Inspired by neuroscience and pattern recognition, a kernel is
constructed which gives a setting for learning theory in
bioinformatics. In particular this applies to the peptide binding
problem to give more accurate results than previous ones. The
protein folding problem is studied in the same framework.