Predicting Success of Zinc Finger Targeting with Machine Learning
Author Information
Author(s): Sander Jeffry D, Reyon Deepak, Maeder Morgan L, Foley Jonathan E, Thibodeau-Beganny Stacey, Li Xiaohong, Regan Maureen R, Dahlborg Elizabeth J, Goodwin Mathew J, Fu Fengli, Voytas Daniel F, Joung J Keith, Dobbs Drena
Primary Institution: Massachusetts General Hospital
Hypothesis
Can machine learning classifiers accurately predict the success of zinc finger protein targeting using the OPEN method?
Conclusion
The ZiFOpT classifier can effectively identify DNA sequences likely to be successfully targeted by zinc finger proteins, significantly reducing experimental effort.
Supporting Evidence
- ZiFOpT achieved an overall accuracy of 87% and specificity of 90% on a dataset of 135 validated ZFP target sites.
- The classifier's performance was validated on an independent test set of 140 sites, achieving 88% accuracy.
- Using confidence scores, users can improve the accuracy of target site predictions significantly.
Takeaway
Scientists created a computer program that helps find the best spots in DNA to target with special proteins, making it easier to change genes.
Methodology
The study used Naïve Bayes and Support Vector Machine classifiers to analyze DNA sequences and predict their targeting success.
Potential Biases
Potential biases in the dataset could affect the generalizability of the classifier's predictions.
Limitations
The classifiers may not account for all variables affecting targeting success, and the dataset was limited to specific target sites.
Statistical Information
P-Value
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