Citation

Hamim, Sultanul Arifeen and Sadi, Md Tafhimul Haque and Mridha, M. F. and Hossen, Md. Jakir (2026) Bio-MutClassNet: A Context-Aware Hybrid Transformer Framework for Precision Oncology via Attention-Based Genetic Mutation Classification. IEEE Open Journal of the Computer Society, 7. pp. 1026-1037. ISSN 2644-1268

Text 5.pdf - Published Version Restricted to Repository staff only Download (2MB)

Abstract

The automatic categorization of genetic mutations from medical narratives is a key application in personalized oncology, but this remains an intricate problem because of both linguistic complexity and a very imbalanced class in medical texts. Conventional deep learning methods, which solely utilize fixed word embedding or non-diversified architectures, are unable to discern semantically subtle differentials in benign and pathogenic variants. In this work, we present Bio-MutClassNet, a context-aware hybrid framework for genetic mutation classification from clinical narratives. The model leverages domain-specific transfer learning through BioBERT embeddings and integrates complementary convolutional and recurrent neural components to capture both local lexical patterns and long-range contextual dependencies. Instead of relying on static feature concatenation, we employ an adaptive branch-level attention fusion strategy that dynamically weights heterogeneous representations according to input characteristics. Evaluated on the widely used MSKCC dataset using stratified cross-validation, Bio-MutClassNet achieves an accuracy of 95.2% and an F1-score of 0.95, demonstrating consistent performance improvements over traditional machine learning approaches and standalone biomedical transformer baselines under comparable experimental settings. In addition, we incorporate explainability analysis to examine whether model decisions align with biologically meaningful terminology, supporting greater transparency in AI-assisted mutation classification.

Item Type:	Article
Uncontrolled Keywords:	Genetic mutation classification, deep learning, LSTM, BiLSTM, CNN, GRU, precision oncology, personalized healthcare
Subjects:	R Medicine > R Medicine (General) > R855-855.5 Medical technology
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	04 Jun 2026 07:33
Last Modified:	04 Jun 2026 07:33
URII:	http://shdl.mmu.edu.my/id/eprint/15942

Downloads

Edit (login required)