ProtDCal

ProtDCal generates numerical descriptors from protein sequences and 3D structures to encode sequence-structure-function relationships for downstream bioinformatics analyses.

Key Features:

• Extensive Descriptor Generation: Produces tens of thousands of features by combining sequence-based and 3D-structural descriptors to comprehensively encode protein data.
• Local and Group-Based Encoding: Implements novel local and group-based encoding methods that capture local residue context and group-level properties.
• 3D-Structure-Based Descriptors: Generates a wide variety of non-redundant 3D-structure-derived descriptors that complement sequence-based features.
• Predictive Modeling Support: Provides descriptor sets used to train and evaluate classification models, including models for N-linked glycosylation site prediction.

Scientific Applications:

• Protein Encoding for Machine Learning: Supplies numerical feature vectors for training and applying machine learning models on protein data.
• Protein Classification: Enables classification tasks using comprehensive sequence and structure descriptors.
• Similarity Analysis: Supports similarity and comparative analyses of proteins based on multi-modal descriptors.
• Function Prediction: Facilitates prediction of protein function through integrated sequence and structure features.
• Glycosylation Site Prediction: Has been applied to train and evaluate predictors of N-linked glycosylation sites.

Methodology:

Generates tens of thousands of sequence-based and 3D-structure-based numerical descriptors using novel local and group-based encoding methods, and uses these descriptors to train and evaluate classification models for tasks such as N-linked glycosylation site prediction.