SSE-PSSM

SSE-PSSM enhances protein secondary structure prediction by constructing a secondary structure element-based position-specific scoring matrix that generates informative features for machine learning-based SSP.

Key Features:

• Secondary structure element-based PSSM: Constructs a position-specific scoring matrix derived from secondary structure elements rather than solely from amino acid sequences.
• Feature generation for machine learning: Produces an informative feature set from the SSE-PSSM intended for use in machine learning algorithms for SSP.
• Compatibility with amino acid PSSMs: Integrates with conventional amino acid-based PSSMs to augment existing feature representations.
• Benchmarking and validation: Demonstrated consistent outperformance of traditional amino acid-based PSSMs in independent tests with training and testing datasets sharing less than 25% sequence identity.
• Reported accuracy improvements: Preliminary combined tests showed average improvements of 2.0% for three-state and 5.2% for eight-state SSP accuracies.

Scientific Applications:

• Three-state and eight-state protein secondary structure prediction: Improves accuracy metrics for both three-state and eight-state SSP tasks.
• Machine learning-based SSP enhancement: Serves as a source of enriched features to boost performance of machine learning models for secondary structure prediction.
• Integration with SSP techniques: Can be combined with state-of-the-art SSP methods to attempt to surpass current accuracy limitations.

Methodology:

SSE-PSSM constructs a position-specific scoring matrix based on secondary structure elements, derives feature sets for machine learning-based SSP, integrates with conventional amino acid PSSMs, and was validated using independent tests with training and testing datasets sharing less than 25% sequence identity.