Introduction Amylases of varying specificity catalyse the degradation of starch and related oligoand poly-saccharides 111. These enzymes occur widely in plants, animals and micro-organisms and many, in particular those of bacterial or fungal origin, are used in the starch industry. Malt a-amylases are important in the production of beer and other alcoholic beverages. Despite the industrial demand for altered physical characteristics and biochemical behaviour, there are few reports of engineered amylases. Successful redesign of an enzyme generally requires (i) knowledge about the three-dimensional structure, preferably of an enzymesubstrate analogue complex, (ii) knowledge about the mechanism of action and (iii) access to the cloned cDNA and an expression system. Although the crystal structure is determined for three a-amylases [2-41, two cyclodextrin glucanotransferases [ 5, 61, and a /?-amylase [7], the catalytic and binding mechanisms of the enzymes remain to be described in molecular detail. Thus, conflicting identification of catalytic groups has arisen through interpretations based on the tertiary structures of the three different a-amylases [2-41. Site-specific mutation of these residues has not identified their respective functional roles, but rather indicated a more complicated mechanism of action than initially anticipated. A large number of amylolytic enzymes have been characterized in terms of industrially interesting properties such as thermostability, pH-activity dependence, and substrate specificity [ 11. The corresponding cDNAs and suitable expression systems are often available, but correlations with the enzyme structure that could form the basis for protein engineering have not yet been made. Despite the very low sequence similarity different types of amylases seem to adopt the same polypeptide-fold pattern [81 11. It may be possible, therefore, to design amylolytic enzymes with altered properties on the basis of sequence comparison using known properties of functional side-chains, the tertiary structure of a related enzyme, and the predicted secondary structure. Random mutagenesis is an alternative to site-directed mutagenesis and is a particularly attractive approach if a strong selection system is available. It does not rely on structural information, but comparison of the threedimensional structure, ideally of both wild-type and mutant enzymes, will help to interpret the effect of the side-chain substitutions and suggest further mutations. Random mutations thus mapped were evaluated by using a computer-generated model of the enzyme [ 121 and identified functionally important residues in Bacillus stearothemtophilus aamylase.