NN-breastcancerproteins

NN-breastcancerproteins predicts proteins associated with breast cancer from protein sequence descriptors using machine learning to prioritize biomarkers and therapeutic targets.

Key Features:

• Molecular Descriptors: Uses six distinct sets of protein sequence descriptors to capture biochemical properties relevant to breast cancer proteins.
• Machine Learning Integration: Evaluates 13 machine learning methods with a univariate feature selection process applied across five descriptor families and identifies the multilayer perceptron (artificial neural network) as the most effective classifier.
• High Predictive Performance: Reports AUROC 0.980 ± 0.0037 and accuracy 0.936 ± 0.0056, validated by 3-fold cross-validation.
• Comprehensive Protein Ranking: Ranks proteins across biological categories, including cancer immunotherapy proteins (e.g., RPS27, SUPT4H1, AKT3), metastasis driver proteins (e.g., S100A9, MAPK1), and RNA-binding proteins (e.g., RPS27L, MRPL54).

Scientific Applications:

• Biomarker and therapeutic-target prediction: Prioritizes proteins associated with breast cancer for use as candidate biomarkers and therapeutic targets.
• Drug design and personalized medicine: Provides ranked protein candidates to inform target selection in drug design and personalized medicine strategies.
• Mechanistic studies: Supports investigation of molecular mechanisms by ranking proteins linked to immunotherapy, metastasis, and RNA-binding functions.

Methodology:

Computational workflow uses six protein sequence descriptor sets, univariate feature selection across five descriptor families, evaluation of 13 machine learning methods, selection of a multilayer perceptron classifier, and performance assessment by 3-fold cross-validation.