Machine learning pinpoints gene-expression biomarkers of lung transplant rejection

Researchers have identified a set of candidate gene-expression biomarkers for lung allograft rejection by combining transcriptomic analysis with machine learning, according to a Research Note published in BMC Research Notes on 28 October 2026. The team, led by Masoume Avateffazeli and Elham Nazari, set out to find reproducible molecular markers that could complement the current diagnostic gold standard, histopathological assessment of transbronchial biopsies, which is limited by sampling error and inter-observer variability.

The analysis drew on publicly available microarray data from 457 transbronchial biopsies (dataset GSE150156) to identify differentially expressed genes between rejection or injury-associated biopsies and non-rejection biopsies. The researchers detected 1,412 differentially expressed genes, and then performed functional enrichment and protein-protein interaction analyses to characterize the biological roles of the genes involved.

To narrow the field to the most promising markers, the team applied two machine-learning algorithms, L1-regularized logistic regression and random forest. This prioritization step produced 18 candidate biomarkers whose area under the receiver operating characteristic curve exceeded 90 percent in the discovery dataset.

External validation followed in an independent dataset of 190 third-bronchial-bifurcation mucosal biopsies (GSE125004). In this external evaluation, 13 genes were significantly differentially expressed, and 11 of those showed concordant directions of expression across both datasets, a consistency the authors regard as evidence of reproducibility.

Among the validated candidates, CYBB and LCP2 emerged as particularly notable because they had not previously been evaluated as individual biomarkers in lung transplant rejection. The authors state that the integrative transcriptomic and machine-learning approach identified reproducible candidate biomarkers for the molecular assessment of lung allograft rejection, and they note that prospective studies are warranted to validate the findings and support the development of targeted diagnostic assays.

Subject of Research: Identification of candidate gene-expression biomarkers for lung allograft rejection using transcriptomic analysis and machine learning.

Article Title: Gene-expression biomarkers of lung allograft rejection identified through transcriptomic analysis and machine learning

Article References: Avateffazeli, M., Khayami, R., Abedini, A., Mohamadnia, A., Soleimani, S., Judaki, A. A., Moradkhani, A., Shafaghi, S., & Nazari, E. (2026). Gene-expression biomarkers of lung allograft rejection identified through transcriptomic analysis and machine learning. BMC Research Notes. https://doi.org/10.1186/s13104-026-08093-x

Image Credits: AI Generated

DOI: 10.1186/s13104-026-08093-x

Keywords: lung transplantation, allograft rejection, gene expression, biomarkers, machine learning, transcriptomics, microarray, transbronchial biopsy, CYBB, LCP2, diagnostic assays, BMC Research Notes