Determination of the native folded structure for a particular protein is a milestone towards understanding its function, and in most cases, can be done experimentally. However, the ability to predict in silico protein structure and related features would represent a fundamental breakthough in structural biology. The ability to predict domains in proteins is amongst the most important tasks needed for effective functional classification, homology-based structure prediction, structural genomics, as it makes function prediction easier. In this paper, we present the DomAnS, protein domain prediction approach, that is based on pattern alignment. DomAnS allows rapid screening for potential domain regions with the ability to recognize the most promising regions where domains might exists. The combination of the DomAnS algorithm with specialized databases that contains all known domains, allows us to find domain regions without solving 3D structure. Our approach has been tested on CASP7 data, and for 28 targets gave the best overall score.

Keywords: protein domain prediction, domain patterns database, sequence analysis, sequence patterns