Mathematical filtering using wavelet transform was applied to analyse
hydropathy signals of membrane proteins. The accuracy of our
localization of transmembrane approaches that of well-established methods.
The analysis of hydrophobicity plots using wavelets presents advantages
with respect to other "filtering" methods based on fixed windows or
Fourier transforms and compared to "training" (neutral networks)
techniques. Although the method embodies principles that have long been
appreciated, its simplicity makes it a very useful tool for the evaluation
of protein membrane-spanning segments. Generalization of use of wavelets
should be encouraged in other aspects of bioinformatics.