Improving the performance of DomainParser for structural domain partition using neural network

Abstract

Structural domains are considered as the basic units of protein folding, evolution, function and design. Automatic decomposition of protein structures into structural domains, though after many years of investigation, remains a challenging and unsolved problem. Manual inspection still plays a key role in domain decomposition of a protein structure. We have previously developed a computer program, DomainParser, using network flow algorithms. The algorithm partitions a protein structure into domains accurately when the number of domains to be partitioned is known. However the performance drops when this number is unclear (the overall performance is 74.5% over a set of 1317 protein chains). Through utilization of various types of structural information including hydrophobic moment profile, we have developed an effective method for assessing the most probable number of domains a structure may have. The core of this method is a neural network, which is trained to discriminate correctly partitioned domains from incorrectly partitioned domains. When compared with the manual decomposition results given in the SCOP database, our new algorithm achieves higher decomposition accuracy (81.9%) on the same data set.

Protein contacts, inter-residue interactions and side-chain modelling

“…Hence, methods for phylogenetic analyses and protein modelling usually perform better for single domains [58]. Automatic domain parsing generally makes the assumption that interdomain interaction (under a correct domain assignment) is weaker than the intradomain interaction (PUU [59], DOMAK [60] and 3Dee [61,62], DETECTIVE [63], DALI [64], STRUDL [65], DomainParser [66,67], Protein Domain Parser [68] and DDOMAIN [69]). These approaches maximize the number of contacts within a domain.…”

DDOMAIN: Dividing structures into domains using a normalized domain–domain interaction profile

“…(5) The DomainParser set. We downloaded the test data set used by DomainParser 2 (Guo et al 2003). It contains 1317 chains whose domains were assigned by SCOP 1.59.…”

“…Dividing a protein into domains is useful for more accurate function and structure determination because the methods for phylogenetic analysis and protein modeling usually work best for single domains (Ponting and Russell 2002). Both sequencebased (Chivian et al 2003;Heger and Holm 2003;Linding et al 2003;Bateman et al 2004;Letunic et al 2004;Liu and Rost 2004;Nagarajan and Yona 2004;von Ohsen et al 2004;Sim et al 2005) and structure-based methods (Holm and Sander 1994;Siddiqui and Barton 1995;Sowdhamini and Blundell 1995;Swindells 1995;Holm and Sander 1998;Taylor 1999;Xu et al 2000;Alexandrov and Shindyalov 2003;Guo et al 2003;Kundu et al 2004) have been developed for identification of protein domains.The problem of dividing a protein structure into domains is not yet solved. This is in part due to the lack of an unambiguous definition of domains.…”

“…Examples are PUU (Holm and Sander 1994), DOMARK (Siddiqui and Barton 1995), DETECTIVE (Swindells 1995), DALI (Holm and Sander 1998), STRUDL (Wernisch et al 1999), DomainParser (Xu et al 2000;Guo et al 2003), and protein domain parser (PDP) (Alexandrov and Shindyalov 2003). Others (Sowdhamini and Blundell 1995;Pugalenthi et al 2005) define domains by clustering substructures based on their spatial distances.…”

Analysis of protein contacts into Protein Units

“…Methods for phylogenetic analyses or protein modeling usually perform best for single domains [29]. The commonly used principle for automatic domain parsing is that interdomain interaction under a correct domain assignment is weaker than the intradomain interaction (PUU [30], DOMAK [31], 3Dee [32,33], DETECTIVE [34], DALI [35], STRUDL [36], DomainParser [37,38], Protein Domain Parser [39] and DDOMAIN [40]). Innovative approaches have been used in this context, e.g., graph theory [41] and Normal Mode Analysis approach [42].…”