Stratification of Breast Cancer by Integrating Gene Expression Data and Clinical Variables

Abstract

Breast cancer is a heterogeneous disease. Although gene expression profiling has led to the definition of several subtypes of breast cancer, the precise discovery of the subtypes remains a challenge. Clinical data is another promising source. In this study, clinical variables are utilized and integrated to gene expressions for the stratification of breast cancer. We adopt two phases: gene selection and clustering, where the integration is in the gene selection phase; only genes whose expressions are most relevant to each clinical variable and least redundant among themselves are selected for further clustering. In practice, we simply utilize maximum relevance minimum redundancy (mRMR) for gene selection and k-means for clustering. We compare the results of our method with those of two commonly used only expression-based breast cancer stratification methods: prediction analysis of microarray 50 (PAM50) and highest variability (HV). The result is that our method outperforms them in identifying subtypes significantly associated with five-year survival and recurrence time. Specifically, our method identified recurrence-associated breast cancer subtypes that were not identified by PAM50 and HV. Additionally, our analysis discovered three survival-associated luminal-A subgroups and two survival-associated luminal-B subgroups. The study indicates that screening clinically relevant gene expressions yields improved breast cancer stratification.

Modeling Basins of Attraction for Breast Cancer Using Hopfield Networks

“…Stratification of patients may improve treatment outcomes through the identification of molecular targets common to a group of patients (He et al, 2019). For instance, trastuzumab is a monoclonal antibody, used as adjuvant treatment against breast and stomach cancers that overexpress the HER2 protein (Wang et al, 2019).…”

Integrating Somatic Mutations for Breast Cancer Survival Prediction Using Machine Learning Methods

“…We followed the protocol from our previously published studies ( He et al, 2017 , 2019 ), and we first removed the genes with missing values in more than 10% of samples for gene expression, CNV, gene methylation, protein expression, and somatic mutations. After that, flat variables that had the same values in more than 80% of the samples (non-informative) were discarded except in the case of somatic mutations ( Yuan et al, 2014 ; He et al, 2019 ). According to the previous study ( He et al, 2019 ), the RNA-Seq gene expression level 3 transcription was log2 transformed and RSEM-normalized ( Li and Dewey, 2011 ).…”

“…After that, flat variables that had the same values in more than 80% of the samples (non-informative) were discarded except in the case of somatic mutations ( Yuan et al, 2014 ; He et al, 2019 ). According to the previous study ( He et al, 2019 ), the RNA-Seq gene expression level 3 transcription was log2 transformed and RSEM-normalized ( Li and Dewey, 2011 ). Regarding the CNV features, we directly utilized the gene-level copy number values that were estimated using the GISTIC2 method ( Mermel et al, 2011 ; Yuan et al, 2017 ; Yuan et al, 2019 , 2020a , b ).…”

“…However, this large number of features may cause poor performance due to dimensionality and high redundancy ( Jain and Zongker, 1997 ; Jie et al, 2015 ). Therefore, according to our previous study ( He et al, 2019 ), mRMR was adopted in the present study to select the most useful features for the prognostic model.…”

cancerAlign: Stratifying tumors by unsupervised alignment across cancer types

“…Existing computational approaches mainly exploited two machine learning techniques to advance tumor stratification: dimensionality reduction [22][23][24][25][26][27] and network-based aggregation of individual mutations [1,[17][18][19]28,29]. For example, NBS used molecular networks to aggregate individual gene mutation into higher level functions and structures in cancer cells [1].…”