Respiratory Research Analysis
Q3 2025 respiratory research coalesced around interpretable AI, tractable host–pathogen targets, and deployable platforms. Methodologically, an AI deep generative single‑cell model (UNAGI) and externally validated EHR/NLP and oximetry‑only tools advanced scalable diagnosis and therapy discovery. Infectious disease science matured from mechanism to translation: structural virology mapped HKU5 merbecovirus ACE2 usage, a chronic A. baumannii model enabled persistence/coinfection studies, and an NDV
Summary
Q3 2025 respiratory research coalesced around interpretable AI, tractable host–pathogen targets, and deployable platforms. Methodologically, an AI deep generative single‑cell model (UNAGI) and externally validated EHR/NLP and oximetry‑only tools advanced scalable diagnosis and therapy discovery. Infectious disease science matured from mechanism to translation: structural virology mapped HKU5 merbecovirus ACE2 usage, a chronic A. baumannii model enabled persistence/coinfection studies, and an NDV‑based COVID‑19 booster proved immunogenic and locally manufacturable. Clinical trial signals included a four‑month clofazimine TB regimen and disease‑directed efficacy of inhaled molgramostim for autoimmune PAP. Targeted oncology progressed with first‑line limertinib efficacy in EGFR‑mutant NSCLC, while integrative biomarker‑genetic analyses nominated IL‑6 inhibition for pleural infection.
Selected Articles
1. A deep generative model for deciphering cellular dynamics and in silico drug discovery in complex diseases.
UNAGI models time-series single-cell transcriptomic dynamics to prioritize drug candidates; in idiopathic pulmonary fibrosis, it predicted repurposable agents and validated nifedipine’s anti-fibrotic effects in human precision-cut lung slices with proteomic support.
Impact: Bridges single-cell disease trajectories with actionable drug prioritization and ex vivo validation, accelerating respiratory therapeutic discovery.
Clinical Implications: Offers a preclinical prioritization platform for repurposing and early-phase trials in pulmonary fibrosis; informs candidate and biomarker selection.
Key Findings
- Captured time-resolved single-cell disease trajectories and improved drug perturbation modeling.
- Predicted candidates for pulmonary fibrosis; nifedipine showed anti-fibrotic effects ex vivo.
- Proteomic validation supported inferred dynamics; framework generalized to other diseases.
2. A chronic Acinetobacter baumannii pneumonia model to study long-term virulence factors, antibiotic treatments, and polymicrobial infections.
The authors developed a low-dose intranasal, tlr4-mutant mouse model sustaining A. baumannii lung infection for ≥3 weeks, revealing a late-stage adhesin (InvL) required for persistence, distinguishing antibiotics that sterilize versus allow persister formation, and demonstrating outcome modulation by coinfecting species.
Impact: Enables mechanistic and therapeutic study of chronic respiratory infection, antibiotic persistence, and polymicrobial interactions—areas poorly captured by acute models.
Clinical Implications: Provides a platform to evaluate regimens targeting persisters and to identify late-stage virulence targets (e.g., InvL) that could guide novel therapeutics; human-relevant validation is the next step.
Key Findings
- Established a ≥3-week chronic A. baumannii lung infection model with low inocula in tlr4-mutant mice.
- Identified InvL as a persistence-critical late-stage adhesin.
- Differentiated sterilizing antibiotics from those associated with persister formation; coinfections modulated outcomes.
3. Phase 2/3 study evaluating safety, immunogenicity, and noninferiority of single booster dose of AVX/COVID-12 vaccine.
A double-blind, active-controlled phase 2/3 noninferiority RCT (n=4,056) of an NDV-LaSota HexaPro-S booster (AVX/COVID-12) showed favorable safety and robust neutralizing antibody responses to ancestral SARS-CoV-2 and Omicron BA.2/BA.5, with CD8 IFN-γ responses, supporting locally manufacturable, low-cost deployment.
Impact: Demonstrates an affordable, locally manufacturable vaccine booster with variant-specific immunogenicity, directly addressing equity and cold-chain barriers in LMICs.
Clinical Implications: NDV-based boosters can be considered by public health programs where mRNA logistics are impractical; effectiveness against infection/hospitalization and durability against new variants should be further evaluated.
Key Findings
- Double-blind, active-controlled phase 2/3 noninferiority design with 4,056 participants.
- Neutralizing antibodies to ancestral and Omicron BA.2/BA.5 plus CD8+ IFN-γ responses.
- Platform enables cost-effective local manufacturing suitable for LMIC deployment.
4. HKU5 bat merbecoviruses engage bat and mink ACE2 as entry receptors.
Using pseudotyped/full-length viruses, cryo-EM, and mutagenesis, HKU5 merbecoviruses were shown to use Pipistrellus abramus ACE2 and bind mink/stoat ACE2, but not human ACE2 or DPP4; sera from MERS-CoV vaccines poorly neutralized HKU5, underscoring surveillance and pan-merbecovirus vaccine needs.
Impact: Defines receptor usage and structural interfaces for HKU5, directly informing zoonotic risk assessment, host-range prediction, and priorities for broad merbecovirus immunogen design.
Clinical Implications: Guides receptor-focused surveillance in bats and mustelids, cautions limited cross-protection from existing MERS vaccines, and informs antigen selection for pan-merbecovirus strategies.
Key Findings
- HKU5 uses bat (Pipistrellus abramus) ACE2 and does not use human ACE2 or DPP4.
- Cryo-EM and mutagenesis reveal a distinctive spike–ACE2 interface versus other ACE2-using coronaviruses.
- HKU5 engages mink and stoat ACE2, implicating mustelids as potential intermediate hosts.
5. Transparent artificial intelligence-enabled interpretable and interactive sleep apnea assessment across flexible monitoring scenarios.
Apnea Interact Xplainer, an interpretable AI, was trained and externally validated on 15,807 polysomnograms across seven multi-ethnic cohorts, achieving 0.738–0.810 accuracy for four-level severity classification, R²=0.92–0.96 for AHI prediction, and 0.970 sensitivity for early detection using oximetry-only inputs.
Impact: Addresses adoption barriers by providing externally validated, interpretable AI that functions with limited-signal inputs, enabling scalable screening and transparent clinician–patient workflows.
Clinical Implications: Supports oximetry-based screening pathways, routine hypoxic burden reporting alongside AHI, and triage strategies for prioritizing full polysomnography or urgent therapy.
Key Findings
- Externally validated on 15,807 PSGs with severity accuracy 0.738–0.810 and AHI prediction R²=0.92–0.96.
- Achieved 0.970 sensitivity for early detection using oximetry-only input.
- Delivered interpretable visualizations that support clinical auditability and decision-making.
6. Four-month clofazimine regimen for susceptible pulmonary TB: a randomized clinical trial.
CORTAIL, a multicenter randomized non-inferiority trial (n=322), found a 16-week clofazimine-based regimen comparable to the standard 24-week therapy in relapse, sputum conversion, and bacteriologic cure, with acceptable safety.
Impact: Provides high-level RCT support for safely shortening first-line treatment of drug-susceptible pulmonary TB from six to four months.
Clinical Implications: If validated broadly, a 4-month clofazimine-containing regimen could be adopted as a new standard, especially where adherence to longer courses is challenging.
Key Findings
- Non-inferiority for relapse at 3 months post-treatment within predefined margin.
- No significant difference in 1-year relapse, sputum conversion, or bacteriologic cure.
- Acceptable safety profile across 11 centers.
7. Open-source computational pipeline flags instances of acute respiratory distress syndrome in mechanically ventilated adult patients.
An interpretable, open-source pipeline operationalized the Berlin Definition using radiology reports and clinician notes to automatically flag ARDS in mechanically ventilated adults; external validation achieved 93.5% sensitivity with a 17.4% false-positive rate, far exceeding human documentation.
Impact: Addresses ARDS under-recognition with a reproducible, externally validated tool that can be embedded in EHRs to standardize care and streamline trial enrollment.
Clinical Implications: Hospital deployment could increase timely ARDS recognition, enable protocolized ventilator management, real-time quality metrics, and improve clinical trial readiness.
Key Findings
- Interpretable classifiers mapped Berlin Definition elements from text sources.
- External validation yielded 93.5% sensitivity with a 17.4% false-positive rate.
- Automated identification far exceeded human documentation rates.
8. Phase 3 Trial of Inhaled Molgramostim in Autoimmune Pulmonary Alveolar Proteinosis.
In a 48‑week double‑blind phase 3 RCT (n=164), once‑daily inhaled molgramostim (300 μg) significantly improved DLCO at 24 and 48 weeks versus placebo and produced clinically meaningful improvement in SGRQ total score at 24 weeks, with similar adverse event rates between groups.
Impact: Largest rigorous RCT showing a targeted inhaled biologic improves physiology and patient‑reported outcomes in a rare pulmonary disease, closing a key therapeutic gap.
Clinical Implications: Molgramostim may become a disease‑directed therapy for autoimmune PAP, potentially reducing whole‑lung lavage and improving gas exchange and quality of life; guideline and reimbursement updates will be needed.
Key Findings
- 24‑week DLCO improvement: +9.8 pp with molgramostim vs +3.8 pp with placebo (difference 6.0; P<0.001).
- DLCO benefit persisted to 48 weeks; SGRQ total score improved at 24 weeks.
- Adverse event rates were similar between groups.
9. Efficacy and safety of limertinib versus gefitinib as first-line treatment for locally advanced or metastatic non-small-cell lung cancer with EGFR-sensitising mutation: a randomised, double-blind, double-dummy, phase 3 trial.
A multicenter, double-blind phase 3 RCT (n=337) showed limertinib more than doubled ICR-assessed PFS versus gefitinib (20.7 vs 9.7 months; HR 0.44) with comparable grade ≥3 adverse events, supporting limertinib as a first-line EGFR TKI option.
Impact: Phase 3 superiority on disease control in EGFR-mutant NSCLC likely influences first-line standards, especially where gefitinib remains in use.
Clinical Implications: Consider limertinib among first-line EGFR TKI choices; comparative data versus osimertinib and CNS outcomes will refine positioning.
Key Findings
- Median ICR-assessed PFS 20.7 vs 9.7 months (HR 0.44; p<0.0001).
- Grade ≥3 treatment-related AEs were 25% in each arm.
- Randomised, double-blind, double-dummy design across 56 centers with stratification.
10. The role of interleukin-6 signalling in pleural infection: observational and genetic analyses.
Paired pleural fluid and serum measures (n=76) showed pleural IL‑6 ~5,000‑fold higher than serum, correlating with severity and length of stay; two‑sample Mendelian randomization using IL6R variants (1,601 cases vs 830,709 controls) predicted a large protective effect of IL‑6 pathway inhibition on pleural infection incidence.
Impact: Convergent biomarker evidence and genetic causal inference identify IL‑6 as a tractable therapeutic target in pleural infection, directly informing randomized trials.
Clinical Implications: Supports pleural IL‑6 measurement for risk stratification and prioritizes clinical trials of IL‑6 pathway inhibitors (e.g., tocilizumab) for empyema/pleural infection.
Key Findings
- Pleural IL‑6 median was ~5,000‑fold higher than serum and correlated with severity and length of stay.
- Mendelian randomization using IL6R variants predicted a protective effect of IL‑6 inhibition (OR ~0.23 per SD CRP decrease).
- Integrates biomarker and genetic data to nominate IL‑6 as a therapeutic target.