Feature Selection and Molecular Classification of Cancer Phenotypes: A Comparative Study

Abstract

The classification of high dimensional gene expression data is key to the development of effective diagnostic and prognostic tools. Feature selection involves finding the best subset with the highest power in predicting class labels. Here, we conducted a comparative study focused on different combinations of feature selectors (Chi-Squared, mRMR, Relief-F, and Genetic Algorithms) and classification learning algorithms (Random Forests, PLS-DA, SVM, Regularized Logistic/Multinomial Regression, and kNN) to identify those with the best predictive capacity. The performance of each combination is evaluated through an empirical study on three benchmark cancer-related microarray datasets. Our results first suggest that the quality of the data relevant to the target classes is key for the successful classification of cancer phenotypes. We also proved that, for a given classification learning algorithm and dataset, all filters have a similar performance. Interestingly, filters achieve comparable or even better results with respect to the GA-based wrappers, while also being easier and faster to implement. Taken together, our findings suggest that simple, well-established feature selectors in combination with optimized classifiers guarantee good performances, with no need for complicated and computationally demanding methodologies.

GradWise: A Novel Application of a Rank-Based Weighted Hybrid Filter and Embedded Feature Selection Method for Glioma Grading with Clinical and Molecular Characteristics

“…Feature selection is generally utilized for data analysis, pattern recognition, data mining, and machine learning tasks. This process aims to improve performance (e.g., tumor grading) and classification accuracy rate and provide computational efficiency by removing irrelevant or redundant features and reducing the dimensionality of data [ 12 , 13 , 14 , 15 , 16 , 17 ]. There are various feature selection methods available, such as filter methods, wrapper methods, and embedded methods [ 12 , 18 ], each with its own advantages and limitations.…”

From cancer big data to treatment: Artificial intelligence in cancer research

“…Li et al 54 suggested that developing algorithms that can effectively analyze multiple‐class expression data is crucial because increasing the sample size may not be a sufficient solution to this problem. In another example, Zanella et al 55 performed a comparative study to find the optimal combinations of feature selection techniques and ML algorithms for successful classification of cancer phenotypes using microarray datasets. They evaluated the performance of different feature selectors and classifiers on three benchmark cancer microarray datasets.…”

“…55 performed a comparative study to find the optimal combinations of feature selection techniques and ML algorithms for successful classification of cancer phenotypes using microarray datasets. They evaluated the performance of different feature selectors and classifiers on three benchmark cancer microarray datasets.…”

Machine learning for pan-cancer classification based on RNA sequencing data

“…Filter methods, while computationally less demanding, typically do not consider the subsequent classification model, often resulting in inferior performance compared to wrappers. In contrast, wrappers can be susceptible to overfitting and sensitive to parameter adjustments (Zanella et al, 2022). Recently, various nature-inspired algorithms, such as those based on swarm intelligence and evolutionary principles, have been applied as metaheuristic search methods for wrapper-based feature selection problems and show significant potential in the identification of relevant genes for cancer classification using microarray gene expression measurements (Pham and Raahemi, 2023;Yaqoob et al, 2023).…”