Machine learning-based neddylation landscape indicates different prognosis and immune microenvironment in endometrial cancer

Abstract

Endometrial cancer (EC) is women’s fourth most common malignant tumor. Neddylation plays a significant role in many diseases; however, the effect of neddylation and neddylation-related genes (NRGs) on EC is rarely reported. In this study, we first used MLN4924 to affect the activation of neddylation in different cell lines (Ishikawa and HEC-1-A) and determined the critical role of neddylation-related pathways for EC progression. Subsequently, we screened 17 prognostic NRGs based on expression files of the TCGA-UCEC cohort. Based on unsupervised consensus clustering analysis, patients with EC were classified into two neddylation patterns (C1 and C2). In terms of prognosis, substantial differences were observed between the two patterns. Compared with C2, C1 exhibited low levels of immune infiltration and promoted tumor progression. More importantly, based on the expression of 17 prognostic NRGs, we transformed nine machine-learning algorithms into 89 combinations. The random forest (RSF) was selected to construct the neddylation-related risk score according to the average C-index of different cohorts. Notably, our score had important clinical implications for EC. Patients with high scores have poor prognoses and a cold tumor state. In conclusion, neddylation-related patterns and scores can distinguish tumor microenvironment (TME) and prognosis and guide personalized treatment in patients with EC.

Integrative analysis from multi‐center studies identifies a weighted gene co‐expression network analysis‐based Tregs signature in ovarian cancer

“…To estimate immune cell abundance in different samples, we employed a range of methods including ssGSEA, TIMER, CIBERSORT, QUANTISEQ, MCP‐counter, XCELL, and EPIC. Moreover, the ESTIMATE algorithm was utilized to calculate the immune score and stromal score, serving as indicators of the overall microenvironmental status 29 . Moreover, we downloaded the IMvigor‐210 cohort from existing reference which had undergone immune checkpoint treatment to generate risk scores.…”

“…Each dataset was normalized separately, with the gene mean/SD being normalized to 1. Variable screening, which required a minimum threshold of four variables, was performed using Lasso, CoxBoost, random survival forest (RSF), StepCox [both], and StepCox [backward] in the TCGA‐OV cohort 29 . Subsequently, a combination of 10 machine learning algorithms (lasso, RSF, GBM, Survival‐SVM, SuperPC, ridge regression, plsRcox, CoxBoost, StepCox, and enet) were used to create a composite model, with RSF being identified as the best prognostic model, based on the average C‐index within all cohorts.…”

“…Moreover, the ESTIMATE algorithm was utilized to calculate the immune score and stromal score, serving as indicators of the overall microenvironmental status. 29 Moreover, we downloaded the IMvigor-210 cohort from existing reference which had undergone immune checkpoint treatment to generate risk scores. This enabled us to validate the efficacy of our risk scores in the immunotherapy cohort.…”

“…[backward] in the TCGA-OV cohort. 29 Subsequently, a combination of 10 machine learning algorithms (lasso, RSF, GBM, Survival-SVM, SuperPC, ridge regression, plsRcox, CoxBoost, StepCox, and enet) were used to create a composite model, with RSF being identified as the best prognostic model, based on the average C-index within all cohorts.…”

Transcriptome data‐based status of PI3K/AKT/mTOR pathway indicates heterogeneity and immune modulation in patients with pancreatic ductal adenocarcinoma

“…These approaches can enable the identification of specific molecular features associated with pathway activation and disease progression, as well as the development of predictive models for individual patient response to targeted therapies. 5 For instance, a recent study demonstrated the utility of machine learning in identifying gene expression patterns associated with PI3K/AKT/mTOR status in uveal melanoma, 6 highlighting the potential of these methods for identifying biomarkers of treatment response in PDAC as well. Moreover, the integration of machine learning with other high-throughput omics data, 7,8 such as proteomics and metabolomics, may provide a more comprehensive understanding of the molecular mechanisms underlying PDAC and the role of the PI3K/AKT/mTOR pathway in this disease.…”