XANNpred: Predicting Protein Crystallization Success
Author Information
Author(s): Ian M Overton, C A Johannes van Niekerk, Geoffrey J Barton
Primary Institution: University of Dundee
Hypothesis
Can neural networks effectively predict the propensity of proteins to yield diffraction-quality crystals?
Conclusion
XANNpred outperforms existing algorithms in predicting which proteins are likely to produce diffraction-quality crystals.
Supporting Evidence
- XANNpred achieved accuracy values between 75% to 81% on blind tests.
- XANNpred outperformed other algorithms like XtalPred and PXS.
- XANNpred-PDB and XANNpred-SG were developed for different datasets to enhance prediction accuracy.
Takeaway
Scientists created a computer program that helps choose proteins that are more likely to form good crystals for studying their structure.
Methodology
The study used two neural networks trained on datasets of proteins known to produce diffraction-quality crystals and those that did not.
Potential Biases
Potential bias due to the selection of training datasets.
Limitations
XANNpred may not be strongly predictive for transmembrane proteins.
Statistical Information
P-Value
0.0062
Statistical Significance
p ≤ 0.0062
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website