Sequence-Based Intelligent Model for Identification of Tumor T Cell Antigens Using Fusion Features

Abstract

Cancer is a chronic disease, and recent cases of cancer cause a challenge for the whole world of medicine. Current studies have shown that T-cell antigens play a significant role in cancer treatment. Therefore, identifying tumor T cell antigens is vital for cancer detection and drug development. Several computational methods have been proposed to identify tumor T cell antigens. These models have produced promising results; however, they ignored the sequence’s physiochemical structure and correlation information during the feature transformation process. This study proposes a robust model considering sequence structure and correlation information for identifying tumor T-cell antigens. The proposed model uses three feature encoding (transformation) techniques and a multi-layer deep learning model. Additionally, the proposed model employed a feature fusion approach to consider sequence structure and correlation information during feature transformation. The fusion approach added redundant and noisy features in the feature vectors, which were removed during the feature selection phase. The performance of the proposed model is extensively evaluated by considering different measurement metrics using 10-fold cross-validation. The experimental results showed that the proposed model achieved 99.02% and 98.62% accuracies, improving an average of 14.66% and 23.75% using training and independent datasets, respectively.

Topics & Concepts