- A-Z
- Endocytobiosis and ...
- Volume 11
- Issue 1
- Amino acid sequence...
- Autor(in)
- Seitenbereich
-
001 - 018
- Zusammenfsg.
-
The applicability of artificial neural networks as a fitness function for evolutionary protein design is discussed. In machina design of idealized signal peptidase cleavage-sites served as an example for this method in previous experiments. The results obtained and the design strategy selected are critically reviewed and evaluated. It is demonstrated that neural networks can extract relevant sequence information even from a small set of sequence data provided that the data are representative and appropriately encoded. The physicochemical properties hydrophobicity and side-chain volume showed to be useful for the description of signed peptide features. An encoding scheme for amino acids based on their individual preferences for main-chain folding angles is presented. It is motivated to employ this sequence description for extraction of protein structural features from the amino acid sequence. Furthermore, a computer-based technique for optimization of amino acid sequences termed 'simulated molecular evolution' is evaluated.