MOJ
eISSN: 2374-6920
Submit manuscript...
Research Article Volume 7 Issue 6
An artificial neural network framework for biomarker development to predict distant metastasis of breast cancer from gene expression data
Yingying Ma, Xinping Cu
Received: January 01, 1971 | Published: December 28, 2018
Citation: Yingying M, Xinping C. An artificial neural network framework for biomarker development to predict distant metastasis of breast cancer from gene expression data. MOJ Proteomics Bioinformatics. 2018;7(6):192-198. DOI: 10.15406/mojpb.2018.07.00257
Abstract
Background: In spite of the low number of overlapping genes between the various published gene signatures for prediction of distant metastasis, the signatures had many pathways in common. Identification of the key biological processes, rather than the assessment of signatures based on individual genes, allows not only to build a biological meaningful and robust gene biomarker from functionally related genes, but also provides insight into the mechanism of the disease development.
Methods: In this work, we develop an artificial neural network framework for biomarker development to predict distant metastasis of breast cancer from gene expression data. Our intention for this structure is to potentially identify the underlying biological process in the design network. We follow the evaluation framework by McShane and Hayes from the National Cancer Institute to evaluate the proposed biomarker herein on three important aspects: analytical validity, clinical validity, and clinical utility.
Results: The Gene Recurrence Risk Score (GRRS) from the artificial neural network model developed herein has demonstrated significant association with the probability of distant metastasis within 10 years after adjustment of other standard clinical or pathological factors. On independent assessment TRANSBIG dataset, with distant metastasis within 10 years as clinical outcome, GRRS produced a sensitivity of 86% and a specificity of 37%, a 4% and 3% improvement over Veridex risk score respectively for ER positive group. GRRS also produced a sensitivity of 91% and a specificity of 32%, a 4% and 5% improvement over Veridex risk score respectively for ER negative group. It also showed on par classification accuracy with Veridex risk score for prediction of distant metastasis within 5 years for both ER subgroups.
Conclusion: The Gene Recurrence Risk Score (GRRS) has demonstrated the prognostic value and high classification accuracy for prediction of distant metastasis within 5 years and within 10 years. The artificial neural network framework for biomarker development for this specific task proves to be robust and effective. The model building itself also reveals the necessity to stratify patients by their ER status for classification.
Keywords: gene expression biomarker, distant metastasis of breast cancer prediction, precision oncology, artificial neural network, personalized medicine, biomarkers, herein, gene expression data, tumor bank, TRANSBIG.
©2018 Yingying, et al. This is an open access article distributed under the terms of the, which permits unrestricted use, distribution, and build upon your work non-commercially.