Computational protocol: Computational approaches to selecting and optimising targets for structural biology

Similar protocols

Protocol publication

[…] Having identified a pool of sequences that possess appropriate structure and function relationships according to the project scope, the next logical step is to determine promising candidates for experimental work. As noted above, successful progression of a selected target through to the stage of diffraction-quality crystals is a critical consideration. Algorithms to estimate this include XANNpred, OB-Score, ParCrys, XtalPred, PPCpred and PDPredictor . Approaches focused on key stages of the structural biology pipeline have also been developed, including predictors of soluble expression (e.g. PROSO , SOLpro ) and crystallisation (e.g. PXS , SECRET ). SECRET is limited to only accept sequences of length 46–200 residues . Predictors that focus on a specific experimental stage are particularly useful when protein targets have already reached the given stage in the pipeline, especially in target optimisation to propose alternative constructs; the SERp surface entropy reduction server is an example . Estimating overall success of selected targets with a single predictor is much more appealing than using multiple single-stage predictors. Indeed, a linear combination of multiple predictors suffers from error multiplication and makes candidate target ranking more cumbersome. Consider a strategy combining two predictors to separately estimate soluble expression and crystallisation propensity. If each predictor gave 75% accuracy individually, accuracy for progression through both stages would be only 56%. Moreover, biophysical properties that are advantageous at one stage (e.g. solubility) may conflict with properties required for success at another stage (e.g. crystallisation) . Accordingly, an attractive approach applies a single algorithm to select targets with the right balance of biophysical properties to successfully navigate all stages of the structural proteomics pipeline. As noted above, this strategy is available via the algorithms XANNpred , PPCpred , PDPredictor, XtalPred and ParCrys/OB-Score . Interestingly, PPCpred provides a single prediction for overall success, as well as estimating success at three individual pipeline stages and so informs on expected point(s) of failure. The subsections below give further discussion on the relative merits of these methods, with emphasis on those developed at the Scottish Structural Proteomics Facility (SSPF). Of the algorithms examined, XANNpred was found to be best-performing (Subsection and ). […]

Pipeline specifications

Software tools XANNpred, OB-Score, ParCrys, XtalPred, PPCpred, PDPredictor, SOLpro
Applications Protein structure analysis, Protein physicochemical analysis
Organisms Colocasia esculenta
Diseases Genetic Diseases, Inborn