Efficient Classification of Hallmark of Cancer Using Embedding-Based Support Vector Machine for Multilabel Text

Abstract

The Hallmark of Cancers consists of various biological capabilities of the tumor cell which help the medical experts to understand the development and identification of these cells during various stages of the cancer disease. The hallmark of cancer classification is a widely accepted framework that characterizes the fundamental biological capabilities of cancer cells. This classification is based on the work of Hanahan and Weinberg, who identified 10 hallmark capabilities that collectively enable the development and progression of cancer. The hallmark of cancer classification provides a comprehensive framework for understanding the biological basis of cancer development and progression. It helps researchers to identify the key molecular and cellular pathways that are involved in the disease, which can inform the development of new diagnostic tools and therapies. Multi-label classification aims to assign a set of labels to the samples under study. This paper focuses on creating an improved model by hybridizing the biomedical domain-specific embeddings for all the extracted biomedical features on the machine learning model. The use of domain-specific embeddings adds semantics to the vector-represented text. More specifically the study has tried to improve the efficacy of the multi-label classification as compared with other state-of-art methods using BioWordVec and the MeSH embeddings. The experimental work showed a significant improvement in the performance of our model which is being trained on the machine learning algorithm Support Vector Machine (SVM). The paper also focuses on understanding the label correlation which is studied by conducting a case study with medical domain experts and is also analyzed with the proposed model.