Correlation-centred variable selection of a gene expression signature to predict breast cancer metastasis

Abstract

predictions of distant cancer metastasis based on gene signatures are studied intensively to realise precise diagnosis and treatments. Gene selection i.e. feature selection is a cornerstone to both establish accurate predictions and understand underlying pathologies. Here, we developed a simple but robust feature selection method using a correlation-centred approach to select minimal gene sets that have both high predictive and generalisation abilities. A multiple logistic regression model was used to predict 5-year metastases of patients with breast cancer. Gene expression data obtained from tumour samples of lymph node-negative breast cancer patients were randomly split into training and validation data. Our method selected 12 genes using training data and this showed a higher area under the receiver operating characteristic curve of 0.730 compared with 0.579 yielded by previously reported 76 genes. The signature with the predictive model was validated in an independent dataset, and its higher generalization ability was observed. Gene ontology analyses revealed that our method consistently selected genes with identical functions which frequently selected by the 76 genes. Taken together, our method identifies fewer gene sets bearing high predictive abilities, which would be versatile and applicable to predict other factors such as the outcomes of medical treatments and prognoses of other cancer types. Breast cancer is the most common cancer in women worldwide 1. It is a heterogeneous disease that causes difficulties for its diagnosis and treatment. Thus, multiple clinicopathological factors, such as the tumour size and lymph node metastasis, and established biomarkers, such as the estrogen receptor (ER), human epidermal growth factor receptor 2 (HER2), and progesterone receptor, have been used to characterise specific features of breast cancer, including its subtypes. To realise more accurate characterisation of breast cancer, gene expression has been introduced to understand molecular-level pathology and prediction of treatment outcomes of breast cancer, which would contribute to the design of treatments 2,3. Currently, MammaPrint 4 , Oncotype DX 5 , and MapQuant Dx 6 are commercially available diagnostic tests to accurately estimate the recurrence of breast cancer. These tests employ relatively large numbers of genes, that is, 70, 21, and 97, respectively. Gene expression is also used for metastasis prediction and subtype identification of breast cancers 7-9. The use of gene expression in combination with known biomarkers and clinicopathological factors is considered as a critical issue for precision medicine. There are three processes in the establishment of a prediction model using microarray data: (1) feature selection, (2) development of a mathematical model, and (3) validation of the developed model. The purpose of feature selection is to identify a minimal predictive gene set by removing redundant and irrelevant features 10-12. This process is important because the subsequent analyses depend on the selec...

Quantifying the contribution of transcription factor activity, mutations and microRNAs to CD274 expression in cancer patients

“…TFs with non-zero values and a significant ( p < 0.001) effect on PD-L1 expression were included in the final MLR model. As a side note, although we utilized only fully independent target genes in our estimates of TF activity, potential multicollinearity of the involved genes might still weaken the model’s generalization ability 36 . However, multicollinearity analysis by calculation of the Variation Inflation Factor (VIF) values of the model’s independent variables resulted in VIFs lower than 4, indicating that the selected descriptors were not highly correlated (Supplementary file 1 , Table S1 ).…”

The combined impact of IFRS mandatory adoption and institutional quality on the IPO companies’ underpricing

“…VIF is an abbreviation for the variance inflation factor test, which is used to determine the existence of a multicollinearity issue in the OLS model. If VIF exceeds a threshold value of 10 then it results in misleading T -statistic values (Hikichi et al , 2020);*** p < 0.01,** p < 0.05,* p < 0.1 are considered significant when the T -statistics in brackets are corrected for heteroscedasticity. Table 2 provides definitions for all variables…”

“…VIF is an abbreviation for the variance inflation factor test, which is used to determine the existence of a multicollinearity issue in the OLS model. If VIF exceeds a threshold value of 10 then it results in misleading T-statistic values(Hikichi et al, 2020); *** p < 0.01, ** p < 0.05, * p < 0.1 are considered significant when the T-statistics in brackets are corrected for heteroscedasticity.Table 2 provides definitions for all variables IFRS mandatory adoption financial reports and establishing defined measurement and recognition requirements. Adopting IFRS increases the accuracy of financial reporting in IPO prospectuses and reduces investor ex ante uncertainty regarding the proper valuation of the IPO offering price, minimizing information asymmetry and IPO underpricing.…”

Advancement of prognostic models in breast cancer: a narrative review

“…To solve the multicollinearity of genes and improve the generalization ability of the model, Hikichi et al (91) applied the same dataset as which Wang et al (90) used to build 76-Gene signature to develop a simple but robust feature selection method using a correlation-centered approach. They obtained a 12-gene set with both high predictive and generalization abilities and constructed a prediction model for 5-year DM in patients with early-stage breast cancer, of which the prediction efficiency was similar to that of Rotterdam 76-gene assay.…”