Related Experiment Videos
The influence of gapped positions in multiple sequence alignments on secondary structure prediction methods
1Bioinformatics Section, Faculty of Sciences, Vrije Universiteit, De Boelelaan 1081A, 1081 HV Amsterdam, The Netherlands.
Computational Biology and Chemistry
|November 24, 2004
Summary
This study introduces SymSSP, an algorithm that improves protein secondary structure prediction by preserving evolutionary information lost in traditional methods. SymSSP enhances prediction accuracy by reintroducing lost data and using a novel consensus strategy.
Area of Science:
- Bioinformatics
- Computational Biology
- Structural Biology
Background:
- Leading protein secondary structure prediction methods rely on multiple sequence alignments.
- Current methods often delete gapped positions, losing crucial position-specific evolutionary information.
Purpose of the Study:
- To investigate the impact of information loss on secondary structure prediction accuracy.
- To develop an algorithm, SymSSP, that mitigates information loss and improves prediction quality.
Main Methods:
- Designed SymSSP to post-process predicted secondary structures using evolutionary information.
- Reintroduced information lost through gap deletion in multiple sequence alignments.
- Employed a novel dynamic programming routine for optimally segmented consensus secondary structure prediction.
Main Results:
- SymSSP improved segmentation quality compared to majority voting across tested methods (PHD, PROFsec, SSPro2, JNET).
- The consensus-deriving dynamic programming strategy enhanced prediction accuracy.
- Investigated noise from prediction errors, finding helix and strand edge predictions are often incorrect.
Conclusions:
- SymSSP effectively reintroduces lost evolutionary information, enhancing protein secondary structure prediction.
- The novel consensus strategy outperforms traditional methods like majority voting.
- Prediction errors at helix/strand edges remain a challenge for current methods.