SCEC

SCEC predicts protein structural classes from amino acid sequences, classifying proteins into the four major structural categories all-α, all-β, α/β, and α+β using evolutionary profile information.

Key Features:

• Sequence representation: Introduces a sequence representation based on PSI-BLAST profile-based collocation of amino acid (AA) pairs.
• Evolutionary information: Integrates PSI-BLAST-derived evolutionary profiles to capture complex relationships within protein sequences.
• Classification categories: Targets the four structural classes: all-α, all-β, α/β, and α+β.
• Predictive performance: Achieves accuracy rates ranging from 61% to 96% across evaluated datasets.
• Benchmarking and comparison: Evaluated on six benchmark datasets and reported error rate reductions of 14%–26% relative to recent methods.
• Classifier evaluation: Assessed with five representative classifiers, including support vector machines (SVMs).
• Sequence-similarity dependence: Reports accuracy reductions of 20%–35% for datasets with lower sequence identity levels (25%, 30%, 40%).

Scientific Applications:

• Structural class assignment: Assigns proteins to all-α, all-β, α/β, and α+β structural classes from sequence data.
• Protein folding analysis: Provides insights into protein folding patterns via structural class predictions.
• Function and interaction inference: Supports inference of protein function and interactions based on predicted structural class.
• Method development and benchmarking: Serves as a comparative benchmark for structural class prediction techniques.

Methodology:

Uses PSI-BLAST profile-based collocation of amino acid (AA) pairs for sequence representation and evaluates models with five representative classifiers including support vector machines on six benchmark datasets.