Predicting Protein Structure and Accessibility
Author Information
Author(s): Pollastri Gianluca, Martin Alberto JM, Mooney Catherine, Vullo Alessandro
Primary Institution: University College Dublin
Hypothesis
Can machine learning systems improve the prediction of protein secondary structure and solvent accessibility by using homology information?
Conclusion
The developed predictive systems significantly outperform traditional ab initio methods when homology templates are available.
Supporting Evidence
- The predictive systems achieved approximately 90% accuracy for secondary structure prediction with over 30% sequence similarity.
- Porter_H outperformed the baseline by significant margins across various template qualities.
- The systems are capable of processing up to 20,000 queries a day.
Takeaway
Scientists created computer programs that can guess how proteins are shaped and how accessible they are, and they work better when they can look at similar proteins.
Methodology
The study used machine learning techniques to analyze protein sequences and structural templates from the PDB to predict secondary structure and solvent accessibility.
Limitations
The performance of the predictive systems may decrease with lower quality templates and marginal sequence similarity.
Statistical Information
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website