A Novel Fuzzy Classifier Model for Cancer Classification Using Gene Expression Data
Open Access
- 17 October 2023
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 115161-115178
Abstract
In the pursuit of better cancer classification, many studies have been conducted to identify the genes associated with cancer. However, the high dimensionality of gene expression data and the limited relevance of a few genes pose significant challenges to this endeavor. Existing gene selection methods yield divergent gene lists, further complicating the classification process. To overcome this issue, we developed a novel approach called Fuzzy Gene Selection (FGS), which combines the strengths of various gene selection methods in the field. FGS was developed using three feature selection techniques (Mutual Information, F-ClassIf, and Chi-squared) to rank genes based on their importance. These methods generated scores and rankings for each gene. Fuzzification and Defuzzification techniques were then applied to combine these scores into a single best score for each gene. This approach aids in identifying genes of significance in cancer classification, especially in multi-class scenarios. Classifiers often produce convergent decisions in such cases, where the predicted probabilities for different classes do not always correspond to the correct predicted class with the highest probability. To address this, we developed a novel Fuzzy classifier that leverages the contributions from each node’s traditional deep classifier. This novel approach combines the strengths of traditional deep classifiers at individual nodes, enabling the Fuzzy classifier to make more robust and informed predictions. The Fuzzy classifier (FC) has demonstrated improvements in accuracy and the generalization of the proposed algorithm to accurately classify different cancer types.Keywords
Funding Information
- Ministry of Higher Education and Scientific Research in the Republic of Iraq, in 2017 (22223)
This publication has 23 references indexed in Scilit:
- A deep auto-encoder model for gene expression predictionBMC Genomics, 2017
- Survey on recent cancer classification systems for cancer diagnosis2017
- Feature Selection and Cancer Classification via Sparse Logistic Regression with the Hybrid L1/2 +2 RegularizationPLOS ONE, 2016
- Using epigenomics data to predict gene expression in lung cancerBMC Bioinformatics, 2015
- Identification of Thyroid Carcinoma Related Genes with mRMR and Shortest Path ApproachesPLOS ONE, 2014
- The Cancer Genome Atlas Pan-Cancer analysis projectNature Genetics, 2013
- NCBI GEO: archive for functional genomics data sets—updateNucleic Acids Research, 2012
- Introduction to Microarray Data Analysis2005
- CT scanning: A major source of radiation exposureSeminars in Ultrasound, CT and MRI, 2002
- Gene Expression Omnibus: NCBI gene expression and hybridization array data repositoryNucleic Acids Research, 2002