Parallel cascade identification as a means for automatically classifying protein sequences into structure/function groups

Abstract

Current methods for automatically classifying protein sequences into structure/function groups, based on their hydrophobicity profiles, have typically required large training sets. The most successful of these methods are based on hidden Markov models, but may require hundreds of exemplars for training in order to obtain consistent results. In this paper, we describe a new approach, based on nonlinear system identification, which appears to require little training data to achieve highly promising results.

Interpreting Microarray Data and Related Applications Using Nonlinear System Identification

“…To increase the accuracy of estimating the coefficients, the impulse response function h k ( j) can first be scaled so that the linear element's output u k (i) has unity mean square [40]. The new residual is then calculated from Eq.…”

“…For example, consider the problem of predicting the structure/function family of a novel protein given only its primary amino acid sequence. The amino acid sequences can be regarded as the inputs and their corresponding families as the outputs [40,51]. First, some means is used to map the amino acid sequence into a corresponding numerical sequence and similarly to numerically designate the families to be distinguished.…”

“…To distinguish between, say, the globin and calcium-binding families, the profiles of one or more exemplars from each family were spliced together to form a training input [40]. The corresponding training output was defined to have ideal value -1 over globin segments of the training input and 1 over calcium-binding segments.…”

“…A novel sequence was classified by feeding its corresponding profile through the identified model and then computing the average of the resulting output starting with the (R þ 1)th point. If the average was less than zero, the sequence was classified as globin and otherwise as calcium binding [40]. While effective classifiers were built encoding amino acids by the Rose scale, the resulting training inputs had some drawbacks for nonlinear system identification.…”

On the Advantages of Multi-Input Single-Output Parallel Cascade Classifiers

“…7 Parallel Cascade Identification (PCI), a method of nonlinear system identification, 6 has shown itself to be a powerful tool in the identification of such systems, which are often dynamic, complex, and have high order nonlinearities. Recent applications of PCI to bioinformatics include the prediction of protein coding potential from DNA sequences, 9 the identification of adenosine triphosphate (ATP) binding sites on proteins, 4 and the classification of protein sequences into structure/function families.…”