Gene Expression Analysis for Early Lung Cancer Prediction Using Machine Learning Techniques: An Eco-Genomics Approach

Abstract

Cancer as a multifactorial disorder develops due to the complex interaction between gene and environment. A person may be susceptible to cancer due to his individual genetic makeup. Cancer causes maximum death worldwide as per data given by the World Health Organization. Some cases are reported with particular genetic makeup. Hence, the proper understanding of eco-genomics of cancer is necessary to interpret the underlying cause and risk factor for cancer. Combining huge gene expression date available in cyberspace and advanced machine learning technique may bring out the strongest candidate genes, which individually or as a part of the complex system, have more accurate prognostic value to determine someone's susceptibility toward cancer. In this paper, we have analyzed gene expression data for the lung cancer available in the Kent Ridge Bio-Medical Dataset Repository. The microarray gene expression data are analyzed to select and predict the optimal subset of genes, which are the most probable causing agent of lung cancer. INDEX TERMS Gene expression analysis, machine learning technique, eco-genomics, multilayer perceptron, information gain attribute ranking.

A Survey of Machine Learning Approaches Applied to Gene Expression Analysis for Cancer Prediction

“…In a study of microarray gene expression data from Lung adenocarcinoma samples (86 tumor samples and 10 non-tumor samples collected from Kent Ridge Bio-Medical Dataset Repository available from [60] with 7129 genes) an Info gain feature selection technique was applied to identify genes strongly associated with cancer samples using 70% of samples as a training set and 30% of samples as a test set. This study applied three classifier techniques to discriminate tumor and non-tumor samples after choosing the candidate genes that had known relevance to lung cancer [2]. Several selected genes were evaluated for biological relevance in lung cancer pathology.…”

“…Abnormal growth of cells occurs because of the complex interaction between genes (deregulated due to mutation and epigenetic modifications) and the environment (i.e. carcinogens) [2]. Consequently, whole-genome expression analysis has become an important tool to identify relevant genes pathways that are deregulated and drive abnormal cellular proliferation and metastatic spread.…”

“…Whole-genome expression (transcriptomic) analysis has the potential to provide early cancer prediction, for diagnosis, determining clinical outcomes, and the potential for disseminated disease. In addition, molecular cancer classification can enhance the success of personalized treatments including immune-checkpoint inhibitors including anti-PD1 and anti-CTLA-4 [2]. Measuring gene expression differences of thousands of genes between healthy and unhealthy tissue using transcriptomic approaches such as microarrays and more recently RNA-Seq has required investigators to develop bioinformatic pipelines that include mathematical and statistical methods to analyze these large novel datasets.…”

Research Proposal: Lung Cancer Prediction and Classification using Machine learning Models

“…Almost 10% to 15% of all lung cancers in the lungs are SCLC and it is once in a while called oat cell carcinoma [3]. This kind of lung cancer in the lungs will in general develop and spread quicker than NSCLC.…”

Meta-Learning Based Recommendation of Ensemble Methods for Gene Expression Classification

“…New opportunities for a better understanding of treated tissues was brought up by gene expression profiling tools like microarrays [39]. Microarrays are methods based on hybridization that allow a global view of cells and their gene expression levels [33], thus allowing the measurement of the proteins being produced. This technology has helped researchers to achieve better results on understanding healthy and unhealthy states.…”