Identifying Domain Linkers Using Neural Networks
Author Information
Author(s): Miyazaki Satoshi, Kuroda Yutaka, Yokoyama Shigeyuki
Primary Institution: University of Tokyo
Hypothesis
Can a neural network effectively identify domain linkers in a large protein sequence database?
Conclusion
The study shows that combining neural network predictions with low-complexity region detection improves the identification of domain boundaries in proteins.
Supporting Evidence
- The neural network identified 8133 putative domain linkers with a cutoff of 0.95.
- 75% of the evaluated putative domain linkers were correctly located within 20 residues of a domain terminus.
- Combining predictions from the neural network and low-complexity regions improved sensitivity without decreasing specificity.
Takeaway
The researchers used a computer program to find parts of proteins that connect different sections, helping scientists understand protein structures better.
Methodology
The study employed a neural network trained on a small dataset to predict domain linkers from the SWISSPROT database.
Limitations
The predictions may underestimate the actual performance due to the classification of linkers and the ambiguity in domain boundaries.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website