A Bioinformatic Approach To Identifying Sensitizing Therapeutics For Rheumatoid Arthritis

Background: Rheumatoid Arthritis (RA) is a systemic and chronic autoimmune
disorder that has significant impact on quality of life and work capacity for the
patient. Treatment and management of RA is aimed at gaining control of inflammation and alleviating pain, however, achieving remission and low disease activity with minimal toxicity is frequently not possible with the current suite of drugs.
Additionally with escalating de novo drug development costs, alternate bioinformatic approaches which scope the potential to repurpose already licenced
compounds and their ability to work synergistically have become attractive.
OBJECTIVES:
1)Develop, test and refine a bespoke connectivity mapping based bioinformatic
pipeline, DrugExpress, with capability of mining treatment response gene
signatures using public RNAseq and microarray datasets.
2)Verify treatment response signatures and prioritise novel drugs to generate a
list of robust sensitising drugs.
Methods: Public RNAseq (GSE198529) and microarray datasets (GSE93777 and
GSE24742) were mined based on the presence of DAS score, clinical features,
sample size and technological platform. R/Bioconductor packages and Perseus
software were used to pre-process and carry out differential expression (DE) analysis to obtain a list of differentially expressed genes (DEGs), based on the cut off
criteria of padj < 0.05 and a minimal 2-fold change of expression, and derive gene
signatures characteristic of treatment response. DEGs from multiple datasets were
combined to create a master gene list consisting of 21 DEGs. Gene Ontology (GO)
enrichment analysis was performed on the genes in the master list to identify the
top biological pathways. A list of inferred genes involved in each of these pathways
was compiled to determine the overall direction of expression with response to
treatment. DEGs in the master list were extracted and mapped to Affymetrix probeset IDs to create treatment response gene signatures to be entered into a sscMap
search for candidate drugs. Connectivity mapping (CMap) analysis was used to
establish networks between DEGs in the gene signature and FDA approved drugs.
P-value and connection score of each reference drug in the CMap was obtained
and used to determine statistical significance and perturbation stability.
Results: GO enrichment analysis performed on the master gene list identified
the top 3 pathways: transport of connexons to the plasma membrane, oligomerization of connexins into connexons, and gap junction assembly. CMap analysis
identified a total of 36 statistically significant compounds with a perturbation stability score of 1. Candidate compounds were selected based on whether they
enhanced a theoretical response phenotype. The top 3 ranked compounds (by
p value) which would induce theoretical reduction in disease activity were: ciclosporin (p-value = 0.0001), demecarium bromide (p-value = 0.0002) and 2-aminobenzenesulfonamide (p-value = 0.0002).
Conclusion: The analysis using the DrugExpress pipeline illustrates the process of treatment response DEG extraction from public expression datasets. These response signatures can be used for CMap analysis to predict new compounds for treatment refractive patients and can potentially simulate the effective
changes observed in responders. Next step in the pipeline would be to implement in silico toxicity screening methods to identify any contraindications with existing treatments.