Related Experiment Videos
Probabilistic scoring measures for profile-profile comparison yield more accurate short seed alignments
David Mittelman1, Ruslan Sadreyev, Nick Grishin
1Howard Hughes Medical Institute Department of Biochemistry, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-9050, USA.
Bioinformatics (Oxford, England)
|August 13, 2003
Summary
Probabilistic methods, including log-odds and prof_sim, are best for seeding local profile-profile alignments. Modified COMPASS and Picasso functions proved most effective for scoring protein sequence comparisons.
Area of Science:
- Bioinformatics
- Computational Biology
- Genomics
Background:
- Automatic protein sequence comparison is crucial for understanding protein families.
- Profile-profile comparisons enhance sensitivity for distantly related sequences.
- Accurate scoring of position matches is key to profile-profile comparison.
Purpose of the Study:
- To evaluate various scoring functions for profile-profile alignment.
- To identify the most effective methods for generating accurate ungapped alignments.
Main Methods:
- Implementation of previously reported and novel scoring functions.
- Comparison of scoring functions based on alignment accuracy.
- Focus on generating short, ungapped alignments.
Main Results:
- Probabilistic methods (log-odds, prof_sim) are suitable for initial alignment seeding.
- Modified COMPASS and Picasso scoring functions demonstrated high effectiveness.
- The study identified superior scoring systems for local profile-profile alignments.
Conclusions:
- Probabilistic scoring approaches are recommended for initiating local profile-profile alignments.
- Optimized scoring functions, derived from COMPASS and Picasso, enhance alignment accuracy.