Prediction of early breast cancer patient survival using ensembles of hypoxia signatures

Abstract

BackgroundBiomarkers are a key component of precision medicine. However, full clinical integration of biomarkers has been met with challenges, partly attributed to analytical difficulties. It has been shown that biomarker reproducibility is susceptible to data preprocessing approaches. Here, we systematically evaluated machine-learning ensembles of preprocessing methods as a general strategy to improve biomarker performance for prediction of survival from early breast cancer.ResultsWe risk stratified breast cancer patients into either low-risk or high-risk groups based on four published hypoxia signatures (Buffa, Winter, Hu, and Sorensen), using 24 different preprocessing approaches for microarray normalization. The 24 binary risk profiles determined for each hypoxia signature were combined using a random forest to evaluate the efficacy of a preprocessing ensemble classifier. We demonstrate that the best way of merging preprocessing methods varies from signature to signature, and that there is likely no ‘best’ preprocessing pipeline that is universal across datasets, highlighting the need to evaluate ensembles of preprocessing algorithms. Further, we developed novel signatures for each preprocessing method and the risk classifications from each were incorporated in a meta-random forest model. Interestingly, the classification of these biomarkers and its ensemble show striking consistency, demonstrating that similar intrinsic biological information are being faithfully represented. As such, these classification patterns further confirm that there is a subset of patients whose prognosis is consistently challenging to predict.ConclusionsPerformance of different prognostic signatures varies with pre-processing method. A simple classifier by unanimous voting of classifications is a reliable way of improving on single preprocessing methods. Future signatures will likely require integration of intrinsic and extrinsic clinico-pathological variables to better predict disease-related outcomes.

Hypoxia-Associated Prognostic Markers and Competing Endogenous RNA Co-Expression Networks in Breast Cancer

“…Inna Y. Gong et al. explored several datasets from GEO database based on four published hypoxia signatures [Buffa ( 12 ), Winter ( 44 ), Hu ( 45 ), and Sorensen ( 46 )], and confirmed to a certain extent that hypoxia-related gene signatures had potential to be used as biomarkers to predict survival of early breast cancer ( 47 ). By contrast, Maud H W Starmans et al.…”

“…Enriched studies indicate that hypoxia-related gene signatures generated in vitro and in vivo have prognostic power in breast cancer and other cancers. Inna Y. Gong et al explored several datasets from GEO database based on four published hypoxia signatures [Buffa (12), Winter (44), Hu (45), and Sorensen (46)], and confirmed to a certain extent that hypoxia-related gene signatures had potential to be used as biomarkers to predict survival of early breast cancer (47). By contrast, Maud H W Starmans et al identified 295 up-regulated and 164 downregulated genes under hypoxia in breast (MCF7), colon (HT29) and prostate (DU145) carcinoma cells in vitro, but they found that none of these in vitro derived signatures consisting of hypoxia-induced genes are prognostic when in a much larger cohort of breast cancer patients in vivo (48).…”

Identification of a gene expression signature associated with breast cancer survival and risk that improves clinical genomic platforms

“…An ensemble of linear classifiers is trained for feature selection and diversity among them is induced by bootstrap resampling. Several authors have applied ensemble methods in feature selection to improve the robustness of the gene markers found ( Abeel et al , 2010 ; Gong et al , 2018 ; Zhao et al , 2011 ). The ensemble of lists is combined by selecting the genes associated to the IHC variables with highest stability.…”

“…Within this problem, some integrative approaches attempted to combine protein interactome network information with gene expression data to more accurately predict prognosis ( Das et al , 2015 ). Others are based on ensemble methods ( Abeel et al , 2010 ; Gong et al , 2018 ), that try to reduce the variance of the estimator considering different feature selection algorithms or different preprocessing methods.…”

Syndecan-1 Promotes Angiogenesis in Triple-Negative Breast Cancer through the Prognostically Relevant Tissue Factor Pathway and Additional Angiogenic Routes

“…For survival analysis, in the R statistical environment, we utilized the Kaplan–Meier Plotter database via the statistical package “survival” to calculate Kaplan–Meier survival curves and the number-at-risk. Furthermore, the hazard ratio (and 95% confidence intervals) and log-rank p were calculated for each gene [ 27 ]. The statistical analysis was considered significant when p < 0.05.…”