Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Jul 24:15:1441838.
doi: 10.3389/fimmu.2024.1441838. eCollection 2024.

Machine learning-based derivation and validation of three immune phenotypes for risk stratification and prognosis in community-acquired pneumonia: a retrospective cohort study

Qiangqiang Qin #  1 , Haiyang Yu #  1 , Jie Zhao #  2 , Xue Xu  1 , Qingxuan Li  3 , Wen Gu  1 , Xuejun Guo  1
Affiliations

Machine learning-based derivation and validation of three immune phenotypes for risk stratification and prognosis in community-acquired pneumonia: a retrospective cohort study

Qiangqiang Qin et al. Front Immunol. .

Abstract

Background: The clinical presentation of Community-acquired pneumonia (CAP) in hospitalized patients exhibits heterogeneity. Inflammation and immune responses play significant roles in CAP development. However, research on immunophenotypes in CAP patients is limited, with few machine learning (ML) models analyzing immune indicators.

Methods: A retrospective cohort study was conducted at Xinhua Hospital, affiliated with Shanghai Jiaotong University. Patients meeting predefined criteria were included and unsupervised clustering was used to identify phenotypes. Patients with distinct phenotypes were also compared in different outcomes. By machine learning methods, we comprehensively assess the disease severity of CAP patients.

Results: A total of 1156 CAP patients were included in this research. In the training cohort (n=809), we identified three immune phenotypes among patients: Phenotype A (42.0%), Phenotype B (40.2%), and Phenotype C (17.8%), with Phenotype C corresponding to more severe disease. Similar results can be observed in the validation cohort. The optimal prognostic model, SuperPC, achieved the highest average C-index of 0.859. For predicting CAP severity, the random forest model was highly accurate, with C-index of 0.998 and 0.794 in training and validation cohorts, respectively.

Conclusion: CAP patients can be categorized into three distinct immune phenotypes, each with prognostic relevance. Machine learning exhibits potential in predicting mortality and disease severity in CAP patients by leveraging clinical immunological data. Further external validation studies are crucial to confirm applicability.

Keywords: community-acquired pneumonia; immune phenotype; machine learning; risk stratification; unsupervised clustering.

PubMed Disclaimer

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

Figure 1
Figure 1
The flowchart of this research.
Figure 2
Figure 2
Consensus Clustering and visualization. (A) Identification of three immune phenotypes of community acquired pneumonia (CAP) patients by consensus clustering. (B) Cumulative distribution function (CDF) curve illustrated consensus distribution for each phenotype. (C) T-distributed stochastic neighbor embedding (t-SNE) method successfully divided CAP patients into three distinct immune phenotypes. The purple dot represent patients belong to phenotype A. Patients with phenotype B are represented by a yellow dot, and those with phenotype C by a blue dot. CAP, community acquired pneumonia; t-SNE, T-distributed stochastic neighbor embedding; CDF, Cumulative distribution function.
Figure 3
Figure 3
Association and variation between clinical immunological indicators and three phenotypes. Chord diagram (A–D) of the association between clinical immunological variables and each phenotype in training cohort. Different phenotypes were shown in different colors: phenotype A is purple, phenotype B is blue, and phenotype C is green. Rank plot (E–G) of variable mean among various phenotypes in training cohort. Variables were normalized by mean and standard error.
Figure 4
Figure 4
Primary and secondary outcomes among three distinct immune phenotypes in training cohort. (A) Survival curves for various phenotype patients during their hospitalization. (B) Survival curves for various phenotype patients over 28 days. Blue line represents Phenotype A patients, red for Phenotype B patients, and green for Phenotype C patients. CAP patients in Phenotype A had a better prognosis than those in Phenotype A and C (P<0.05).Phenotype C CAP patients experience extended hospital stays (C) and ICU stays (F), prolonged ventilation days (D), and fewer ICU-free days (E) in comparison to patients with the other two phenotypes. Green represents Phenotype A patients, light blue for Phenotype B patients, and dark blue for Phenotype C patients. Patients with phenotype C comprise a greater proportion of patients requiring assisted ventilation (G) and those with severe pneumonia (H). Differences are observed in patient composition with respect to ventilation and the presence of severe pneumonia. P<0.001.
Figure 5
Figure 5
Heatmap dipicted C-index of various machine learning method in training and validation cohort for patients’ outcome (A) and pneumonia severity (B).
Figure 6
Figure 6
Robust performance of machine learning algorithm. (A) Time dependent bar and line graph of 9 machine learning methods at 7 days, 14days, and 21 days in training cohort. (B) Time dependent bar and line graph of 9 machine learning methods at 7 days, 14days, and 21 days in validation cohort. (C) The performance of SuperPC method and conventional PSI and CURB-65 evaluation criteria in training cohort. (D) Time dependent ROC curve of SuperPC method at 7 days, 14 days, 21 days in training cohort. (E) The performance of SuperPC method and conventional PSI and CURB-65 evaluation criteria in validation cohort. (F) Time dependent ROC curve of SuperPC method at 7 days, 14 days, 21 days in validation cohort. The performance of Random forest method and conventional PSI and CURB-65 evaluation criteria in training (G) and validation (H) cohort for predicting severe pneumonia. (I) Sankey plot illustrated the relationship between immune phenotypes and conventional pneumonia severity index (PSI) and CURB-65 evaluation criteria in Training cohort.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Phua J, Dean NC, Guo Q, Kuan WS, Lim HF, Lim TK. Severe community-acquired pneumonia: timely management measures in the first 24 hours. Crit Care (London England). (2016) 20:237. doi: 10.1186/s13054-016-1414-2 - DOI - PMC - PubMed
    1. Torres A, Chalmers J, Dela Cruz C, Dominedò C, Kollef M, Martin-Loeches I, et al. . Challenges in severe community-acquired pneumonia: a point-of-view review. Intensive Care Med. (2019) 45:159–71. doi: 10.1007/s00134-019-05519-y - DOI - PMC - PubMed
    1. Peyrani P, Arnold FW, Bordon J, Furmanek S, Luna CM, Cavallazzi R, et al. . Incidence and mortality of adults hospitalized with community-acquired pneumonia according to clinical course. Chest. (2020) 157:34–41. doi: 10.1016/j.chest.2019.09.022 - DOI - PubMed
    1. Zhang ZX, Yong Y, Tan WC, Shen L, Ng HS, Fong KY. Prognostic factors for mortality due to pneumonia among adults from different age groups in Singapore and mortality predictions based on PSI and CURB-65. Singapore Med J. (2018) 59:190–8. doi: 10.11622/smedj.2017079 - DOI - PMC - PubMed
    1. Parsonage M, Nathwani D, Davey P, Barlow G. Evaluation of the performance of CURB-65 with increasing age. Clin Microbiol infection. (2009) 15:858–64. doi: 10.1111/j.1469-0691.2009.02908.x - DOI - PubMed

Grants and funding

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. This work received funding from the Shanghai Science and Technology Commission (Grant No. 22Y11901700).

LinkOut - more resources