TAP

TAP predicts tumor-associated antigens (TAAs) for cancer immunotherapy by applying machine learning classifiers to physicochemical and biochemical properties.

Key Features:

• Machine Learning Integration: Evaluates fifteen machine learning algorithms and identifies a quadratic discriminant classifier with reported accuracy 0.7384 and AUC 0.817.
• Comprehensive Feature Calculation: Computes 544 physicochemical and biochemical properties per antigen using a custom Python script and the AAindex database.
• Robust Dataset Construction: Uses curated antigen datasets sourced from TANTIGEN and IEDB for training and validation.
• Validation and Testing: Performs 10-fold cross-validation and testing on an independent dataset to assess model generalizability.

Scientific Applications:

• Tumor Antigen Identification: Prioritizes candidate TAAs for selection as targets in cancer immunotherapy research.
• Vaccine and Immunotherapy Development: Supports development of personalized cancer vaccines and other immunotherapeutic strategies by providing predicted antigen candidates.

Methodology:

Antigens were curated from TANTIGEN and IEDB, 544 AAindex-derived properties per antigen were calculated using a Python script, fifteen classifiers were evaluated (with quadratic discriminant classifier selected), and models were assessed by 10-fold cross-validation followed by independent dataset testing.