Identifying microRNAs using one-class machine learning
Author Information
Author(s): Yousef Malik, Jung Segun, Showe Louise C, Showe Michael K
Primary Institution: The Wistar Institute
Hypothesis
Can one-class machine learning methods effectively identify microRNAs when the negative class is not well defined?
Conclusion
One-class methods can be as effective as two-class methods for identifying microRNAs, especially when the negative class is poorly characterized.
Supporting Evidence
- One-class methods showed average accuracies of 70–80% compared to 90% for two-class methods.
- Using the EBV genome, one-class machine learning performed as well as or better than two-class approaches.
- One-class methods eliminate the need to define a negative class, which can be ambiguous.
Takeaway
This study shows that we can find tiny RNA molecules called microRNAs using a simpler method that doesn't need to guess what the opposite type looks like.
Methodology
The study compared one-class machine learning methods to two-class methods using various classifiers like naïve Bayes and Support Vector Machines.
Potential Biases
The arbitrary selection of negative examples can introduce bias in the results.
Limitations
The performance of one-class methods may be affected by the selection of features and the definition of the negative class.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website