A comparison of survival analysis methods for cancer gene expression RNA-Sequencing data

Abstract

Identifying genetic biomarkers of patient survival remains a major goal of large-scale cancer profiling studies. Using gene expression data to predict the outcome of a patient"s tumor makes biomarker discovery a compelling tool for improving patient care. As genomic technologies expand, multiple data types may serve as informative biomarkers, and bioinformatic strategies have evolved around these different applications. For categorical variables such as a gene"s mutation status, biomarker identification to predict survival time is straightforward. However, for continuous variables like gene expression, the available methods generate highly-variable results, and studies on best practices are lacking. We investigated the performance of eight methods that deal specifically with continuous data. K-means, Cox regression, concordance index, D-index, 25 th -75 th percentile split, median-split, distribution-based splitting, and KaplanScan were applied to four RNA-sequencing (RNA-seq) datasets from the Cancer Genome Atlas. The reliability of the eight methods was assessed by splitting each dataset into two groups and comparing the overlap of results. Gene sets that had been identified from the literature for a specific tumor type served as positive controls to assess the accuracy of each biomarker using receiver operating characteristic (ROC) curves. Artificial RNA-Seq data were generated to test the robustness of these methods under fixed levels of gene expression noise. Our results show that methods based on dichotomizing tend to have consistently poor performance while C-index, D-index and k-means perform well in most settings. Overall, the Cox regression method had the strongest performance based on tests of accuracy, reliability, and robustness.

The shape of gene expression distributions matter: how incorporating distribution shape improves the interpretation of cancer transcriptomic data

“…It is worthwhile recognizing that the use of thresholds to evaluate survival analysis differences is inherently arbitrary. A previous study took a broader view into investigating how threshold-based cut-offs versus more robust, less arbitrary approaches such as the concordance index (C-index), D-index, and K-means performed for survival analysis models based on gene expression data sets from TCGA [23]. Raman et al also contrasted performance of these metrics against the distribution-based ones that were included in this study.…”

Influence of tumor-infiltrating immune cells on local control rate, distant metastasis, and survival in patients with soft tissue sarcoma

“…Gene expression units used were log2(fpkm+1) and treated as continuous variables, as previously suggested. 25,26 CD19, MS4A1 (Membrane Spanning 4-Domains A1; CD20), CD22 and CD79A were used as genes characteristic for B-cells, while CD68,ITGAM (Integrin Subunit Alpha M; CD11B) and CD163 were chosen as genes characteristic for macrophages. CD3D, CD3E, CD52 and CD6 were selected as genes characteristic for T-cells.…”

Exploring Additional Valuable Information From Single-Cell RNA-Seq Data

“…Furthermore, Li H. P. et al (2017) identified single-cell biomarkers that can stratify the colorectal tumors from TCGA and GEO databases into subgroups with divergent survival. For survival analysis, Raman et al (2019) revealed that highly variable results are usually obtained from different methods, and Cox regression (Li, 2003) is superior to other compared approaches based on tests of reliability, accuracy, and robustness. Cox regression is a flexible method that can improve the accuracy of estimation between gene expression level and patient survival by enabling the inclusion of multiple covariates to accommodate explanatory variables.…”