Engineering Letter, 16:4, EL_16_4_07 ____________________________________________________________________________________ Prediction of Protein–Protein Interactions Using Pairwise Alignment and Inter-Domain Linker Region Nazar Zaki Abstract— One of the central problems in modern biology is to identify the complete set of interactions among the proteins in a cell. The structural interaction of proteins and their domains in networks is one of the most basic molecular mechanisms for biological cells. Structural evidence indicates that, interacting pairs of close homologs usually interact in the same way. In this article, we make use of both homology and inter-domain linker region knowledge to predict interaction between protein pairs solely by amino acid sequence information. High quality core set of 150 yeast proteins obtained from the Database of Interacting Proteins (DIP) was considered to test the accuracy of the proposed method. The strongest prediction of the method reached over 70% accuracy. These results show great potential for the proposed method. Index Terms—Protein-protein interaction, pairwise alignment, protein domain, inter-domain linker regions. I. INTRODUCTION A. The importance of protein–protein interactions The term protein-protein interaction refers to the association of protein molecules and the study of these associations from the perspective of biochemistry, signal transduction and networks. Protein-protein interactions occur at almost every level of cell function, in the structure of sub-cellular organelles, the transport machinery across the various biological membranes, the packaging of chromatin, the network of sub-membrane filaments, muscle contraction, signal transduction, and regulation of gene expression, to name a few [1]. Abnormal protein-protein interactions have implications in a number of neurological disorders; include Creutzfeld-Jacob and Alzheimer's diseases. Because of the importance of protein-protein interactions in cell development and disease, the topic has been studied extensively for many years and a large number of approaches to detect protein-protein interactions have been developed. Each of these approaches has strengths and weaknesses, especially with regard to the sensitivity and specificity of the method. B. Current methods to predict protein–protein interactions One of the major goals in functional genomics is to Manuscript received October 13, 2008. U.S. Nazar Zaki is an Assistant Professor with the College of Information Technology, United Arab Emirates University. Al-Ain 17555 UAE, (fax: +971-3-7626309; e-mail: nzaki@uaeu.ac.ae). determine protein interaction networks for whole organisms, and many of the experimental methods have been applied to study this problem. Co-immunoprecipitation is considered to be the gold standard assay for protein-protein interactions, especially when it is performed with endogenous proteins [2]. The protein of interest is isolated with a specific antibody. Interaction partners that stick to this protein are subsequently identified by western blotting. Interactions detected by this approach are considered to be real, but this method can only verify interactions between suspected interaction partners. Thus, this method is not a screening approach to identify unknown protein-protein interactions. The yeast two-hybrid screen investigates the interaction between artificial fusion proteins inside the nucleus of yeast [3]. This approach can identify binding partners of a protein in an unbiased manner, but this method suffers from high false-positive rate which makes it necessary to verify the identified interactions by co-immunoprecipitation. Tandem Affinity Purification (TAP) detects interactions within the correct cellular environment [4], which is a big advantage over the yeast two-hybrid approach. However, the TAP tag method requires two successive steps of protein purification, and thus this method cannot readily detect transient protein-protein interactions. This method is not an efficient means to detect physical protein-protein interactions that exist in different cellular environments either, which is especially important when studying the interaction network in an organism’s genome a very significant in the post-genomic era. Quantitative immunoprecipitation combined with knock-down (QUICK) relies on co-immunoprecipitation, quantitative mass spectrometry (SILAC) and RNA interference (RNAi) was introduced. This method detects interactions among endogenous non-tagged proteins [5], and thus this method’s results have the same high confidence as co-immunoprecipitation. However, this method also depends on the availability of suitable antibodies. These high-throughput methods have contributed tremendously in the creation of databases containing large sets of protein interactions, such as Database of Interacting Proteins (DIP) [6], MIPS [7] (developed at the Martinsried Institute for Protein Sequences) and Human Protein Reference Database (HPRD) [8]. In addition, several in silico methods have been developed to predict protein–protein interactions based on features such as gene context [9]. These include gene fusion [10], gene neighborhood [11] and phylogenetic profiles [12]. However, most of the in silico (Advance online publication: 20 November 2008) Engineering Letter, 16:4, EL_16_4_07 ____________________________________________________________________________________ methods seek to predict functional association, which often implies but is not restricted to physical binding. Despite the availability of the mentioned methods of predicting protein-protein interaction, the accuracy and coverage of these techniques have proven to be limited. Computational approaches remain essential both to assist in the design and validation of the experimental studies and for the prediction of interaction partners and detailed structures of protein complexes [13]. C. Computational approaches to predict protein-protein interaction Some of the earliest techniques predict interacting proteins through the similarity of expression profiles [14], coordination of occurrence of gene products in genomes, description of similarity of phylogenetic profiles [12] or trees [15], and studying the patterns of domain fusion [16]. However, it has been noted that these methods predict protein–protein interactions in a general sense, meaning joint involvement in a certain biological process, and not necessarily actual physical interaction [17]. Most of the recent works focus on employing the protein domain knowledge to predict the protein-protein interaction [18]-[22]. The motivation for this choice is that molecular interactions are typically mediated by a great variety of interacting domains [23]. It is thus logical to assume that the patterns of domain occurrence in interacting proteins provide useful information for training protein-protein interaction prediction methods [24]. An emerging new approach in the protein interactions field is to take advantage of structural information to predict physical binding [25]-[26]. Although the total number of complexes of known structure is relatively small, it is possible to expand this set by considering evolutionary relationships between proteins. It has been shown that in most cases close homologs (>30% sequence identity) physically interact in the same way with each other. However, conservation of a particular interaction depends on the conservation of the interface between interacting partners [27]. In this paper, we propose to predict protein-protein interaction using only sequence information. The proposed method combines evolutionary and structural relationships between protein pair to predict the interaction between them. Evolutionary relationships will be incorporated by measuring the similarity between protein pair using Pairwise Alignment. Structural relationships will be incorporated in terms of protein domain knowledge. We are encouraged by the fact that compositions of contacting residues in protein sequence are unique, and that incorporating evolutionary and predicted structural information improves the prediction of protein–protein interactions [28]. II. METHOD In this paper, we present a simple yet effective method to predict protein-protein interaction solely by amino acid sequence information. Figure 1, illustrates the overview of the proposed method. It consists of three main steps: (a) extracting the evolutionary relationships by measuring regions of similarity that may reflect functional, structural or evolutionary relationships between protein sequences (b) downsize the protein sequences of interest by predicting and eliminating inter-domain linker regions (c) scanning and detecting domain matches in all the protein sequences of interest. Two proteins may interact if they share similar domains. Fig. 1. Overview of the proposed method. A. Similarity Measures between Protein Sequences Using Pairwise Alignment One of the most fundamental tools in the field of bioinformatics is sequence alignment. By aligning sequences to one another, it is possible to evaluate how similar the sequences are and identify conserved regions in sets of related sequences. Therefore, our proposed method starts by measuring the protein-protein interaction sequence similarity, which reflects the evolutionary and homology relationships. Two protein sequences may interact by the mean of the amino acid similarities they contain [24]. This work is motivated by the observation that the Smith-Waterman (SW), algorithm (Advance online publication: 20 November 2008) Engineering Letter, 16:4, EL_16_4_07 ____________________________________________________________________________________ [29], which measures the similarity score between two sequences by a local gapped alignment, provides a relevant measure of similarity between protein sequences. This similarity incorporates biological knowledge about protein evolutionary structural relationships [30]. The Smith-Waterman similarity score SW ( x1 , x2 ) between two protein sequences Then the SW similarity score matrix Matrix _ a will be calculated as: x1 and x2 is the score of the best local alignment with gaps between the two protein sequences computed by the Smith-Waterman dynamic programming algorithm. Let us denote by µ a possible local alignment between x1 and x2 , defined by a number n of aligned residues, and by the indices 1 ≤ i1 < ... < in ≤ x1 and 1 ≤ j1 < ... < jn ≤ x2 of the aligned residues in x1 and x2 ∏( x1 , x2 ) the set of all possible local alignments between x1 and x2 , and by p ( x1 , x2 , µ ) the score of the local alignment µ ∈ ∏( x1 , x2 ) between x1 and x2 , the Smith-Waterman score SW ( x1 , x2 ) between sequences x1 and x2 can be respectively. Let us also denote by written as: SW ( x1 , x2 ) = max p ( x1 , x2 , µ ) (1) µ ∈∏ ( x1 , x 2 ) The similarity matrix can be calculated as follow: SW(x1, x1) SW(x1, x2) SW(x2, x1) SW(x2, x2) Matrix_a = : : ... ... : SW(x1, xm) SW(x2, xm) : SW(xm, x1) SW(xm, x2) ... SW(xm, xm) (2) where m is the number of the protein sequences. For example, suppose we have the following randomly selected protein-protein interaction dataset: YDR190C, YPL235W, YDR441C, YML022W, YLL059C, YML011C, YGR281W and YPR021C represented by x1 , x2 , x3 , x4 , x5 , x6 , x7 and x8 respectively. The From Matrix _ a , higher score may reflect interaction between two proteins. SW ( x1 , x2 ) and SW ( x2 , x1 ) scores are equal to 465; SW ( x3 , x4 ) and SW ( x4 , x3 ) scores are equal to 553, which confirm the interaction possibility. However, SW ( x5 , x6 ) and SW ( x6 , x5 ) scores are equal to 24; SW ( x7 , x8 ) and SW ( x8 , x7 ) scores are equal to 36, which are not the highest scores. To correct these errors more biological information is needed, which lead us to the second part of our method. B. Identify and Eliminating Inter-domain Linker Regions The results could be further enhanced by incorporating inter-domain linker regions knowledge. The next step of our algorithm is to predict inter-domain linker regions solely by amino acid sequence information. Our intention here is to identify and eliminate all the inter-domain linker regions from the protein sequences of interest. By doing this step, we are actually downsizing the protein sequence to shorter ones with only domains, which may produce better alignment scores. In this case, the prediction is made by using linker index deduced from a data set of domain/linker segments from SWISS-PROT database [31]. DomCut developed by Suyama et al [32] is employed to predict linker regions among functional domains based on the difference in amino acid composition between domain and linker regions. Following [32], we defined the linker index Si for amino acid residue i and it is calculated as follows: Lin ker Si = − ln fi Domain fi (3) interaction between these 8 proteins is shown in Figure 2. Where f i Lin ker is the frequency of amino acid residue the linker region and f i Domain i in is the frequency of amino acid residue i in the domain region. The negative value of Fig. 2. The interaction between the randomly selected proteins. Si means that the amino acid preferably exists in a linker region. A threshold value is needed to separate linker regions as shown in Figure 3. Amino acids with linker score greater than the set threshold value will be eliminated from the protein sequence of interest. (Advance online publication: 20 November 2008) Engineering Letter, 16:4, EL_16_4_07 ____________________________________________________________________________________ Fig. 3. An example of linker preference profile generated using Domcut. In this case, linker regions greater than the threshold value 0.093 will be eliminated from the protein sequence. When applying the second part of the method, the matrix Matrix _ a will be calculated as follows: several patterns, rules and profiles from PROSITE against our protein sequences of interest. Running ps_scan through the 8 proteins identifies the following Domains: YDR441C ( x 3 ) → PS00103 YML022W ( x 4 ) → PS00103 YGR281W ( x7 ) → PS00211, PS50893 and PS50929 YPR021C ( x 8 ) → PS50929 Which reveals structural relationships between proteins x3 x7 and x8 . This step is illustrated in Figure 4. Based on this relationship, SW ( x3 , x 4 ) and SW ( x7 , x8 ) will be calculated as follow: and x4 ; and proteins From Matrix _ a , its clearly noted that, more evidence is shown to confirm the interaction possibility between proteins x7 and x8 , and therefore, the result is furthermore enhanced. SW ( x3 , x4 ) = SW ( x3 , x4 ) + k * 300 (4) In the following part of the method, protein domain knowledge will be incorporated in Matrix _ a for better accuracy. SW ( x7 , x8 ) = SW ( x7 , x8 ) + k * 300 (5) C. Detecting Domain Matches and Associated Structural Relationships in Proteins In this part of the method, protein domains knowledge will be incorporated in Matrix _ a . Protein domains are highly informative for predicting protein-protein interaction as it reflects the potential structural relationships between them. In this implementation, we employed ps_scan [33] to scan one or Where k is the number of Domains the two interacting proteins share. Unfortunately, these results have not added more accuracy in this case, however, it confirmed the interacting possibilities between proteins x3 and x4 ; x7 and x8 . Fig. 4. An example of domain PS00103 found in proteins YDR441C and YML022W. (Advance online publication: 20 November 2008) Engineering Letter, 16:4, EL_16_4_07 ____________________________________________________________________________________ III. EXPERIMENTAL WORK To test our method, we obtained the protein-protein interaction data from the Database of Interacting Proteins (DIP). The DIP database catalogs experimentally determined interactions between proteins. It combines information from a variety of sources to create a single, consistent set of protein-protein interactions in Saccharomyces cerevisiae. The data stored within the DIP database were curated, both, manually by expert curators and also automatically using computational approaches that utilize the knowledge about the protein-protein interaction networks. This knowledge is extracted from the most reliable, core subset of the DIP data [34]. The DIP version we used contains 4749 proteins involved in 15675 interactions for which there is domain information [6]. However, only high quality core set of 2609 yeast proteins was considered in this experimental work. This core set is involved in 6355 interactions, which have been determined by at least one small-scale experiment or two independent experiments [35]. Furthermore, we selected proteins interacts with only one protein and not involved in any other interactions. This process results in a dataset of 150 proteins with 75 positive interactions as shown in Figure 5. The intention here is to design a method capable of predicting protein interaction partner, which facilitate a way to construct protein-protein interaction using only protein sequences information. We started our experimental work by measuring the protein-protein sequence interaction similarity using Smith-Waterman algorithm as implemented in FASTA [36]. The default parameters are used: gap opening penalty and extension penalties of 13 and 1, respectively, and a substitution matrix BLOSUM62 matrix. Various types of substitution matrices have been used over the years. Some were based on theoretical considerations, however, the most successful, based on analysis of alignments of numerous homologs of well-studies proteins from many different species [37]. The choice of which substitution matrix to use is not trivial because there is no one correct scoring scheme for all circumstances. The BLOSUM matrix is another very common used amino acid substitution matrix that depends on data from actual substitutions. This procedure produces the matrix Matrix _ a150 X 150 . This matrix was then enhanced by incorporating inter-domain linker regions information. In this case, only well defined domains with sequence length ranging from 50 to 500 residues were considered. We skipped all the frequently matching (unspecific) domains. A trashed value of 0.093 is used to separate the linker regions. Any residue generates an index greater than the threshold value results in eliminating it. This procedure downsized the protein sequences without losing the biological information. In fact, running the SW algorithm on a sequence having pure domains, results in better accuracy. A linker preference profile is generated using the linker index values along an amino acid sequence using a siding window. A window of size w = 15 is used because it gives the best performance. Further more, protein domains knowledge will be incorporated in Matrix _ a150 X 150 . In this implementation, ps_scan [33] is used to scan one or several patterns, rules and profiles from PROSITE against the 150 protein sequences. All frequently matching (unspecific) patterns and profiles are skipped. The ps_scan requires two compiled external programs from the PFTOOLS package: "pfscan" used to scan a sequence against a profile library and "psa2msa" which is necessary for the "-o msa" output format only. The Matrix _ a150 X 150 is then used to predict the protein interaction network. Two proteins may interact if the similarity score between them is the highest. Fig. 5. Dataset of core interaction proteins used in the experimental work. IV. RESULTS AND DISCUSSION The performance of the proposed method is measured by how well it can predict the protein-protein interaction network. Prediction accuracy, whose value is the ratio of the number of correctly predicted interactions between protein pairs to the total number of interactions and non-interactions possibilities in network, is the best index for evaluating the performance of a predictor. However, approximately 20% of the data are truly interacting proteins, which leads to a rather unbalanced distribution of interacting and non-interacting cases. This concept is shown in Figure 6. Suppose we have 6 proteins, then the interacting pairs are 1→2, 3→4 and 5→6, which result in 3 interactions cases (in red) out of a total of 15 interaction possibilities (12 non-interaction cases in blue). Fig. 6. Distribution of interacting and non-interacting cases. To assess our method objectively, another two indices are introduced in this paper, namely specificity and sensitivity commonly used in the evaluation of information retrieval. A high sensitivity means that many of the interactions that occur in reality are detected by the method. A high specificity (Advance online publication: 20 November 2008) Engineering Letter, 16:4, EL_16_4_07 ____________________________________________________________________________________ indicates that most of the interactions detected by the screen are also occurring in reality Sensitivity and specificity are combined measures of true positive ( tp ), true negative ( tn ), false positive ( expressed as: Where, tp = interacting two protein sequences classified interacting, fn = non-interacting two protein sequences classified interacting, fp ) and false negative ( fn ) and can be fp = interacting two protein sequences classified non-interacting, tn = non- interacting two protein sequences classified non-interacting. tp Sensitivity (Sens) = tp + fn tn Specificity (Spec) = tn + fp Based on the above mentioned performance measures, our algorithm was able to achieve encouraging results. In Figures 7 and 8, we summarized the sensitivity and specificity results based on the three stages of the method. The figures clearly show improvement in sensitivity but not much in specificity and that’s because of the big number of non-interacting possibilities. Sensitivity Accuracy (100%) Accuracy (100%) Specificity 0.703 0.7025 0.702 0.7015 0.701 0.7005 0.7 0.6995 0.699 0.7026 0.7026 0.6992 0.61 0.6 0.59 0.58 0.57 0.56 0.55 0.54 0.6067 0.5733 0.5467 Using only SW Eliminating inter- Adding protein Algorithm domain linker domains regions evidence Using only SW Eliminating inter- Adding protein Algorithm domain linker domains evidence regions The 3 steps of the method The 3 steps of the method (a) (b) Accuracy (100%) Fig. 7. Sensitivity and specificity results. 70.3 70.2 70.1 70 69.9 69.8 69.7 69.6 70.19 70.17 69.82 Using only SW Algorithm Eliminating inter- Adding protein domain linker domains evidence regions The 3 steps of the method Fig. 8. Overall accuracy. The overall performance evaluation results are summarized in Table 1. Table 1: Overall performance evaluation Similarity Measure Using SW Algorithm Eliminated Inter-domain Linker Regions Adding Structural Domain Evidence tp fp tn fn Sens Spec RFP* Accu 82 86 91 68 64 59 15523 15597 15597 6677 6603 6603 0.5467 0.5733 0.6067 0.6992 0.7026 0.7026 0.3008 0.2974 0.2974 60.82 70.17 70.19 *Rate of False Positive (RFP), which defined as the fraction of negative test sequences that score as high as or better than the positive sequence fp for ( fp + tn ) > 0 [38]. RFP = ( fp + tn) (Advance online publication: 20 November 2008) Engineering Letter, 16:4, EL_16_4_07 ____________________________________________________________________________________ V. CONCLUSION In this article we make use of both evolutionary and structural similarities among domains of known interacting proteins to predict putative protein interaction pairs. When tested on a sample data obtained from the Database of Interacting Proteins (DIP), the proposed method shows great potential and a new vision to predict protein-protein interaction. It proves that the combination of methods predicts domain boundaries or linker regions from different aspects and the evolutionary relationships would improve accuracy and reliability of the prediction as a whole. However, it is difficult to directly compare the accuracy of our proposed method because all of the other existing methods use different criteria for assessing the predictive power. Moreover, these existing methods use completely different characteristics in the prediction. One of the immediate future works is to consider the entire protein-protein interaction network and not to restrict our work on binary protein-protein interaction. ACKNOWLEDGMENT The Author would like to thank Dr. M. Suyama for providing valuable information about DomCut method. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] I. Donaldson, J. Martin, B. Bruijn, B. Wolting, C. Lay, V. Tuekam, B. Zhang, S. Baskin, B. Bader, G. D. Michalickova, K. Pawson, T. and C. W. Hogue, “PreBIND and Textomy--mining the biomedical literature for protein-protein interactions using a support vector machine”, BMC Bioinformatics, vol. 4, 2003. E. Gharakhanian, J. Takahashi, J. Clever, and H. Kasamatsu, “In vitro Assay for Protein-Protein Interaction: Carboxyl-Terminal 40 Residues of Simian Virus 40 Structural Protein VP3 Contain a Determinant for Interaction with VP1”, PNAS, vol. 85(18), 1998, pp. 6607-6611. P. L. Bartel and S. Fields, The yeast two-hybrid system. In Advances in Molecular Biology, New York: Oxford University Press, 1997. G. Rigaut, A. Shevchenko, B. Rutz, M. Wilm, M. Mann, and B. Seraphin,”A generic protein purification method for protein complex characterization and proteome exploration”, Nature Biotechnology, vol. 17, 1999, pp. 1030-1032. M. Selbach, and M. Mann, “Protein interaction screening by quantitative immunoprecipitation combined with knockdown (QUICK)”, Nature Methods, vol. 3, 2006, pp. 981-983. L. Salwinski, C. S. Miller, A. J. Smith, F. K. Pettit, J. U. Bowie, D. Eisenberg, “The Database of Interacting Proteins: 2004 update”, Nucleic Acids Res., vol. 1, 2004, pp. 449-51. H. W. Mewes, “MIPS: analysis and annotation of proteins from whole genomes”, Nucleic Acids Res., vol. 32, 2004, pp. 41-44. S. Peri,”Human protein reference database as a discovery resource for proteomics”, Nucleic Acids Res., vol. 32, 2004, pp. 497–501. J. Espadaler, “Detecting remotely related proteins by their interactions and sequence similarity”, Proc. Natl Acad. Sci. USA, vol. 102, 2005, pp. 7151–7156. E. Marcotte, “Detecting protein function and protein–protein interactions from genome sequences”, Science, vol. 285, 1999, pp. 751–753. T. Dandekar, ”Conservation of gene order: a fingerprint of proteins that physically interact”, Trends Biochem. Sci., vol. 23, 1998, pp. 324–328. M. Pellegrini, E. M. Marcotte, M. J. Thompson, D. Eisenberg, and T. O. Yeates, “Assigning protein functions by comparative genome analysis: protein phylogenetic profiles,” In the proceedings of National Academy of Sciences, USA, vol. 96, 1999, pp. 4285–4288. A. Szilàgyi, V. Grimm, A. K Arakaki and J. Sholnick, “Prediction of physical protein-protein interactions”, Phys. Biol. 2005, pp. 1-16. E. M. Marcotte, M. Pellegrini, M. J. Thompson, T. O. Yeates, and D. Eisenberg, “A combined algorithm for genome-wide prediction of protein function,” Nature, vol. 402, 1999, pp. 83–86. [15] F. Pazos and A. Valencia, “Similarity of phylogenetic trees as indicator of protein-protein interaction,” Protein Engineering, vol. 14, 2001, pp. 609- 614. [16] J. Enright, I. N. Ilipoulos, C. Kyrpides, and C. A. Ouzounis, “Protein interaction maps for complete genomes based on gene fusion events,” Nature, vol. 402, 1999, pp. 86–90. [17] D. Eisenberg, E. M. Marcotte, I. Xenarios, and T. O. Yeates, “Protein function in the post-genomic era,” Nature, vol. 405, 2000, pp. 823-826. [18] J. Wojcik and V. Schachter, “Protein-Protein interaction map inference using interacting domain profile pairs,” Bioinformatics, vol. 17, 2001, pp. 296-305. [19] W. K. Kim, J. Park, and J. K. Suh, “Large scale statistical prediction of protein-protein interaction by potentially interacting domain (PID) pair,” Genome Informatics, vol. 13, 2002, pp. 42-50. [20] S. K. Ng, Z. Zhang, and S. H. Tan, “integrative approach for computationally inferring protein domain interactions,” Bioinformatics, 19, 2002, pp. 923-929. [21] S. M. Gomez, W. S. Noble, and A. Rzhetsky, “Learning to predict protein-protein interactions from protein sequences,” Bioinformatics, 19, 2003, pp. 1875-1881. [22] C. Huang, F. Morcos, S. P. Kanaan, S. Wuchty, A. Z. Chen, and J. A. Izaguirre, “Predicting Protein-Protein Interactions from Protein Domains Using a Set Cover Approach”, IEEE/ACM Transactions on Computational Biology and Bioinformatics, vol. 4, 2007. [23] T. Pawson and P. Nash, “Assembly of cell regulatory systems through protein interaction domains,” Science, vol. 300, 2003, pp. 445-452. [24] N. M. Zaki, S. Deris and H. Alashwal, “Protein Protein Interaction Detection Based on Substring Sensitivity Measure", International J. of Biomedical Sci., vol. 1, 2006, pp. 148-154. [25] P. Aloy and R. B. Russell, “InterPreTS: protein interaction prediction through tertiary structure”, Bioinformatics, vol. 19, 2003, pp. 161–162. [26] L. Lu, “Multiprospector: an algorithm for the prediction of protein–protein interactions by multimeric threading”, Proteins, vol. 49, 2002, pp. 350–364. [27] J. Espadaler, O. Romero-Isart, R. M. Jackson and B. Oliva1 "Prediction of protein–protein interactions using distant conservation of sequence patterns and structure relationships", Bioinformatics, vol. 21, 2005, pp. 3360–3368. [28] O. Keskin, “A new, structurally nonredundant, diverse data set of protein–protein interfaces and its implications”, Protein Sci., vol. 13, 2004, pp. 1043–1055. [29] T. Smith and M. Waterman, “Identification of common molecular subsequences”, J. Mol. Bio., vol. 147, 1981, pp. 195-197. [30] H. Saigo, J. Vert, N. Ueda and T. Akutsu, “Protein homology detection using string alignment kernels,” Bioinformatics, Vol. 20 no. 11, 2004, pp. 1682-1689. [31] A. Bairoch and R. Apweiler, “The SWISS-PROT protein sequence database and its supplement TrEMBL in 2000”, Nucleic Acids Res., 28, 2000, pp. 45–48. [32] M. Suyama and O. Ohara, "DomCut: prediction of inter-domain linker regions in amino acid sequences", Bioinformatics, vol 19, 2003, pp. 673-674. [33] A. Gattiker, E. Gasteiger, A. Bairoch, “ScanProsite: a reference implementation of a PROSITE scanning tool”, Applied Bioinformatics, vol. 1, 2002, pp. 107-108. [34] I. Xenarios, L. Salwínski, X. J. Duan, P. Higney, S. Kim and D. Eisenberg, “DIP, the Database of Interacting Proteins: a research tool for studying cellular networks of protein interactions” , Nucleic Acids Research, Oxford University Press, vol. 30, 2002, pp. 303-305. [35] C. M. Deane, L. Salwinski, I. Xenarios, and D. Eisenberg, “Protein interactions: two methods for assessment of the reliability of high throughput observations,” Molecular & Cellular Proteomics, vol. 1, 2002, pp. 349-56. [36] W. R. Pearson, “Rapid and sensitive sequence comparisons with FASTAP and FASTA Method”, Enzymol, Vol. 183, 1985, pp. 63. [37] Z. Marketa, and O. B. Jeremy, “Understanding bioinformatics”, Garland Science, Taylor & Francis Group, LLC, 2008. [38] N. M. Zaki, S. Deris, and R. M. Illias, “Application of string kernels in protein sequence classification”, Applied Bioinformatics, vol. 4, 2005, pp. 45-52. (Advance online publication: 20 November 2008)