Title: Identifying breast cancer molecular class using integrated feature selection and deep learning model

Authors: Monika Lamba; Geetika Munjal; Yogita Gigras

Addresses: The Northcap University, Gurugram, India ' Amity University, Uttar Pradesh, India ' The Northcap University, Gurugram, India

Abstract: The extraction of molecular subcategory is one such valuable evidence concerning breast cancer in determining its cure and prognosis. This manuscript has framed a model for molecular subtype-based feature selection known as CFS-BFS followed by classification using deep learning. The proposed model captures significant genes by utilising pre-processing ladder along with the combination of filter and wrapper-based technique CFS-BFS. The obtained genes are assessed via numerous machine learning methodologies where it is remarked that carefully chosen significant genes are more profitable in explaining this molecular problem using deep learning. The study has attained the maximum precision and beats brilliantly in terms of recall, F-score, TP_Rate, fallout, and MCC. Hence, proposed paradigm is recognised as one of the best effective technique determining the outstanding recital with all the chosen micro-array gene expression datasets for significant obtained genes. The genes identified by integrated model are also validated using Kaplan-Meier survival graph to show their credibility in breast cancer prognosis. Survival analysis show that selected genes using integrated approach can separate luminal, non-luminal subcategory utilising various factors including age, disease free survival, and relapse free survival.

Keywords: feature selection; deep learning; breast cancer; molecular subtype; SMOTE; best first search; CFS-BFS.

DOI: 10.1504/IJBRA.2023.131278

International Journal of Bioinformatics Research and Applications, 2023 Vol.19 No.1, pp.19 - 42

Received: 17 Jun 2022
Accepted: 04 Jan 2023
Published online: 05 Jun 2023 *