Bronchoalveolar Lavage Lipidomic Profiles Can Predict Short-Term Changes in Lung Function in Lung Transplant Recipients

2020 
Purpose Spirometric lung function is one of the central clinical parameters to assess allograft function after lung transplantation (LTX). Reduction in FEV1 can be the result of infections, rejections episodes and most importantly development of CLAD. Currently, there is no means to predict short-term changes in lung function after LTX. Methods Bronchoalveolar samples of patients after standard double LTX were taken during follow-up bronchoscopy at multiple different time points. Lipidomic, metabolomic, flow cytometric and bacterial 16S rRNA gene sequencing (microbiome) data were analyzed. We then used a machine learning approach to predict future changes in FEV1 from these sample data to characterize patient lung function trajectories. We trained support vector machine (SVM) regressors on the collected lipidomic, metabolomic, microbiome and flow cytometry datasets. Changes in FEV1 within 30, 60 or 90 days after lavage sample collection were used as response variables. To train hyper-parameters and evaluate model accuracy, a nested leave-one-out cross validation scheme was used. Model accuracy was benchmarked against a model trained on clinical metadata. Results At a prediction timeframe of 30 days, lipidomic data were available from 34 samples of 20 patients. Lipidomics showed the highest predictive power for short-term changes 30 and 60 days after sample collection. Prediction accuracy (R2) of intra-alveolar lipid composition for a 30-day projection was 0.24, meaning that BAL lipid profiles could explain more than 20 percent of total variation in relative change in FEV1 for this time span. R² for clinical metadata alone was only 0.1, and metabolomics, microbiome and FACS analysis of BAL showed no predictive accuracy at this time point. Conclusion Our results suggest that the intra-alveolar lipid composition is a powerful predictor of short-term changes in lung allograft function. This could potentially facilitate pre-emptive therapeutic interventions.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    0
    References
    0
    Citations
    NaN
    KQI
    []