The Proteomic Profile of Interstitial Lung Abnormalities

Research priorities for progressive pulmonary fibrosis in the UK

INTRODUCTION

Health research bodies recommend patient involvement and engagement in research and healthcare planning, although their implementation is not yet widespread. This deficiency extends to progressive pulmonary fibrosis (PPF), where crucial aspects remain unknown, including causal mechanisms, curative treatments and optimal symptom management. This study addresses these gaps by seeking stakeholders' perspectives to guide research and treatment directions.

METHOD

A priority-setting partnership was established to explore stakeholders' priorities in the diagnosis, treatment, management and care of PPF, including idiopathic pulmonary fibrosis which is the archetypal PPF. Stakeholders included people living with PPF, their carers, relatives and healthcare professionals involved in their management.

RESULTS

Through an online open-ended survey, 2542 responses were collected from 638 stakeholders. Thematic analysis identified 48 specific research questions, which were then cross-referenced with academic literature to pinpoint research gaps. Following the evidence check, 44 unanswered questions were shortlisted by 834 stakeholders in a second online survey. Ultimately, a top 10 priority list was established through consensus.The prioritised research questions include (1) improved diagnosis accuracy and timing, (2) development of new treatments, (3) enhanced accuracy in primary care, (4) optimal timing for drug and non-drug interventions, (5) effective cough treatment, (6) early intervention for PPF, (7) improved survival rates, (8) symptom reduction, (9) impact of interventions on life expectancy and (10) new treatments with reduced side effects.

CONCLUSION

Stakeholders' priorities can be summarised into five areas: early diagnosis, drug and non-drug treatments, survival and symptom management. Ideally, these topics should guide funding bodies and health policies.

Automated Interstitial Lung Abnormality Probability Prediction at CT: A Stepwise Machine Learning Approach in the Boston Lung Cancer Study

Background It is increasingly recognized that interstitial lung abnormalities (ILAs) detected at CT have potential clinical implications, but automated identification of ILAs has not yet been fully established. Purpose To develop and test automated ILA probability prediction models using machine learning techniques on CT images. Materials and Methods This secondary analysis of a retrospective study included CT scans from patients in the Boston Lung Cancer Study collected between February 2004 and June 2017. Visual assessment of ILAs by two radiologists and a pulmonologist served as the ground truth. Automated ILA probability prediction models were developed that used a stepwise approach involving section inference and case inference models. The section inference model produced an ILA probability for each CT section, and the case inference model integrated these probabilities to generate the case-level ILA probability. For indeterminate sections and cases, both two- and three-label methods were evaluated. For the case inference model, we tested three machine learning classifiers (support vector machine [SVM], random forest [RF], and convolutional neural network [CNN]). Receiver operating characteristic analysis was performed to calculate the area under the receiver operating characteristic curve (AUC). Results A total of 1382 CT scans (mean patient age, 67 years ± 11 [SD]; 759 women) were included. Of the 1382 CT scans, 104 (8%) were assessed as having ILA, 492 (36%) as indeterminate for ILA, and 786 (57%) as without ILA according to ground-truth labeling. The cohort was divided into a training set (n = 96; ILA, n = 48), a validation set (n = 24; ILA, n = 12), and a test set (n = 1262; ILA, n = 44). Among the models evaluated (two- and three-label section inference models; two- and three-label SVM, RF, and CNN case inference models), the model using the three-label method in the section inference model and the two-label method and RF in the case inference model achieved the highest AUC, at 0.87. Conclusion The model demonstrated substantial performance in estimating ILA probability, indicating its potential utility in clinical settings. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Zagurovskaya in this issue.

Machine Learning of Plasma Proteomics Classifies Diagnosis of Interstitial Lung Disease

Rationale: Distinguishing connective tissue disease-associated interstitial lung disease (CTD-ILD) from idiopathic pulmonary fibrosis (IPF) can be clinically challenging. Objectives: To identify proteins that separate and classify patients with CTD-ILD and those with IPF. Methods: Four registries with 1,247 patients with IPF and 352 patients with CTD-ILD were included in analyses. Plasma samples were subjected to high-throughput proteomics assays. Protein features were prioritized using recursive feature elimination to construct a proteomic classifier. Multiple machine learning models, including support vector machine, LASSO (least absolute shrinkage and selection operator) regression, random forest, and imbalanced Random Forest, were trained and tested in independent cohorts. The validated models were used to classify each case iteratively in external datasets. Measurements and Main Results: A classifier with 37 proteins (proteomic classifier 37 [PC37]) was enriched in the biological process of bronchiole development and smooth muscle proliferation and immune responses. Four machine learning models used PC37 with sex and age score to generate continuous classification values. Receiver operating characteristic curve analyses of these scores demonstrated consistent areas under the curve of 0.85-0.90 in the test cohort and 0.94-0.96 in the single-sample dataset. Binary classification demonstrated 78.6-80.4% sensitivity and 76-84.4% specificity in the test cohort and 93.5-96.1% sensitivity and 69.5-77.6% specificity in the single-sample classification dataset. Composite analysis of all machine learning models confirmed 78.2% (194 of 248) accuracy in the test cohort and 82.9% (208 of 251) in the single-sample classification dataset. Conclusions: Multiple machine learning models trained with large cohort proteomic datasets consistently distinguished CTD-ILD from IPF. Many of the identified proteins are involved in immune pathways. We further developed a novel approach for single-sample classification, which could facilitate honing the differential diagnosis of ILD in challenging cases and improve clinical decision making.

Proteomic Biomarkers of Survival in Idiopathic Pulmonary Fibrosis

Rationale: Idiopathic pulmonary fibrosis (IPF) causes progressive lung scarring and high mortality. Reliable and accurate prognostic biomarkers are urgently needed. Objectives: To identify and validate circulating protein biomarkers of IPF survival. Methods: High-throughput proteomic data were generated using prospectively collected plasma samples from patients with IPF from the Pulmonary Fibrosis Foundation Patient Registry (discovery cohort) and the Universities of California, Davis; Chicago; and Virginia (validation cohort). Proteins associated with three-year transplant-free survival (TFS) were identified using multivariable Cox proportional hazards regression. Those associated with TFS after adjustment for false discovery in the discovery cohort were advanced for testing in the validation cohort, with proteins maintaining TFS association with consistent effect direction considered validated. After combining cohorts, functional analyses were performed, and machine learning was used to derive a proteomic signature of TFS. Measurements and Main Results: Of 2,921 proteins tested in the discovery cohort (n = 871), 231 were associated with differential TFS. Of these, 140 maintained TFS association with consistent effect direction in the validation cohort (n = 355). After cohorts were combined, the validated proteins with the strongest TFS association were latent-transforming growth factor β-binding protein 2 (hazard ratio [HR], 2.43; 95% confidence interval [CI] = 2.09-2.82), collagen α-1(XXIV) chain (HR, 2.21; 95% CI = 1.86-2.39), and keratin 19 (HR, 1.60; 95% CI = 1.47-1.74). In decision curve analysis, a proteomic signature of TFS outperformed a similarly derived clinical prediction model. Conclusions: In the largest proteomic investigation of IPF outcomes performed to date, we identified and validated 140 protein biomarkers of TFS. These results shed important light on potential drivers of IPF progression.

Proteomic Biomarkers of Quantitative Interstitial Abnormalities in COPDGene and CARDIA Lung Study

Rationale: Quantitative interstitial abnormalities (QIAs) are early measures of lung injury automatically detected on chest computed tomography scans. QIAs are associated with impaired respiratory health and share features with advanced lung diseases, but their biological underpinnings are not well understood. Objectives: To identify novel protein biomarkers of QIAs using high-throughput plasma proteomic panels within two multicenter cohorts. Methods: We measured the plasma proteomics of 4,383 participants in an older, ever-smoker cohort (COPDGene [Genetic Epidemiology of Chronic Obstructive Pulmonary Disease]) and 2,925 participants in a younger population cohort (CARDIA [Coronary Artery Disease Risk in Young Adults]) using the SomaLogic SomaScan assays. We measured QIAs using a local density histogram method. We assessed the associations between proteomic biomarker concentrations and QIAs using multivariable linear regression models adjusted for age, sex, body mass index, smoking status, and study center (Benjamini-Hochberg false discovery rate-corrected P ⩽ 0.05). Measurements and Main Results: In total, 852 proteins were significantly associated with QIAs in COPDGene and 185 in CARDIA. Of the 144 proteins that overlapped between COPDGene and CARDIA, all but one shared directionalities and magnitudes. These proteins were enriched for 49 Gene Ontology pathways, including biological processes in inflammatory response, cell adhesion, immune response, ERK1/2 regulation, and signaling; cellular components in extracellular regions; and molecular functions including calcium ion and heparin binding. Conclusions: We identified the proteomic biomarkers of QIAs in an older, smoking population with a higher prevalence of pulmonary disease and in a younger, healthier community cohort. These proteomics features may be markers of early precursors of advanced lung diseases.

The Role of Lung Microbiome in Fibrotic Interstitial Lung Disease-A Systematic Review

Interstitial Lung Disease (ILD) involves lung disorders marked by chronic inflammation and fibrosis. ILDs include pathologies like idiopathic pulmonary fibrosis (IPF), connective tissue disease-associated ILD (CTD-ILD), hypersensitivity pneumonitis (HP) or sarcoidosis. Existing data covers pathogenesis, diagnosis (especially using high-resolution computed tomography), and treatments like antifibrotic agents. Despite progress, ILD diagnosis and management remains challenging with significant morbidity and mortality. Recent focus is on Progressive Fibrosing ILD (PF-ILD), characterized by worsening symptoms and fibrosis on HRCT. Prevalence is around 30%, excluding IPF, with a poor prognosis. Early diagnosis is crucial for optimizing outcomes in PF-ILD individuals. The lung microbiome comprises all the microorganisms that are in the respiratory tract. Relatively recent research try to evaluate its role in respiratory disease. Healthy lungs have a diverse microbial community. An imbalance in bacterial composition, changes in bacterial metabolic activities, or changes in bacterial distribution within the lung termed dysbiosis is linked to conditions like COPD, asthma and ILDs. We conducted a systematic review of three important scientific data base using a focused search strategy to see how the lung microbiome is involved in the progression of ILDs. Results showed that some differences in the composition and quality of the lung microbiome exist in ILDs that show progressive fibrosing phenotype. The results seem to suggest that the lung microbiota could be involved in ILD progression, but more studies showing its exact pathophysiological mechanisms are needed.

A Polygenic Risk Score for Idiopathic Pulmonary Fibrosis and Interstitial Lung Abnormalities

Rationale: In addition to rare genetic variants and the MUC5B locus, common genetic variants contribute to idiopathic pulmonary fibrosis (IPF) risk. The predictive power of common variants outside the MUC5B locus for IPF and interstitial lung abnormalities (ILAs) is unknown. Objectives: We tested the predictive value of IPF polygenic risk scores (PRSs) with and without the MUC5B region on IPF, ILA, and ILA progression. Methods: We developed PRSs that included (PRS-M5B) and excluded (PRS-NO-M5B) the MUC5B region (500-kb window around rs35705950-T) using an IPF genome-wide association study. We assessed PRS associations with area under the receiver operating characteristic curve (AUC) metrics for IPF, ILA, and ILA progression. Measurements and Main Results: We included 14,650 participants (1,970 IPF; 1,068 ILA) from six multi-ancestry population-based and case-control cohorts. In cases excluded from genome-wide association study, the PRS-M5B (odds ratio [OR] per SD of the score, 3.1; P = 7.1 × 10-95) and PRS-NO-M5B (OR per SD, 2.8; P = 2.5 × 10-87) were associated with IPF. Participants in the top PRS-NO-M5B quintile had ∼sevenfold odds for IPF compared with those in the first quintile. A clinical model predicted IPF (AUC, 0.61); rs35705950-T and PRS-NO-M5B demonstrated higher AUCs (0.73 and 0.7, respectively), and adding both genetic predictors to a clinical model yielded the highest performance (AUC, 0.81). The PRS-NO-M5B was associated with ILA (OR, 1.25) and ILA progression (OR, 1.16) in European ancestry participants. Conclusions: A common genetic variant risk score complements the MUC5B variant to identify individuals at high risk of interstitial lung abnormalities and pulmonary fibrosis.

Recent Advances of Artificial Intelligence Applications in Interstitial Lung Diseases

Interstitial lung diseases (ILDs) comprise a rather heterogeneous group of diseases varying in pathophysiology, presentation, epidemiology, diagnosis, treatment and prognosis. Even though they have been recognized for several years, there are still areas of research debate. In the majority of ILDs, imaging modalities and especially high-resolution Computed Tomography (CT) scans have been the cornerstone in patient diagnostic approach and follow-up. The intricate nature of ILDs and the accompanying data have led to an increasing adoption of artificial intelligence (AI) techniques, primarily on imaging data but also in genetic data, spirometry and lung diffusion, among others. In this literature review, we describe the most prominent applications of AI in ILDs presented approximately within the last five years. We roughly stratify these studies in three categories, namely: (i) screening, (ii) diagnosis and classification, (iii) prognosis.