Abstract 624P
Background
This study aims to identify a BCAA gene signature and plasma metabolites to facilitate effective prognostic assessments in NSCLC patients.
Methods
Gene expression data and clinical information were retrieved from the GEO datasets. Utilizing a consistent clustering method, BCAA-related subtypes were identified. A risk model was constructed through univariate Cox regression and Lasso regression analyses. The prognostic efficacy of the model was validated across NSCLC, a melanoma, and pan-cancer ICI cohorts. The differences in immune cell infiltration were analyzed using the ESTIMATE and ssGSEA methods. Independent prognostic signatures were further validated through scRNA sequencing, spatial transcriptomics analysis, and multiple immunofluorescence in the CHCAMS cohort (n = 23). Additionally, plasma metabolites were detected via untargeted UPLC-MS/MS in the CHCAMS cohort (n = 133), comprising NSCLC patients receiving anti-PD1 therapy.
Results
Bioinformatics analysis classified NSCLC patients into four subtypes, demonstrating significant differences in prognosis and the tumor microenvironment. Univariate and Lasso Cox regression analyses identified five BCAA-related genes (ACAT2, ALDH2, HMGCS1, MLYCD, and PPM1K) constituting a prognostic signature for NSCLC patients. This signature proved effective as a prognostic predictor, validated through Kaplan-Meier and ROC curve analyses in ICI cohorts (p < 0.05). HMGCS1 exhibited independent prognostic value in ICI cohorts. Furthermore, HMGCS1 expression levels demonstrated a negative correlation with CD8+ T cell infiltration and BCAA degradation, while positively correlating with cholesterol homeostasis across GEO datasets, scRNA sequencing, ST, and mIF cohort analyses (p < 0.05). Plasma metabolite analysis revealed that patients with higher L-leucine and lower cholesterol levels exhibited superior prognosis (p < 0.05).
Conclusions
Our study identifies key BCAA-related genes and plasma metabolites as promising prognostic markers in NSCLC. These findings offer insights for personalized treatment strategies and highlight the potential of metabolic profiling in optimizing patient care.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.