Predicting Glycosylation Sites Using Machine Learning
Author Information
Author(s): Caragea Cornelia, Sinapov Jivko, Silvescu Adrian, Dobbs Drena, Honavar Vasant
Primary Institution: Iowa State University
Hypothesis
Can ensembles of Support Vector Machine classifiers improve the prediction of glycosylation sites in proteins?
Conclusion
Ensembles of Support Vector Machine classifiers provide a more accurate method for identifying glycosylation sites in glycoprotein sequences compared to single classifiers.
Supporting Evidence
- Ensembles of SVM classifiers outperformed single SVM classifiers in predicting glycosylation sites.
- The study utilized a dataset from O-GlycBase containing experimentally verified glycosylation sites.
- The classifiers were evaluated using sequence-based k-fold cross-validation.
Takeaway
This study shows that using a group of machine learning models can help us better find where sugars attach to proteins, which is important for understanding how proteins work.
Methodology
The study used ensembles of Support Vector Machine classifiers trained on a dataset of glycosylation sites to predict glycosylation in protein sequences.
Potential Biases
The classifiers may be biased towards the majority class due to the unbalanced dataset.
Limitations
The dataset is highly unbalanced, which can affect the performance of classifiers.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website