Machine Learning to Find Binding Sites in Human Genes
Author Information
Author(s): Thomas Down, Bernard Leong, Tim JP Hubbard
Primary Institution: Wellcome Trust Sanger Institute
Hypothesis
Can a machine learning model effectively identify candidate binding sites in human protein-coding sequences?
Conclusion
The study successfully trained a computational model that detects signals in coding exons, suggesting the presence of binding sites for proteins that influence RNA splicing.
Supporting Evidence
- The model identified several known splice enhancer elements.
- It effectively distinguished between coding and non-coding exons.
- The neutralization process preserved overall sequence composition while changing the sequence.
- The model included 216 scaffolds, indicating a complex representation of binding sites.
Takeaway
Researchers used a computer program to find hidden signals in gene sequences that help control how genes are spliced together, which is important for making proteins.
Methodology
The study involved extracting coding exons, generating neutralized sequences, and training a machine learning model to distinguish between original and neutralized exons.
Potential Biases
Potential overfitting due to the model being trained only on coding exons.
Limitations
The model may not capture all splicing signals, especially those requiring sequences in the flanks of exons.
Statistical Information
P-Value
0.007, 0.010, 0.005, 0.020
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website