Academic journal article Iranian Journal of Public Health

Improving the Prediction of Survival in Cancer Patients by Using Machine Learning Techniques: Experience of Gene Expression Data: A Narrative Review

Academic journal article Iranian Journal of Public Health

Improving the Prediction of Survival in Cancer Patients by Using Machine Learning Techniques: Experience of Gene Expression Data: A Narrative Review

Article excerpt

Introduction

Today, cancer is one of the major health problems worldwide (1-3). Cancer research in the 21st century has become one of the most common efforts. The International Agency for Research on Cancer, based on 2002 dataset, estimated that the numbers of cancer patients are 25 million and the American Cancer Society in 2004, announced that officially cancer would replace heart disease as the main cause of death (4-6). Despite the many advances in early detection of diseases, cancer patients have a poor prognosis and survival rate for such patients are low (7-9). Correct perception of the biologic behavior of the tumor and its analysis, help to correct choice of treatment and has a potential to improve the consequences of cancer as well. Accurate estimation of prognosis and survival duration is the most important part of a process of clinical decision-making in patients with malignant disease (9). The first step to making sure that cancer patients have received proper care is to improve the ability of physicians to formulate this type of estimation (10).

The prediction of prognosis includes the vast range of decisions about different aspects of cancer treatment (10). The gene expression profiles obtain from different tissue types (11). By comparing the genes expressed in normal tissue and diseased tissue can bring better insight and understanding of the cancer pathology and help to physicians in decision-making. Checking gene expression patterns for attributes associated with the clinical behavior are very important, because these patterns, examine prognosis and leading to the alternative approach to understand the molecular and physiological mechanisms (10-12).

Gene expression pattern analysis offers ways to improve the diagnosis and classification of risk for many cancers (11, 12). Studies have marked the power of analytical methods than histological and clinical criteria in survival prediction. Recently in artificial intelligence domain, developing clinical decision support systems based on machine-learning methods to analyze gene expression data has facilitated and improved the medical prognosis. Studies have shown the higher accuracy of machine learning algorithms than regression models in predicting cancer survival (12, 13). Gene expression data have the potential to prevent errors caused by fatigue and impatience of oncology experts in survival estimation. Analyzing such a data by using machine-learning techniques leads to developing clinical decision support systems for the correct estimation of survival time and so provides proper treatments to patients according to their survival. This achievement can prevent unnecessary surgical and treatment procedures that increase the use of human resources and time that impose unnecessary costs on patients and the health care system (8, 14, 15).

This study has highlighted the advantages of machine learning techniques in survival prediction of cancer patients based on gene expression data.

Methods

This review article was conducted by searching articles between 2000 to 2016 in scientific databases (SCOPUS & Pub Med & Google Scholar & IEEE) and e-Journal (science direct), and by using keywords such as machine learning algorithms, gene expression data, survival and cancer. Non-English and unavailable full texts and also the studies that not defined as a journal article were excluded from this study.

Results

Microarray Technology and Gene Expression Data

Several genes and proteins with abnormal function and expression contribute to the cancer development and its pathogenesis (9, 16, 17). Gene expression measures the level of gene activity in a tissue and thus gives information about the complex activities of cells. This data usually obtain by measuring the activation and function of genes during their translation. Since cancers are associated with genetic abnormalities, gene expression data can display these abnormalities. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.