Identification of genetic profile and biomarkers involved in acute respiratory distress syndrome

Abstract

Purpose: The purpose of this study was to profile genetic causal factors of acute respiratory distress syndrome (ARDS) and early predict patients at high ARDS risk.

Methods: We performed a phenome-wide Mendelian Randomization analysis through summary statistics of an ARDS genome-wide association study (1250 cases and 1583 controls of European ancestry) and 33,150 traits. Transcriptomic data from human blood and lung tissues of a preclinical mouse model were used to validate biomarkers, which were further used to construct a prediction model and nomogram.

Results: A total of 1736 traits, including 1223 blood RNA, 159 plasma proteins, and 354 non-gene phenotypes (classified by Biochemistry, Anthropometry, Disease, Nutrition and Habit, Immunology, and Treatment), exhibited a potentially causal relationship with ARDS development, which were accessible through a user-friendly interface platform called CARDS (Causal traits for Acute Respiratory Distress Syndrome). Regarding candidate blood RNA, four genes were validated, namely TMEM176B, SLC2A5, CDC45, and VSIG8, showing differential expression in blood of ARDS patients compared to controls, as well as dynamic expression in mouse lung tissues. Importantly, the addition of four blood genes and five immune cell proportions significantly improved the prediction performance of ARDS development, with 0.791 of the area under the curve from receiver-operator characteristic, compared to 0.725 for the basic model consisting of Acute Physiology and Chronic Health Evaluation (APACHE) III Score, sex, body mass index, bacteremia, and sepsis. A model-based nomogram was also developed for the clinical practice.

Conclusion: This study identifies a wide range of ARDS relevant factors and develops a promising prediction model, enhancing early clinical management and intervention for ARDS development.

Keywords: Acute respiratory distress syndrome; Biomarker; Causal factor; Mendelian randomization analysis; Phenome-wide association study.

Fig. 1

Flowchart of the study design. This study contains three main stages: identification, validation, and application. In the identification stage, an MR-pheWAS was conducted to profile potential individual causal factors and biomarkers associated with ARDS development. In the validation stage, we used peripheral blood transcriptome analysis and a preclinical mouse model to validate biomarkers implicated in the pathogenesis of ARDS. In the application stage, we developed a CARDS platform including individual causal factors and biomarkers, as well as a risk prediction tool to enhance clinical management and informed decision-making for ARDS

Fig. 2

Results of causal factors on ARDS via MR-pheWAS and corresponding visualization. A Bar chart categorizing 354 non-gene phenotypes into six subtypes manually. B Overview of CARDS (https://mulongdu.shinyapps.io/cards/). CARDS includes three modules: “MR-pheWAS”, “ARDS RNA-Seq”, and “Citation & Contact”. The usage is illustrated using “transmembrane protein” as an example. C Correlation of effect size between blood genes and their encoding plasma proteins according to MR

Fig. 3

Expression patterns of four candidate causal genes. A–D Expression levels of TMEM176B, SLC2A5, CDC45, and VSIG8 in 160 ARDS cases and 142 non-ARDS controls. P values were calculated by Wilcoxon signed-rank test. E–H Gene expression of four genes in mice after exposure to LPS for 1,2,4,18 h (s) from GSE9314. P values were calculated via ANOVA test

Fig. 4

Correlation of gene expression and immune cell fraction in ARDS cases and non-ARDS controls. Correlation matrix plot showed pairwise similarity (Spearman correlation) between four blood genes and abundances of five immune cell types across 142 non-ARDS controls (A) and 160 ARDS cases (B). The size and color shade of the squares in each cell represents the strength of the correlation, with a star (*) indicating statistical significance (P < 0.05)

Fig. 5

Risk prediction models for ARDS. A ROC curve depicting the performance of basic model, basic model with the combination of four blood genes, basic model with the combination of five immune cell proportions, and basic model with above both. B Nomogram for prediction of ARDS occurrence. C Calibration curve of the nomogram for predicting ARDS occurrence. The black columns and red dots show predicted values, while the gray columns and blue dots show actual values. The corresponding calibration was performed via Hosmer–Lemeshow test