Daily Respiratory Research Analysis
Three impactful studies span mechanistic discovery and clinical translation in respiratory medicine: a deep generative model (UNAGI) deciphers disease dynamics in idiopathic pulmonary fibrosis and predicts anti-fibrotic drug candidates validated in human lung tissue; a meta-analysis quantifies the time-to-benefit of low-dose CT lung cancer screening to guide patient selection; and deep learning on chest radiographs accurately detects pulmonary hypertension with catheterization-validated performa
Summary
Three impactful studies span mechanistic discovery and clinical translation in respiratory medicine: a deep generative model (UNAGI) deciphers disease dynamics in idiopathic pulmonary fibrosis and predicts anti-fibrotic drug candidates validated in human lung tissue; a meta-analysis quantifies the time-to-benefit of low-dose CT lung cancer screening to guide patient selection; and deep learning on chest radiographs accurately detects pulmonary hypertension with catheterization-validated performance.
Research Themes
- AI-enabled discovery and diagnosis in respiratory disease
- Evidence-based optimization of lung cancer screening
- Translational single-cell systems biology in pulmonary fibrosis
Selected Articles
1. A deep generative model for deciphering cellular dynamics and in silico drug discovery in complex diseases.
UNAGI models time-evolving single-cell states to elucidate idiopathic pulmonary fibrosis progression and prioritizes drug candidates; nifedipine’s anti-fibrotic effect was confirmed in human precision-cut lung slices. The framework generalizes across diseases (including COVID), combining computational innovation with proteomic and ex vivo validation.
Impact: This work pioneers a disease-informed generative model that links single-cell dynamics to actionable drug predictions validated in human tissue, potentially accelerating therapeutic discovery in pulmonary fibrosis.
Clinical Implications: While not yet clinical, UNAGI can prioritize repurposable agents (e.g., nifedipine) and new targets for idiopathic pulmonary fibrosis, informing preclinical pipelines and the design of early-phase trials.
Key Findings
- UNAGI captured time-resolved single-cell disease dynamics and improved drug perturbation modeling.
- In idiopathic pulmonary fibrosis, UNAGI identified candidate therapeutics; nifedipine’s anti-fibrotic effect was validated in human precision-cut lung slices.
- Proteomic data supported the inferred cellular dynamics, and the approach generalized to other diseases including COVID.
Methodological Strengths
- Integration of time-series single-cell transcriptomics with proteomic validation.
- Ex vivo confirmation in human precision-cut lung slices strengthens translational relevance.
Limitations
- Validation of drug predictions was limited in scope (e.g., nifedipine) and lacks in vivo/clinical outcomes.
- Model performance and generalizability across diverse patient populations and tissue contexts require further evaluation.
Future Directions: Prospective preclinical testing of prioritized candidates (dose-response, mechanism), multi-center single-cell cohorts for external validation, and early-phase trials guided by model-informed biomarkers.
Human diseases are characterized by intricate cellular dynamics. Single-cell transcriptomics provides critical insights, yet a persistent gap remains in computational tools for detailed disease progression analysis and targeted in silico drug interventions. Here we introduce UNAGI, a deep generative neural network tailored to analyse time-series single-cell transcriptomic data. This tool captures the complex cellular dynamics underlying disease progression, enhancing drug perturbation modelling and screening. When applied to a dataset from patients with idiopathic pulmonary fibrosis, UNAGI learns disease-informed cell embeddings that sharpen our understanding of disease progression, leading to the identification of potential therapeutic drug candidates. Validation using proteomics reveals the accuracy of UNAGI's cellular dynamics analysis, and the use of the fibrotic cocktail-treated human precision-cut lung slices confirms UNAGI's predictions that nifedipine, an antihypertensive drug, may have anti-fibrotic effects on human tissues. UNAGI's versatility extends to other diseases, including COVID, demonstrating adaptability and confirming its broader applicability in decoding complex cellular dynamics beyond idiopathic pulmonary fibrosis, amplifying its use in the quest for therapeutic solutions across diverse pathological landscapes.
2. Timing of Screening Benefit for Lung Cancer With Low-Dose CT Imaging.
Across four RCTs, LDCT screening shows a measurable time-to-benefit: approximately 1.78 years to prevent one lung cancer death per 2,000 screened in NLST, with longer horizons for smaller screened cohorts. These estimates argue for incorporating TTB into eligibility decisions, especially for patients with limited life expectancy.
Impact: Quantifying time-to-benefit enables patient-centered screening decisions, aligning LDCT use with life expectancy and reducing low-value care.
Clinical Implications: Incorporate TTB thresholds into LDCT eligibility (e.g., avoid screening those unlikely to live ≥2–3 years), and align shared decision-making with realistic timelines of benefit.
Key Findings
- In NLST, preventing one lung cancer death required screening 2,000 individuals over 1.78 years (95% CI 0.60–5.27).
- TTB scales with the number screened: 2.87, 4.66, and 8.87 years per 1,000, 500, and 200 individuals screened, respectively.
- Findings were consistent across pooled RCT analyses, supporting robustness of TTB estimates.
Methodological Strengths
- Restriction to randomized controlled trials with mortality endpoints.
- Use of an established time-to-benefit analytic framework with sensitivity to cohort size.
Limitations
- Limited number of trials and potential heterogeneity in trial protocols and populations.
- TTB estimates may not capture individual-level variability (e.g., comorbidity, competing risks).
Future Directions: Incorporate patient-level data to refine TTB by comorbidity and frailty strata; integrate TTB into guideline algorithms and decision aids; evaluate impacts on overdiagnosis and resource allocation.
BACKGROUND: Increasing evidence supports lung cancer screening with low-dose CT (LDCT) imaging. However, the benefits of LDCT screening for lung cancer may not be immediate, making it unlikely to benefit patients with limited life expectancy. RESEARCH QUESTION: What is the time to benefit (TTB) from LDCT screening for individuals at high risk for lung cancer? STUDY DESIGN AND METHODS: Population-based randomized controlled trials of lung cancer screening using LDCT imaging and reporting mortality outcomes were systematically searched in PubMed. TTB was estimated for the National Lung Screening Trial (NLST) and the pooled data from 4 trials using an established analysis framework. RESULTS: This analysis included 4 trials encompassing 64,105 individuals. In the NLST (N = 53,452), to prevent 1 death from lung cancer, 2,000 individuals would need to be screened over 1.78 (95% CI, 0.60-5.27) years. On average, it took 2.87 (95% CI, 1.31-6.32), 4.66 (95% CI, 2.64-8.21), and 8.87 (95% CI, 5.12-15.37) years before 1 death from lung cancer was prevented for every 1,000, 500, and 200 individuals screened, respectively. These findings did not vary when added to other trials. INTERPRETATION: Our findings suggest that the clinical benefits of LDCT screening may not be appropriate for individuals with limited life expectancy. Integrating TTB estimates into patient selection criteria could help maximize the benefits of LDCT screening.
3. Deep Learning-Enhanced Noninvasive Detection of Pulmonary Hypertension and Subtypes via Chest Radiographs, Validated by Catheterization.
Deep learning models applied to chest radiographs detected pulmonary hypertension and CHD-PAH with high sensitivity, validated against right heart catheterization in internal and external cohorts. Performance supports use as a screening/triage tool, especially where advanced imaging is limited.
Impact: Provides an accessible, noninvasive screening method for PH with catheterization-validated accuracy, enabling earlier identification and referral in resource-limited settings.
Clinical Implications: Adopt DL-CXR screening to flag suspected PH/CHD-PAH for confirmatory RHC and specialized care; integrate into pathways to reduce diagnostic delay where echocardiography or advanced imaging access is limited.
Key Findings
- CXR-PH-Net achieved AUC 0.964 and sensitivity 0.902 in the internal test set.
- In RHC-confirmed cohorts, sensitivity was 0.902 (AUC 0.872) internally and 0.803 (AUC 0.811) externally.
- CXR-CHD-PAH-Net showed AUCs of 0.908 (internal) and 0.860 (external), with good sensitivity even among mild PH.
Methodological Strengths
- Large retrospective cohort with internal test set and RHC-confirmed internal and external validations.
- Use of catheterization as a reference standard enhances diagnostic validity.
Limitations
- External validation cohort was relatively small (n=90), limiting precision of performance estimates.
- Retrospective design; potential spectrum and selection biases; generalizability across populations requires further study.
Future Directions: Prospective, multi-ethnic validation linked to outcomes; workflow integration studies assessing impact on time-to-diagnosis and referral; exploration of model interpretability and calibration.
BACKGROUND: Pulmonary hypertension (PH) is a complex, life-threatening condition requiring noninvasive, accessible, and accurate diagnostic tools, particularly in resource-limited settings. Early and precise identification of PH and its subtypes is critical for effective management and timely intervention. RESEARCH QUESTION: Can deep learning (DL) methods applied to chest radiography (CXR) accurately detect PH and its subtype, congenital heart disease-associated pulmonary arterial hypertension (CHD-PAH)? STUDY DESIGN AND METHODS: A retrospective cohort study was conducted with 4,576 patients, including 2,288 patients with PH, who underwent CXR followed by right heart catheterization (RHC) or transthoracic echocardiography. DL models were developed and validated for detecting PH (CXR-PH-Net model) and CHD-PAH (CXR-CHD-PAH-Net model). Internal testing used a data set of 2,140 patients (1,070 patients with PH), and additional validation included an RHC-confirmed internal cohort (1,158 patients) and an external RHC cohort (90 patients) from 2 independent hospitals. Model performance was evaluated primarily using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. RESULTS: The CXR-PH-Net model achieved a sensitivity of 0.902 and an AUC of 0.964 for PH detection in the internal test set. In the RHC-confirmed cohort, sensitivity was 0.902 (AUC, 0.872) internally and 0.803 (AUC, 0.811) externally. The CXR-CHD-PAH-Net model demonstrated sensitivities of 0.859 and 0.870 with AUCs of 0.908 and 0.860 in the internal and external data sets, respectively. Meanwhile, the CXR-CHD-PAH-Net model showed favorable sensitivity in detecting CHD-PAH among patients with mild PH, with values of 0.813 and 0.846 in the internal and external datasets, respectively. INTERPRETATION: The CXR-PH-Net and CXR-CHD-PAH-Net models demonstrated high sensitivity as screening tools for PH and CHD-PAH, potentially facilitating early detection and triage for further evaluation, particularly in resource-limited settings. Further validation in diverse populations is warranted to enhance clinical generalizability. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov; No.: NCT05566002; URL: www. CLINICALTRIALS: gov.