PredNTS

PredNTS predicts nitrotyrosine sites in proteins, a post-translational modification generated by reactive nitrogen species, using machine learning on multiple sequence-derived features.

Key Features:

• Sequence feature integration: Incorporates K-mer representations, composition of k-spaced amino acid pairs (CKSAAP), AAindex profiles, and binary encoding schemes for comprehensive sequence encoding.
• Feature selection: Applies recursive feature elimination to identify the most informative features using a random forest classifier.
• Ensemble scoring: Refines predictions by linearly combining successive random forest probability scores from models employing different encoding schemes.
• Classifier choice and comparison: Utilizes a random forest classifier and reports comparisons with other machine learning algorithms to assess relative performance.
• Performance evaluation: Assessed by five-fold cross-validation and independent dataset testing, with a reported AUC of 0.910.
• Curated datasets: Evaluations and model building are performed using curated datasets.

Scientific Applications:

• Nitrotyrosine site prediction: Identification of specific tyrosine nitration sites within protein sequences.
• Post-translational modification analysis: Supporting studies of protein function affected by tyrosine nitration.
• Experimental prioritization: Prioritizing candidate nitration sites for experimental validation and hypothesis generation.

Methodology:

Integrates K-mer, CKSAAP, AAindex and binary encodings; uses recursive feature elimination with a random forest classifier for feature selection; constructs predictions by linearly combining successive random forest probability scores from models trained on different encodings; evaluates performance by five-fold cross-validation and independent-dataset testing (AUC 0.910) with comparisons to other machine learning algorithms.