Sequence-based prediction of protein crystallization, purification and production propensity

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• X-ray crystallography is key for protein structure determination but has low success rates.
• Existing in silico prediction tools show declining accuracy with recent crystallization data.
• There is a need for improved computational methods to select crystallization targets.

Purpose of the Study:

• To develop a novel computational approach for predicting protein crystallization success.
• To improve the accuracy of predicting diffraction-quality crystal formation.
• To identify key factors influencing protein production, purification, and crystallization.

Main Methods:

• A new dataset and annotation protocol were used to track crystallization progress.
• The predictor, PPCpred, utilizes sequence-derived inputs including energy, hydrophobicity, disorder, and amino acid composition.
• The method predicts success for the entire crystallization process and identifies failure points.

Main Results:

• PPCpred significantly outperforms existing alignment-based and modern crystallization propensity predictors.
• The tool accurately predicts protein production, purification, and diffraction-quality crystal formation.
• Analysis revealed intuitive factors like amino acid content (Cys, His, Ser) and disorder influence success.

Conclusions:

• PPCpred offers a more accurate and reliable method for selecting protein crystallization targets.
• The tool is particularly beneficial for achieving high true positive rates in predictions.
• Understanding sequence-derived features aids in optimizing the protein crystallization workflow.