||May 29, 2006
||Dr. Paul Horton, CBRC (Computational Biology Research Center), National Institute of Advanced Industrial Science and Technology
||Protein Subcellular Localization Prediction with WoLF PSORT
|| I will present a new program for predicting protein subcellular
localization from amino acid sequence. WoLF PSORT is a major update
to the PSORTII program, based on new sequence data and incorporating
new features with a feature selection procedure. Following
SWISS-PROT, we divided eukaryotes into three groups: fungi, plant, and
animal. For the 2113 fungi proteins divided into 14 categories; we
found that, combined with BLAST, WoLF PSORT yields a cross-validated
accuracy of 83%, eliminating about 1/3 of the errors made when using
BLAST alone. For 12771 animal proteins a combined accuracy of 95.6%
is obtained, eliminating 1/4 of BLAST alone errors, and for 2333 plant
proteins the accuracy can be improved to 86% from 84%.
In the talk I will use the disease protein treacle as a running
example to illustrate the type of information one may obtain with the
WoLF PSORT server. Time permitting, I may also briefly discuss some
practical issues of maintaining a popular server (which has been
accessed by approximately 20,000 unique URLs) on a shoe string budget.
Kenta Nakai, Minoru Kanehisa.
A Knowledge Base for Predicting Protein Localization Sites in
Genomics, 14, 897-911(1992).
Paul Horton, Kenta Nakai.
Better Prediction of Protein Cellular Localization Sites with the k
Nearest Neighbors Classifier.
Proceeding of the Fifth International Conference on Intelligent Systems
for Molecular Biology,
WoLF PSORT (http://wolfpsort.org)
Paul Horton, Keun-Joon Park, Takeshi Obayashi, Kenta Nakai.
Protein Subcellular Localization Prediction with WoLF PSORT.
the 4th Annual Asia Pacific Bioinformatics Conference APBC06
, 39-48 (2006).