||January 25, 2010
||Dr. Shanfeng Zhu, Associate Professor, Fudan University, China
||Improve the MHC II-peptide binding prediction with ensemble approaches
In contrast to biological experiments, computational approaches for MHC peptide binding prediction can reduce the time and financial cost significantly. An accurate predictor will be of great benefit to immunologists for understanding the underlying mechanism of immune recognition as well as facilitating the process of epitope recognition and vaccine design. Although various computational prediction approaches have been developed, the performance of each approach is far from perfection. To make full use of current prediction methods and achieve a higher predictive performance, we developed MetaMHC that integrates the outputs of possible predictors based on ensemble approaches. The superiority of our ensemble predictors has been demostrated by the good prediction performacne in the experiment of not only internal cross validation, but also independent test dataset.
An additional note is that our predictor awarded the best performance team in the category of HLA A*0101 (9-mer) in recently-held Machine Learning in Immunology Competition
(http://www.kios.org.cy/ICANN09/MLI.html) out of 20 submissions in the world.