32 common differentially expressed genes and tested on the validation cohorts. The model with FS genes outperformed the model with common DE genes with AUC 0.89 ± 0.04 vs 0.86 ± 0.05. The FS genes were further validated and thresholded on the 5 independent datasets resulting in 10 upregulated genes, TNFAIP6, S100A8, TNFSF10, DRAM1, LY96, QPCT, KYNU, ENTPD1, CLIC1, ATP6V0E1, that are involved in innate immune system pathways, including neutrophil degranulation and apoptosis and expressed in granulocytes, dendritic cells, and macrophages; and 3 downregulated genes, HSP90AB1, NCL, CIRBP, involved in metabolic processes and T-cell receptor regulation of apoptosis and expressed in lymphoblasts.

To investigate the clinical utility of the FS genes, RAScore was composed and found to be highly correlated with DAS28 (r = 0.33 ± 0.03, p = 7e-9) and able to distinguish RA and OA samples (t-test, p = 2.3e-6). However, it did not show any difference between RF-positive and RF-negative RA sub-phenotypes (t-test, p = 0.9) suggesting the generalizability of this score in clinical applications. The RAScore was also able to monitor the treatment effect among RA patients (t-test of treated vs untreated, p = 2e-4) and separate polyJIA from healthy individuals in 7 independent pediatric cohorts (t-test, p = 2e-4).

Conclusion: This novel list of biomarkers, identified through a robust feature selection procedure on public data and validated using multiple independent data sets, coupled with the RAScore may be useful in the early diagnosis and disease and treatment monitoring of RA.

« Back to ACR Convergence 2020

ACR Meeting Abstracts - https://acrabstracts.org/abstract/uncovering-novel-biomarkers-for-rheumatoid-arthritis-from-feature-selection-and-machine-learning-approaches-on-synovium-and-blood-gene-expression-data/