Prediction of MHC Binding Peptides Using Artificial Neural Networks

		/vol02/vbb/	© Copyright 1995
		Volume 02	Received: Accepted:	---- ----

Prediction of MHC Binding Peptides Using Artificial Neural Networks Vladimir Brusic, George Rudy and Leonard C. Harrison
Abstract
	major function of the immune system is to recognise proteins foreign to the host organism. In vertebrates, peptides derived from foreign proteins are bound by major histocompatibility complex (MHC) molecules on the surface of host cells and thereby recognised by T-cells of the immune system. Binding of peptide to MHC is necessary for immune recognition, but only certain peptides can bind to particular MHC molecules. We have studied the prediction of MHC/peptide binding using an artificial neural network (ANN). A database comprising over 4000 peptide sequences known to bind MHC molecules was compiled from published sources. Training sets were drawn from these data and used to train a fully connected 3-layer back-propagation network. Three distinct representations of the input data were investigated. We compared predictions for peptide binding to human HLA-A2 and mouse MHC molecules to published experimental data not yet included in the database. Overall, the predictive value using ANNs was 78% for binding to HLA-A2 and 88% for . The type of data representation used had little effect on prediction. As few as 100 training cycles were sufficient, and the network appeared resistant to over-training effects after as many as 2000 cycles.

Full Text

Multimedia Links
(none)

Reference Links
(none)

Citation Reference

Get viewers
for PS & PDF

[CI Editor] [Site Manager]