Comprehensive characterization of long QT syndrome-associated genes in cancer and development of a robust prognosis model

Abstract

Cancer is the leading public health problem worldwide. However, the side effects accompanying anti-cancer therapies, particularly those pertaining to cardiotoxicity and adverse cardiac events, have been the hindrances to treatment progress. Long QT syndrome (LQTS) is one of the major clinic manifestations of the anti-cancer drug associated cardiac dysfunction. Therefore, elucidating the relationship between the LQTS and cancer is urgently needed. Transcriptomic sequencing data and clinic information of 10,531 patients diagnosed with 33 types of cancer was acquired from TCGA database. A pan-cancer applicative gene prognostic model was constructed based on the LQTS gene signatures. Meanwhile, transcriptome data and clinical information from various cancer types were collected from the GEO database to validate the robustness of the prognostic model. Furthermore, the expression level of transcriptomes and multiple clinical features were integrated to construct a Nomo chart to optimize the prognosis model. The ssGSEA analysis was employed to analysis the correlation between the LQTS gene signatures, clinic features and cancer associated signalling pathways. Our findings revealed that patients with lower LQTS gene signatures enrichment levels exhibit a poorer prognosis. The correlation of enrichment levels with the typical pathways was observed in multiple cancers. Then, based on the 17 LQTS gene signatures, we construct a prognostic model through the machine-learning approaches. The results obtained from the validation datasets and training datasets indicated that our prognostic model can effectively predict patient outcomes across diverse cancer types. Finally, we integrated this model with clinical features into a nomogram, demonstrating its potential as a valuable prognostic tool for cancer patients. Our study sheds light on the intricate relationship between LQTS and cancer pathways. A LQTS feature based clinic decision tool was developed aiming to enhance precision treatment of cancer.

Keywords: biomarker; cancer; long QT syndrome; prognosis.

FIGURE 1

The workflow of this study. Analysis of LQTS‐associated genes in cancer was performed using data in the TCGA and GEO databases.

FIGURE 2

The expression and prognostic value of LQTS‐associated genes in cancers. (A) The expression of long QT syndrome genes in cancer. (B) Bubble map of differential expression of 17 long LQTS‐associated genes between tumour and adjacent normal tissue. (C) The ROC curve and AUC value of long QT syndrome genes in cancer.

FIGURE 3

Correlation between the enrichment levels of LQTS gene and prognosis of cancer. (A) There were significant differences in the enrichment levels of LQTS‐associated genes in cancer. (B) Correlation between enrichment levels of LQTS‐associated genes and cancer stage. (C–E) Kaplan–Meier analysis of the association between enrichment levels of long QT syndrome genes and OS (C), DSS (D), PFI (E).

FIGURE 4

Correlation between the enrichment levels of LQTS‐associated genes and cancer associated signalling pathway. (A) Correlation map of 17 long QT syndrome genes. (B and C) Heat and radar maps demonstrate the correlations between long QT syndrome genes and cancer associated signalling pathway.

FIGURE 5

A LQTS gene‐based model for prognosis of cancer. (A and B) The coefficients calculated by multivariate Cox regression using LASSO‐COX regression. (C) Kaplan–Meier overall survival (OS) curves for pan‐cancer patients. (D) AUC curve of the prognostic model for pan‐cancer patients.

FIGURE 6

Survival analysis of prognostic models in multiple cancer types. Overall survival (OS) analysis of Prognostic model in TCGA dataset and other verification set by Kaplan–Meier curve. (A–J) TCGA cohorts, ACC (A), BRCA (B), KIRC (C), KIRP (D), MESO (E), LUAD (F), PAAD (G), LIHC (H), THYM(I), SKCM(J). (K–O) GEO cohorts, ACC (K), HNSC (L), BRCA (M), LUAD (N), DLCB (O). N: E‐MTAB‐1980 cohort, KIRC (P).

FIGURE 7

Differences in tumour‐related pathway activity in different groups of the TCGA cohort. The 83 pathway‐related functions are displayed in different colours at the bottom of the heat map. The biology process activity shows the activity scores after standardization of 83 pathways. Stage, Gender, OS, and Age are different clinical indicators of patients in the pan‐cancer cohort.

FIGURE 8

Correlation analysis between risk score and different clinical indicators. (A) Correlation of risk score with Age. (B) Correlation of risk score with OS time. (C) Correlation of risk score with PFI time. (D) Correlation of risk score with DFI time. (E) Correlation of risk score with DSS time. (F) Heat map of correlation between risk score and relevant clinical indicators. (G) Prognosis of LQTS genes and clinical features of correlation.

FIGURE 9

Establishment and prognostic value of nomogram based on prognostic model of LQTS‐associated genes. (A) Nomograph on the basis of LQTS score can predict 1, 3 and 5 year survivorship in pan‐cancer patients. (B) Nomograph calibration curves for predicting 1‐, 3‐, and 5‐year survivorship. (C) Kaplan–Meier curve showing a comparison of OS among low‐ and high‐risk groups. (D) AUC curve of the nomograph. (E) AUC values of nomograms and other clinical diagnostic features for predicting 1‐year survival times. (F) AUC values of nomograms and other clinical diagnostic features for predicting 3‐year survival times.