SADeepcry

SADeepcry predicts protein crystallization propensity across production, purification, and crystallization stages using deep learning models applied to protein sequences.

Key Features:

• Stage-Specific Crystallization Prediction: Estimates success probabilities for protein production, purification, and crystallization to support experimental planning in X-ray diffraction (XRD) workflows.
• Self-Attention and Auto-Encoder Architecture: Extracts sequence-derived structural and physicochemical features and captures global long-range dependencies within protein sequences.

Scientific Applications:

• Protein Structure Determination Optimization: Prioritizes protein targets for X-ray diffraction–based structural studies by predicting crystallization likelihood.

Methodology:

SADeepcry integrates optimized self-attention and auto-encoder neural networks to encode protein sequence, structural, and physicochemical features, modeling long-distance spatial dependencies to predict crystallization propensity at multiple experimental stages.