Related Experiment Videos
Super paramagnetic clustering of protein sequences.
Igor V Tetko1, Axel Facius, Andreas Ruepp
1GSF National Research Center for Environment and Health, Institute for Bioinformatics, Ingolstädter Landstrasse 1, D-85764 Neuherberg, Germany. i.tetko@gsf.de
BMC Bioinformatics
|April 5, 2005
Summary
The global Super Paramagnetic Clustering (gSPC) algorithm enhances protein family detection by improving clustering accuracy and sequence coverage compared to existing methods. This advancement aids in the automatic annotation of entire genomes.
Area of Science:
- Bioinformatics
- Computational Biology
- Genomics
Background:
- Detecting sequence homologues is crucial for protein family discovery and automated annotation.
- Challenges include diverse protein family functions, inhomogeneity, and varying sizes, complicating existing clustering methods.
Purpose of the Study:
- To analyze the Super Paramagnetic Clustering (SPC) and its extension, the global SPC (gSPC) algorithm.
- To evaluate the performance of gSPC in clustering protein sequences for family identification.
Main Methods:
- The study employed clustering algorithms analogous to ferromagnetic physics.
- Algorithms analyzed include SPC, global SPC (gSPC), and Tribble-MCL (TRIBE-MCL).
- Performance was assessed using SwissProt, SCOP databases, and MIPS FunCat annotation of bacterial genomes.
Main Results:
- gSPC demonstrated up to 30% improvement in specificity and sensitivity over SPC and TRIBE-MCL on SwissProt and SCOP.
- gSPC covered approximately 12% more sequences than other methods during bacterial genome annotation.
- All three algorithms yielded similar results for MIPS FunCat 1.3 annotation.
Conclusions:
- gSPC offers higher accuracy and greater sequence coverage than TRIBE-MCL.
- gSPC is a valuable tool for automatic protein family detection.
- The algorithm facilitates unsupervised annotation of complete genomes.