Abstract Rationale The lack of reliable diagnostic and prognostic biomarkers represents an unmet clinical need in sarcoidosis. We leverage bronchoalveolar lavage fluid (BALF) and proteomics to develop classifiers for diagnostic accuracy and disease progression in pulmonary sarcoidosis. Methods Sarcoidosis cases met ATS criteria, were untreated, and had limited extrapulmonary involvement were categorized as progressive (P, n = 36) or non-progressive (NP, n = 46) based on pulmonary function decline, radiographic worsening, or treatment need (Table 1). BALF from 104 sarcoidosis patients (median age 53 years, 55% male), 37 healthy controls (HC), and 29 disease controls (DC) was studied. Medium- and low-abundance BALF proteins were trypsin-digested, labeled with TMTpro 18-plex tags, and processed via 2D-LC MS/MS. Peptide spectral matching was conducted in Proteome Discoverer 3.1.1.93 against the Human Universal Proteome database (downloaded 8/18/2023). Data were log-transformed, missing values imputed with SVDmiss, and batch effects corrected using ComBat algorithm. The processed proteins were selected for classifier development employing sparse distance-weighted discrimination, logistic regression, and random forest techniques; classifier performance was evaluated using AUC under leave-one-out cross-validation. Results Age differed significantly among HC, DC, and sarcoidosis groups but not between P and NP cases. Sex and race showed non-significant trends. PFTs and Scadding stage varied significantly across groups. Out of 2,637 proteins, 1,585 with less than 50% missing quantification were retained for analysis. A stepwise LR model with six proteins achieved an AUC of 0. 93 for distinguishing HC from sarcoidosis, while an eight-protein model distinguished DC from sarcoidosis: AUC 0. 89. A single-protein model differentiated P- from NP-sarcoidosis with an AUC of 0. 76. Since PFTs were unavailable for HC, the HC versus sarcoidosis comparison was adjusted for age, sex, and race (AUC = 0. 93). For DC versus sarcoidosis and P- versus NP, models were adjusted for age, sex, race, FVC, and FEV 1, resulting in AUCs of 0. 95 and 0. 78, respectively. We tested 93 peptides representing these proteins to identify potential transitions for Parallel Reaction Monitoring-MS, to enable a clinically translatable assay for robust protein quantification. 53 peptide transitions show promise for quantifying proteins that distinguish HC, DC, and sarcoidosis cases, and also predicting disease progression. Conclusion Promising findings support the development of BALF protein classifiers that differentiate between sarcoidosis and HC/DC and between progressive and non-progressive forms. If validated in independent cohorts, these classifiers could serve as the basis for clinically useful diagnostic and prognostic biomarkers in sarcoidosis. This abstract is funded by: NIH R01HL153613
Bhargava et al. (Fri,) studied this question.