Predicting Gene Expression from Sequence: A Reexamination
Author Information
Author(s): Yuan Yuan, Guo Lei, Shen Lei, Jun S Liu
Primary Institution: Department of Statistics, Harvard University
Hypothesis
Can gene expression patterns be accurately predicted from the information in their respective promoter sequences?
Conclusion
The naïve Bayes classifiers developed in this study achieved a higher prediction accuracy of 79% compared to the previous Bayesian network model's 73%.
Supporting Evidence
- The naïve Bayes classifiers correctly predicted expression patterns for 79% of the genes.
- The previous Bayesian network model achieved a prediction accuracy of 73%.
- The study found that the cross-validation procedure used by BT may have overestimated their method's prediction accuracy by about 10%.
- Adding information about motif orientation and position did not improve the naïve Bayes classifier's predictive power.
Takeaway
This study found a simpler way to predict how genes work based on their DNA sequences, and it worked better than a more complicated method used before.
Methodology
The study used naïve Bayes classifiers trained on motif matching scores from promoter sequences to predict gene expression patterns.
Potential Biases
The feature selection process may introduce bias if the motifs are not representative of the true regulatory elements.
Limitations
The study's results may not generalize beyond the specific dataset and methods used, and the naïve Bayes model assumes independence among features.
Participant Demographics
The study focused on 2,587 genes from the yeast Saccharomyces cerevisiae.
Statistical Information
P-Value
1 × 10−64
Statistical Significance
p < 1 × 10−21
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website