POSSUM

POSSUM generates numerical sequence feature descriptors derived from Position-Specific Scoring Matrices (PSSMs) for use in machine-learning analyses of protein sequences and protein attribute prediction.

Key Features:

• PSSM-based feature generation: Produces numerical sequence feature descriptors derived from Position-Specific Scoring Matrices (PSSMs) that encode evolutionary information in protein sequences.
• Twenty-one descriptor types: Generates 21 distinct types of PSSM-based feature descriptors.
• Machine-learning compatibility: Outputs numerical descriptors suitable for feature extraction, feature selection, and benchmarking in machine-learning models.
• Protein attribute relevance: Provides descriptors specifically intended to support predictors of protein attributes.

Scientific Applications:

• Feature extraction for ML: Supplying PSSM-derived numerical descriptors to train classifiers and regressors on protein sequence data.
• Feature selection and benchmarking: Enabling selection and benchmarking of PSSM-based features during model development and evaluation.
• Protein attribute prediction: Enhancing the input feature set for predictors of protein attributes by incorporating evolutionary information from PSSMs.

Methodology:

Generates numerical sequence feature descriptors from Position-Specific Scoring Matrices (PSSMs), producing 21 distinct PSSM-based descriptor types.