Improving MHC-I Binding Prediction with Multiple Tools
Author Information
Author(s): Brett Trost, Mik Bickis, Anthony Kusalik
Primary Institution: University of Saskatchewan
Hypothesis
Peptides predicted by a number of tools are more likely to bind than those predicted by just one tool.
Conclusion
Combining the results of individual tools leads to better identification of high-affinity binders with fewer false positives.
Supporting Evidence
- The heuristic-based method significantly outperformed individual tools at high specificity thresholds.
- The combined tool showed improved sensitivity for identifying binders compared to the best individual tool.
- Using known nonbinders for testing provides a stronger assessment of each tool's utility.
Takeaway
Using many prediction tools together helps find more good peptides that can fight infections.
Methodology
A heuristic-based method was developed to combine predictions from multiple MHC-binding prediction tools and evaluated using ten-fold cross-validation.
Potential Biases
Potential bias due to the use of known nonbinders for testing the tools.
Limitations
The performance of the combined method may be influenced by the specificities chosen for individual tools.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website