Target selection for structural genomics based on combining fold recognition and crystallisation prediction methods:...

Area of Science:

• Structural Biology
• Bioinformatics
• Proteomics

Background:

• X-ray crystallography is crucial for determining 3D protein structures.
• Identifying suitable protein targets is a bottleneck in structural genomics.
• Predicting crystallization success aids in efficient structure determination.

Purpose of the Study:

• To automatically identify human proteome regions amenable to 3D structure determination via X-ray crystallography.
• To predict the likelihood of these regions yielding diffraction-quality crystals.
• To provide a tool for selecting human proteins based on predicted crystallization success.

Main Methods:

• Combined fold recognition and crystallization prediction algorithms.
• Applied algorithms to the entire human proteome.
• Utilized structure prediction, crystallization prediction, and transmembrane helix prediction algorithms for classification.

Main Results:

• Estimated 40,000 crystallizable regions in the human proteome.
• Currently, only 15% (approx. 6,000 sequences) are solved.
• Categorized remaining regions into 5 classes plus an integral membrane protein (IMP) class.
• Identified ~750 unsolved regions with PDB templates and 'optimal' crystallization likelihood.
• Classified over 10,500 regions as 'very difficult' to crystallize and ~2,500 predicted to contain at least 3 transmembrane helices.

Conclusions:

• The study provides a computational tool (3D-SPECS) for predicting human protein crystallizability.
• Highlights the vast potential of the human proteome for structural studies.
• Facilitates targeted experimental efforts in structural genomics by prioritizing regions with higher predicted success rates.