The human MHC class I system is extremely polymorphic and now the registered number of HLA-I molecules has exceeded 3000. These HLAs have slightly different amino acid sequences, which makes them potentially bind to different sets of peptides. In order to overcome the problems existing in current methods and take into account the MHC sequence and energies of contact residues information, in this paper a method based on artificial neural network is proposed to predict peptides binding to HLAs on unknown alleles with limited or even no prior experimental data. The super-type experiments are implemented to validate our method. In the experiment, we collected 14 HLA-A and 14 HLA-B molecules on Bjoern Peters dataset and did leave-one-out cross-validation on MHC-peptide binding data with different alleles but sharing the same super-type, our method got the best average AUC value as 0.846 compared to gold standard methods such as NetMHC and NetMHCpan. © 2012 Springer-Verlag.
CITATION STYLE
Luo, F., Gao, Y., Zhu, Y., & Liu, J. (2012). Predicting binding-peptide of HLA-I on unknown alleles by integrating sequence information and energies of contact residues. In Communications in Computer and Information Science (Vol. 304 CCIS, pp. 160–165). https://doi.org/10.1007/978-3-642-31837-5_23
Mendeley helps you to discover research relevant for your work.