A model incorporating gene expression and clinical factors predicts response to anti-tumor necrosis factor (TNF) therapy in patients with rheumatoid arthritis (RA), according to study results published in Network and Systems Medicine.

Rheumatoid arthritis is a complex disease and the molecular factors that predict response to treatment are poorly understood. Using gene expression data from patients receiving anti-TNF therapy, investigators aimed to build a biomarker panel, which predicts response to anti-TNF therapies in patients with RA.

Using publicly available data, a comprehensive map of the human protein-protein interactome was generated and used to identify an RA disease module, which contained approximately 200 proteins. Within this module, 66% of the proteins had been previously linked to RA in genome-wide association studies, and the remaining were significantly enriched in similar Gene Ontology biological processes.

Microarray data were obtained from the Gene Expression Omnibus database for 58 women receiving anti-TNF therapy. Using a random forest machine-learning algorithm, the data were used to identify genes for which expression was predictive of response and nonresponse to therapy. Based on the microarray profile and the RA disease module, 37 genes were identified as discriminatory biomarkers for anti-TNF response.


Continue Reading

In addition, RNA sequencing (RNAseq) data were obtained for 143 patients with RA from the Comparative Effectiveness Registry to Study Therapies for Arthritis and Inflammatory Conditions (CERTAIN) study. In addition to gene expression, RNAseq data enabled the identification of single-nucleotide variations (SNVs; formerly single-nucleotide polymorphisms) that were associated with treatment response. An additional 22 SNVs, which were associated with the RA disease model, were identified that were linked to response to anti-TNF therapy.

With the inclusion of clinical factors, a total of 70 biomarkers were identified and used to train a machine-learning algorithm to predict response to anti-TNF therapy. The final model generated, which predicted nonresponse to anti-TNF therapy, consisted of 10 SNPs, 8 gene transcripts, 2 laboratory tests, and 3 clinical measures.

To confirm that the model was broadly generalizable, data from an independent group of 175 patients from the CERTAIN study were used for a validation trial. Patients who were identified by the model as being nonresponders to anti-TNF therapy were 6.57-times more likely to be a true nonresponder than a responder (95% CI, 2.75-15.70). The model predicted nonresponse to therapy with a positive predictive value of 89.7% (95% CI, 79.0-95.7%), a specificity of 86.8% (95% CI, 72.4-94.1%), and a sensitivity of 50.0% (95% CI, 40.8-58.7%). There was no significant difference observed in the predictive power of the model based on ethnicity.

When the relevant gene transcripts and SNVs were mapped to the human interactome, they were all within close proximity to the RA disease module as well as established RA drug targets, including Janus kinase and TNF-α. Pathway enrichment identified genes involved in T-cell signaling as the most enriched pathway in the biomarkers associated with a therapeutic response.

“Customization of treatment regimens to match the individualized disease biology of each patient is a goal of modern medicine,” the researchers concluded. “Development and validation of a drug response algorithm that predicts nonresponse to a targeted therapy using this machine-learning and network medicine approach show great promise for advancing precision medicine in the treatment of RA and other complex autoimmune diseases where costly therapeutic interventions are met with inadequate patient response.”

Disclosure: This study was supported by Scipher Medicine Corporation. Please see the original reference for a full list of authors’ disclosures.

Reference

Mellors T, Withers JB, Ameli A, et al. Clinical validation of a blood-based predictive test for stratification of response to tumor necrosis factor inhibitor therapies in rheumatoid arthritis patients. Network and Systems Medicine. 2020;3(1):91-104.