Non-invasive hair metabolome analysis reveals GERD and thyroid dysfunction in pulmonary fibrosis patients

Idiopathic pulmonary fibrosis (IPF), the most common fibrotic interstitial lung disease (ILD), is characterized by poor (median ∼3 years) survival. While comorbidities such as gastroesophageal reflux disease (GERD) and thyroid dysfunction are associated with increased IPF occurrence, they are frequently underdiagnosed in ILD patients, hampering timely intervention and effective management. Recent advances suggest that the hair metabolome may serve as a non-invasive biomarker for pulmonary fibrosis diagnosis and progression. Building on these findings, our study explores whether comorbidities such as GERD and thyroid dysfunction can similarly be identified through hair metabolomics, potentially enabling earlier detection and personalized treatment approaches in ILD patients.

ILD patient hair metabolome data (n = 53) were obtained from a previous study, in which hair specimens collected were processed for metabolite extraction using 2D-LC/MS-MS. The data were used to train a stacked ensemble machine learning workflow tuned via a rigorous combination of cross-validation, feature selection and testing with a hold-out dataset to diagnose GERD or thyroid dysfunction.

Prediction of GERD achieved AUROC_TEST = 0.824 and PRAUC_TEST = 0.849 with ten metabolites, with the top five being uracil, porphobilinogen, threonylphenylalanine, L-alanyl-L-proline, and valylvaline. Prediction of thyroid dysfunction achieved AUROC_TEST = 0.870 and PRAUC_TEST = 0.781 using two metabolites: leucylproline and desthiobiotin. Key metabolic pathways complementary to these metabolites and with GERD or thyroid dysfunction were identified.

Comorbidities such as GERD and thyroid dysfunction can be non-invasively evaluated from the hair metabolome of ILD patients. Longer-term, this approach may help to improve IPF treatment strategies.