#AI reads Urine#Host and bacterial urine proteomics might predict treatment outcomes for immunotherapy in advanced non-small cell lung cancer patients
Published 09 May, 2025
This research focuses on exploring the potential of host and bacterial urine proteomics in predicting the treatment outcomes of immunotherapy for advanced non - small cell lung cancer (NSCLC) patients. Urinary tests are cost - effective and non - invasive. Urine can not only detect bladder and genito - urinary cancers but also remote malignancies.
The researchers enrolled 33 advanced - stage NSCLC patients who received different immunotherapies, such as nivolumab, pembrolizumab, atezolizumab, or durvalumab - based chemo - immunotherapy. Baseline urine samples were collected before or within one week after the first cycle of immunotherapy, and follow - up urine samples were obtained from 17 patients 120 (±7 days) after the first cycle. Baseline stool samples from 23 patients were used for microbiome genome analysis.
The results showed that a total of 6183 proteins were detected in urine extracellular vesicles (EVs) samples, including 3513 human, 2647 bacterial, 19 fungal, and 4 viral proteins. 186 human proteins showed differential abundance according to progression-free survival (PFS) groups. Proteins associated with long PFS were related to general immune function and innate immunity, while those associated with short PFS were involved in specific pathways like the Endosomal/Vacuolar pathway, Complement cascade, etc.
Among the top differentially abundant human proteins, MPP5, IGKV6 - 21, NT5E, and KRT27 were positive predictors of long PFS, and LMAN2, NUTF2, NID1, TNC, IGF1, BCR, GPHN, and PPBP were negative predictors. For bacterial proteins, 96 showed differential abundance. The abundance of bacterial proteins in urine was higher in patients with long PFS. The gut metagenomic abundance of Escherichia coli and Enterococcus faecalis was positively correlated with their urine EV protein abundance, suggesting that these bacterial proteins might originate from the gut microbiome.
The Random Forest machine learning model showed that the top 20 human urine proteins had an outstanding performance in predicting short vs long PFS, with an AUC of 0.89 and an accuracy of 95%.
However, the study had limitations, such as a relatively small sample size, inability to prove causality, and the absence of a healthy control group. Future studies with larger sample sizes, independent validation cohorts, and experimental validation are needed to confirm these findings.
Front Immunol. 2025 Apr 14:16:1543817. doi: 10.3389/fimmu.2025.1543817. David Dora, Peter Revisnyei, Alija Pasic, Gabriella Galffy, Edit Dulka, Anna Mihucz, Brigitta Rosko, Sara Szincsak, Anton Iliuk, Glen J. Weiss and Zoltan Lohinai
Youhe Gao
Statement: During the preparation of this work the author(s) used Doubao / AI reading for summarizing the content. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.
For earlier AI Reads Urine articles：