Porter

Porter predicts protein secondary structure from amino acid sequences to support analysis of protein function, stability, and interactions.

Key Features:

• Ab initio secondary structure prediction: Predicts protein secondary structure using ab initio methods from sequence information.
• Advanced Neural Network Architecture: Employs ensembles of cascaded Bidirectional Recurrent Neural Networks (BiRNNs) and Convolutional Neural Networks (CNNs) to capture complex patterns in protein sequences.
• Innovative Input Encoding Techniques: Incorporates new input encoding strategies to enhance representation of input data.
• Comprehensive Training Dataset: Trained on an extensive set of protein structures to improve robustness across datasets.
• High Accuracy Rates: Porter 5 achieves 84% accuracy (81% Sequence Order Verification - SOV) for three-state predictions and 73% accuracy (70% SOV) for eight-class predictions, representing a 2% improvement over the previous version.
• Performance Against Competitors: Outperforms or matches the most recent secondary structure predictors tested.
• Homology-Free Testing: When retrained on SCOPe-based datasets that eliminate homology between training and testing samples, it maintains similar performance levels.

Scientific Applications:

• Secondary structure analysis: Provides three-state and eight-class secondary structure predictions to inform studies of protein function, stability, and interactions.
• Drug discovery and molecular biology research: Supplies structural information useful for drug discovery workflows and molecular biology investigations.

Methodology:

Uses ensembles of cascaded BiRNNs and CNNs trained on both single-sequence inputs and evolutionary profile-based inputs, and includes retraining on SCOPe-based datasets for homology-free evaluation.