Structural Descriptor Database: A Tool for Predicting Protein Functions
Author Information
Author(s): Juliana S Bernardes, Jorge H Fernandez, Ana Tereza R Vasconcelos
Primary Institution: Laboratório Nacional de Computação Científica
Hypothesis
Can a web-based tool effectively predict protein functions and functional site positions based on structural properties?
Conclusion
The Structural Descriptor Database (SDDB) outperformed existing methods in predicting active sites and achieved over 70% precision.
Supporting Evidence
- SDDB achieved an average precision of 99.61% and recall of 99.62% for predicting active sites in the Trypsion-like Serine protease data set.
- In the SCOP families experiment, SDDB achieved 84% precision and 70.8% recall for active site predictions.
- The method showed better performance in predicting active sites compared to binding sites due to the conservation of active site residues.
Takeaway
The SDDB is like a smart helper that can guess what a protein does by looking at its shape and structure.
Methodology
The study used Hidden Markov Models (HMM) to predict functional sites based on structural alignments and curated data.
Limitations
The predictions may vary in accuracy depending on the quality of the training data used.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website