Fig. 1. Baseline lipid mediator profiles are predictive of DMARD responsiveness in RA patients.

Plasma was collected from RA patients prior to the initiation of treatment with DMARDs and lipid mediator concentrations established using LC-MS/MS-based lipid mediator profiling (see ‘Methods’ for details). a, b OPLS-DA analysis of peripheral blood lipid mediator concentrations for DMARD responders (Resp) and DMARD non-responders (Non-Resp). a Two-dimensional score plot. Grey circle represents the 95% confidence regions. b Two-dimensional loading plots. Lipid mediators with VIP score >1 are upregulated in Non-Resp and denoted in blue. Results are representative of n = 30 Resp and n = 22 Non-Resp. c Percent accuracy score of prediction models based on the combination of all lipid mediators identified and quantified (AL LM) or individual fatty acid metabolomes as indicated. Clin. Score = clinical score (see ‘Methods’ for parameters included). d ROC curves and AUC values (provided in brackets) for predictive models. e Classification predictions for each class (sensitivity and specificity) of the n-3 DPA model. Green indicates the samples that were predicted as Resp while blue indicates those patients predicted Non-Resp. Percentages indicate true positives (Resp class) and true negatives (Non-Resp class). f Relevance of lipid mediators in the prediction performance of the “ALL LM” model based on decreasing accuracy. g Percent accuracy score of models using the indicated SPM. h ROC curves and AUC values for predictive models based on the indicated SPM. All the models were created using the random forest methodology (“randomForest” package from R). Source data are provided as a Source data file.