ASH (Alignment of Structural Homologs) is a structure alignment program that has been developed for PDBj. Earlier versions of ASH were written by Hiroyuki Toh. The core alignment algorithm based on the double dynamic programming algorithm of Orengo and Taylor was then combined with a new scoring function based on NER (Number of Equivalent Residues) by Daron M. Standley[3,4]. The ASH web page serves as a central portal where both web services and source code can be accessed.
The Toh Lab at Kyushu University provides tools described below:
The ASH similarity score uses several heuristics to determine if two structures belong to the same structural domain. The derivation of the ASH score has been described in detail in an open-access article, so we will only give an overview here.
The ASH similarity score consists of five terms:
1. The Number of Equivalent Residues ( NER ) is a reward for C-alpha atoms that can be superimposed in space
2. A sequence similarity term that is derived from a large set of sequence-unique domains aligned by ASH
3. The C-alpha RMSD scored by the number of aligned residues (also known as the SAS score) is a penalty for C-alpha atoms that can not be superimposed
4. A gap penalty of non-terminal gaps
5. An overall difference term that does not depend on the alignment or superposition but instead looks at the secondary structure content, length, size, and contact order of the two structures
The weights of these five terms has been optimized for recognizing domains that belong to the same SCOP fold or CATH topology.
RASH (Rapid ASH) computes pair-wise alignments in under 1 second. RASH is generally sufficient if the structures are very similar.
GASH (Genetic-algorithm ASH) generates very accurate alignments in about twice the time as RASH. GASH is the best choice if the structures are only partially similar or if multiple solutions are required.