Generation and evaluation of dimension-reduced amino acid parameter representations by artificial neural networks

In order to process data of proteins, a numerical representation for an ami no acid is often necessary. Many suitable parameters can be derived from ex periments or statistical analysis of databases. To ensure a fast and effici ent use of these sources of information, a reduction and extraction of rele vant information out of these parameters is a basic need. In this approach established methods like principal component analysis (PCA) are supplemente d by a method based on symmetric neural networks. Two different parameter r epresentations of amino acids are reduced from five and seven dimensions, r espectively, to one, two, three, or four dimensions by using a symmetric ne ural network approach alternatively with one or three hidden layers. It is possible to create general reduced parameter representations for amino acid s. To demonstrate the ability of this approach, these reduced sets of param eters are applied for the ab initio prediction of protein secondary structu re from primary structure only. Artificial neural networks are implemented and trained with a diverse representation of 430 proteins out of the PDB. A n essentially faster training and also prediction without a decrease in acc uracy is obtained for the reduced parameter representations in comparison w ith the complete set of parameters. The method is transferable to other ami no acids or even other molecular building blocks, like nucleic acids, and t herefore represents a general approach.