StackTTCA: a stacking ensemble learning-based framework for accurate and high-throughput identification of tumor T...

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Identifying tumor T cell antigens (TTCAs) is vital for understanding cancer mechanisms and developing effective anticancer vaccines.
• Current experimental methods for TTCAs discovery are costly and time-consuming.
• Existing machine learning (ML) models for TTCAs identification require improvement in accuracy and precision.

Purpose of the Study:

• To develop a robust and accurate computational framework for identifying tumor T cell antigens (TTCAs).
• To enhance the efficiency and precision of novel TTCAs discovery for anticancer vaccine development.

Main Methods:

• A stacking ensemble learning framework, StackTTCA, was developed.
• 156 baseline models were created using 12 feature encoding schemes and 13 ML algorithms.
• A feature selection strategy was employed to optimize a probabilistic feature vector for the stacked model.

Main Results:

• StackTTCA achieved high accuracy (0.932) and Matthew's correlation coefficient (0.866) on independent tests.
• The proposed framework outperformed several existing ML classifiers and methods.
• StackTTCA enables accurate and large-scale identification of TTCAs.

Conclusions:

• StackTTCA precisely and rapidly identifies true TTCAs, facilitating follow-up experimental verification.
• An online web server was developed to provide convenient high-throughput screening of novel TTCAs.
• The framework supports the advancement of anticancer vaccine development through efficient TTCAs discovery.