A novel prognostic model for lung squamous cell carcinoma based on multi-omics analysis and machine learning

Lung squamous-cell carcinoma (LUSC) is a highly aggressive malignancy with a poor prognosis. Tertiary lymphoid structures (TLS) play a crucial role in the immune response and significantly influence the efficacy of immunotherapy. However, the prognostic and immunological implications of TLS-associated molecular subtypes in LUSC remain unclear. In this study, we applied 10 multi-omics integration strategies to perform a multi-omics analysis of the mRNA expression profiles, DNA methylation, and genomic mutation data of 39 TLSs-related genes, along with long non-coding RNA (lncRNA) expression profiles, to generate integrated consensus subtypes of LUSC. Four molecular subtypes were identified: cancer subtype 1 (CS1), CS2, CS3, and CS4. We observed a significant difference in overall survival between cancer subtype 1 (CS1) and CS3. Subsequently, we identified 33 prognosis-related genes based on differential expression between CS1 and CS3, which were further refined to 20 genes using the least absolute shrinkage and selection operator (LASSO) regression algorithm, and constructed a prognostic signature termed the LUSC-Survival Prediction Index (LUSCSPI). The high-LUSCSPI group demonstrated a poor prognosis and was more likely to benefit from treatment with nine chemotherapeutic agents (shikonin, doxorubicin, CMK, S-Trityl-L-cysteine, paclitaxel, DMOG, gemcitabine, erlotinib, and crizotinib). In contrast, the low-LUSCSPI group exhibited a more favorable prognosis, with thapsigargin and cisplatin identified as promising treatment options. In conclusion, our results highlight the potential of LUSCSPI as an independent prognostic factor for LUSC. Further, the multi-omics consensus approach provides a robust foundation for prognostic stratification in LUSC patients, facilitating personalized treatment and disease management.