Neuropred

Neuropred predicts cleavage sites in neuropeptide precursor sequences at basic amino acid locations using logistic regression models trained on experimentally verified cleavage data from mollusks, mammals, and insects.

Key Features:

• Cleavage site prediction: Predicts cleavage sites at basic amino acid locations within neuropeptide precursor sequences.
• Logistic regression models: Employs binary logistic regression models trained on experimentally verified or published cleavage data and incorporating known amino acid motifs.
• Multiple model options: Provides several prediction models, including a simpler model and a more complex model evaluated on Aplysia californica data.
• Performance metrics: Reports model accuracy metrics, including a 97% correct classification rate, sensitivity, and specificity on the Aplysia californica prohormone dataset.
• Sequence input: Accepts single or multiple amino acid sequences for analysis.
• User-defined functions: Permits integration of user-defined functions into prediction workflows.
• Probability confidence and thresholding: Calculates confidence interval limits for cleavage probabilities and allows thresholding to convert probabilities into cleavage/non-cleavage calls.
• Peptide mass calculation: Computes masses of predicted peptides and includes optional user-specified post-translational modifications in mass calculations.
• Model validation with user data: Accepts user-specified cleavage observations to compute observed-versus-predicted model accuracy statistics.

Scientific Applications:

• Neuropeptide processing analysis: Identification and analysis of prohormone cleavage sites in neuropeptide precursor sequences.
• Mass spectrometry support: Support for mass spectrometry-based discovery and confirmation of novel neuropeptides via predicted peptide masses and optional post-translational modifications.
• Comparative cleavage studies: Comparative analysis of cleavage patterns across organisms such as mollusks, mammals, and insects.
• Model evaluation and refinement: Validation and refinement of cleavage-prediction models using user-provided cleavage data and reported accuracy statistics.

Methodology:

Binary logistic regression analyses were applied to a dataset of 22 Aplysia californica prohormones (750 cleavage sites) to train two models (simple and complex); the tool computes cleavage probabilities with confidence intervals, applies thresholds to call cleavage events, calculates predicted peptide masses including optional post-translational modifications, and produces observed-versus-predicted accuracy statistics when user-specified cleavage data are provided.