Using Protein Structure Alignments to Improve Structure Prediction
Author Information
Author(s): Eric D Scheeff, Philip E Bourne
Primary Institution: San Diego Supercomputer Center, University of California, San Diego
Hypothesis
Can protein structure alignments enhance the prediction of protein structures from sequences?
Conclusion
Combining traditional models with structure alignment-enhanced models improves predictions for remote homologs.
Supporting Evidence
- Models using structure alignments did not improve predictions over sequence-only models for superfamily-level predictions.
- Combining structure alignment-enhanced models with sequence-only models improved results.
- SLAHMMs provided superior fold-level structure assignments compared to sequence-only models.
Takeaway
This study shows that using both regular and structure-based models together helps scientists better guess how proteins are shaped, especially when they are very different from known proteins.
Methodology
The study used iterative protocols to create profile hidden Markov models (HMMs) that incorporated both sequence and structure alignments.
Limitations
SLAHMMs require at least two structural representatives per superfamily, which may not always be available.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website