Background: Asthma is a heterogenous disease and eosinophilic asthma has been defined as high levels of eosinophils in blood or sputum, while neutrophilic asthma is characterised by high sputum neutrophilia.

Aim: We investigate whether eosinophilic and neutrophilic asthma have unique molecular profiles using proteomic data from serum and sputum supernatants.

Methods: Proteomic profiles (1123 and 1128 assayed proteins) were obtained in sputum (n=182) and serum (n=574) from two cohorts of asthma patients. With eosinophilic/neutrophilic high and low asthma as the outcome, LASSO-penalised logistic regression in a stability selection framework and Random Forest variable selection were used to identify sparse sets of serum or sputum proteins jointly discriminating between the two classes.

Results: Overall, 13 serum proteins (including PAPP-A and CCL28) and 1 sputum protein (PAPP-A) were predictive of eosinophilic asthma yielding area under the receiver operating characteristic curve (AUROC) results of 0.83 (95% CI 0.82-0.83) and 0.81 (95% CI 0.80-0.81), respectively. For neutrophilic asthma, 12 serum proteins were identified, including MMP-9 and BLC, yielding an AUROC of 0.91 (95% CI 0.90-0.92) but no sputum proteins were stably selected.

Conclusions: Selected serum protein biomarkers can efficiently predict eosinophilic and neutrophilic asthma status. Such serum proteins may represent a more scalable alternative to sputum for patient characterisation and therapeutic definitions. Of the selected proteins, PAPP-A was associated with eosinophilic asthma in both serum and sputum. It may be involved in the development of asthma subtypes via the regulation of insulin like growth factors.