Protein and Peptide Letters - Volume 15, Issue 5, 2008

We are very pleased to offer this special issue to the readers of Protein and Peptide Letters by selecting the candidate papers from the 2007 International Conference on Intelligent Computing (ICIC), held in Qingdao, Shandong Province, China, 21-24 August 2007. Twelve papers (representing less than one half of one percent of all eligible papers accepted at the 2007 ICIC) were selected for inclusion in this special issue. In recent years, we have witnessed intelligent computing techniques, such as artificial intelligence, machine learning, colony intelligence, and others being dedicated to various research aspects of bioinformatics, neuroinformatics, chemoinformatics, computational biology, system biology, etc. Meanwhile, intelligent computing knowledge has been enriched by the development of more solid mathematical frameworks, elaborating more efficient and powerful algorithms as well as structures. More importantly, it has been driven by its application to many amazing research fields, such as bioinformatics. Currently, intelligent computing techniques for protein bioinformatics are being used to encode biological information, extract biological features, recognize biological patterns, mine and comprehend biological data, build models of biological systems and processes, and thus automatically form theories from the unprecedentedly vast experimental biological data. Its main objective is to find the rule and useful biological information from limited observation examples that cannot be obtained using classical biological methods and theories. It extends the rule to predict and infer protein’s structure, function, structurefunction relationship, interaction networks, and other significant aspects of protein and peptide science. Hence, the intelligent computing technique, an in silico method, is supplementary to conventional experimental methods and makes it possible to use computers to extract knowledge from large amounts of biological and biomedical information. This special issue includes several papers on how to use intelligent computing techniques to solve problems in protein bioinformatics. Three papers in this issue focus on exploring computational algorithms and applications in protein bioinformatics. Busa- Fekete et al. discuss using propagation on unrooted binary trees to perform protein classification. Zhang et al. address an antcolony algorithm for solving protein-folding problems by placing pheromones on the arcs connecting adjacent squares in the lattice. Peng et al. present an artificial immune network-based algorithm for diabetes diagnosis. In the next two papers, Srisawat et al. use combined classifiers for HIV drug-resistance prediction, and Liu, Xia et al. adopt ensemble schemes for protein secondary structure prediction. The next four papers present network or sequence multiple alignment-based protein structure and function predictions. Liu et al. discuss using a pocket similarity network to analyze protein surface patterns. Zhao et al. propose a framework of protein domain function annotation with predicted domain-domain interaction networks. Zhang et al. predict the binding motifs in hepatitis protein C virus NS5A and human proteins using sequence multiple alignment. Gao et al. present their computational phylogenetic analysis for the functional annotation of TBC (Tre-2/Bub2/Cdc16) domain-containing proteins. The last three papers focus on applying special coding methods, support vector machine (SVM) and its modified techniques to protein folding recognition, ligand binding, and long-range interaction site predictions, respectively. More specifically, Chen et al. combine error-correcting output codes and SVM methods for protein fold recognition. Cai et al. apply SVM to the function prediction of DNA-/RNA- binding proteins, GPCR, and drug ADME-associated proteins. Chen et al. present a bounded SVM method for locating key long-range interaction sites by predicted, local B-factors. It should be stressed that recommendations for this special issue were made by the ICIC's International Program Committee, and the final selections were made on the basis of quality, novelty, and theoretical or practical importance. All papers were subjected to two rounds of review with a minimum of three reviewers, reflecting the demand for improving the quality of ICIC papers. We hope that you find this special issue both useful and enjoyable. As guest editor, I would like to take this opportunity to thank all the authors for their contributions to this special issue, the reviewers for their valuable input, insight, and expert comments, and the Editor-in-Chief, Distinguished Professor Ben M. Dunn, for his valuable advice and strong support during the preparation of this special issue.

We present two efficient network propagation algorithms that operate on a binary tree, i.e., a sparse-edged substitute of an entire similarity network. TreeProp-N is based on passing increments between nodes while TreeProp-E employs propagation to the edges of the tree. Both algorithms improve protein classification efficiency.

This paper applies and studies the behavior of three learning algorithms, i.e. the Support Vector machine (SVM), the Radial Basis Function Network (the RBF network), and k-Nearest Neighbor (k-NN) for predicting HIV-1 drug resistance from genotype data. In addition, a new algorithm for classifier combination is proposed. The results of comparing the predictive performance of three learning algorithms show that, SVM yields the highest average accuracy, the RBF network gives the highest sensitivity, and k-NN yields the best in specificity. Finally, the comparison of the predictive performance of the composite classifier with three learning algorithms demonstrates that the proposed composite classifier provides the highest average accuracy.

A new approach based on the implementation of support vector machine (SVM) with the error correcting output codes (ECOC) is presented for recognition of multi-class protein folds. The experimental show that the proposed method can improve prediction accuracy by 4%-10% on two datasets containing 27 SCOP folds.

In this work, in order to reveal protein surface patterns in a systems biology framework, we analyze the similarity among the surface cavities by investigating the features of the pocket similarity network such as the community structure, the small-world property, the scale-free characteristic, and the hubs.

This paper presents a framework for annotating protein domains with predicted domain-domain interaction networks. Specially, domain annotation is formalized as a multi-class classification problem in this work. The numerical experiments on InterPro domains show promising results, which proves the efficiency of our proposed methods.

This paper explores the use of support vector machine (SVM) for protein function prediction. Studies are conducted on several groups of proteins with different functions including DNA-binding proteins, RNA-binding proteins, Gprotein coupled receptors, drug absorption proteins, drug metabolizing enzymes, drug distribution and excretion proteins. The computed accuracy for the prediction of these proteins is found to be in the range of 82.32% to 99.7%, which illustrates the potential of SVM in facilitating protein function prediction.

This paper proposes a flexible ant colony (FAC) algorithm for solving protein folding problems based on the hydrophobic-polar square lattice model. Collaborations of novel pheromone and heuristic strategies in the proposed algorithm make it more effective in predicting structures of proteins compared with other state-of-the-art algorithms.

In this paper, we adopted the bounded support vector machine to locate the key long-range interaction sites by the use of predicted local lowest B-factors. As a result, the key long-range interaction residues can be located based on information of local lowest B-factor sites.

A novel artificial immune network based algorithm for the diagnosis of diabetes is presented. The algorithm's implementation includes: (1) creating the initial immune antibody network; (2) the network is evolved with the learning from foreign antigens; (3) diagnosis process is accomplished by majority vote of the k nearest neighbor antibodies.

This paper proposes an efficient ensemble system to tackle the protein secondary structure prediction problem with neural networks as base classifiers. The experimental results show that the multi-layer system can lead to better results. When deploying more accurate classifiers, the higher accuracy of the ensemble system can be obtained.

From the extensive analysis, we identified three highly conserved sequence segments in HCV NS5A proteins and one binding motif in human proteins. The binding motif of human proteins often forms a full helix or an extended strand-loop structure, and is in good agreement with the experimental findings of previous studies.

Tre-2/Bub2/Cdc16 domain-containing proteins (TBC proteins) participate in wide range cellular processes. With computational approaches, 137 non-redundant TBC proteins from five model organisms were identified and classified into 13 subfamilies base on molecular evolutionary tree. This phylogenetic analysis provides useful functional annotation of newly-identified TBC proteins and guides for further experimentation.

We have examined the structure of two Humanin (HN) analog peptides, HNG and AGA-(C8R)HNG17, in the presence of sodium dodecylsulfate (SDS) and trifluoroethanol (TFE) using CD and sedimentation velocity. Both HNG and AGA-(C8R)HNG17 underwent complex conformational changes with increasing concentrations of SDS and TFE, in contrast to general trend of increasing α-helix with their concentration. To our surprise, both peptides appear to converge into a similar structure in SDS and TFE at higher concentrations; e.g., above 0.05 % SDS or 30-40 % TFE. Sedimentation velocity analysis showed extensive aggregation of HNG at 0.1 mg/ml in PBS in the absence of SDS, but a highly homogeneous solution in 0.1 % SDS, indicating formation of a uniform structure by SDS. These two peptides also formed an intermediate structure both in SDS and TFE at lower concentrations, which appeared to be associated with extensive aggregation. It is interesting that the structure changes of these peptides occur well below the critical micelle concentration of SDS, suggesting that conformational changes are mediated through molecular, not micellar, interactions with SDS.

Measurements of [θ]222 of E. coli phosphatase on heating from 20° to 90° and subsequent cooling to 20° shows a gradual increase in [θ]222 on heating, while cooling shows a symmetric transition centered at 45°. Reheating and cooling shows the same phenomenon. Enzyme heated and cooled once is fully active. The activity of the enzyme depends on its storage conditions (buffer and pH for example), but such changes are least to some extent reversible, especially by heating in different solvents. We conclude the enzyme exists in several forms which are in slow equilibrium with each other, so that the enzyme responds slowly when heated and hence is not at equilibrium during heating/cooling experiments.

Conformational and structural changes of lentil seedlings amine oxidase (LSAO) were studied in the presence of trifluoroethanol (TFE) by spectroscopic and analytical techniques. At TFE concentrations up to 5%, the induction of a structural transition from β-sheet to α-helix and up to 10% TFE a structural transition from α-helix to β-sheet as well as inactivation of the enzyme are observed. At TFE concentrations between 10-35%, LSAO proves to be prone to aggregation and beyond 35% TFE leads to a non-native protein structure with a high α-helix content. The obtained results revealed that the aggregation of LSAO is strongly linked to the nature of secondary structures.

In the present study, carbonic anhydrase (CA) enzyme was purified from rainbow trout (RT) liver with a specific activity of 4318 EUxmg-1 and a yield of 38% using Sepharose-4B-L tyrosine-sulfanilamide affinity gel chromatography. The overall purification was approximately 2260-fold. To check the purity and determine subunit molecular weight of enzyme, SDS-polyacrylamide gel electrophoresis was performed, which showed a single band and MW of approx. 29.4 kDa. The molecular weight of native enzyme was estimated to be approx. 31 kDa by Sephadex-G 200 gel filtration chromatography. Optimum and stable pH were determined as 9.0 in 1 M Tris-SO4 buffer and 8.5 in 1 M Tris-SO4 buffer at 4°C, respectively. The optimum temperature, activation energy (Ea), activation enthalpy (ΔH ) and Q10 from Arrhenius plot for the RT liver CA were 4°C, 2.88 kcal/mol, 2.288 kcal/mol and 1.53, respectively. The purified enzyme had an apparent Km and Vmax of 0.66 mM and 0.126 μmol x min-1 for 4-nitrophenylacetate, respectively. cat k of the CA was found to be 32.8 s-1. The inhibitory effects of low concentrations of different metals (Co(II), Cu(II), Zn(II) and Ag(I)) on CA activity were determined using the esterase method under in vitro conditions. The obtained IC50 values, 50% inhibition of in vitro enzyme activity, were 0.03 mM for cobalt, 30 mM for copper, 47.1 mM for zinc and 0.01 mM for silver. Ki values for these substances were also calculated from Linewaever-Burk plots as 0.050 mM for cobalt, 1.950 mM for copper, 7.035 mM for zinc and 2.190 mM for silver respectively and determined that cobalt and zinc inhibit the enzyme a competitive manner and copper and silver inhibit the enzyme in an uncompetitive manner.

Antimicrobial peptides (AMPs) are generally produced as precursor peptides containing a signal sequence, a pro-region and the mature peptide. A computational analysis of β-defensin precursors predicts cleavage solely by signal peptidase to release the mature peptide, with no pro-region. This supports the extensive transcriptional control of β- defensin expression compared with other AMP genes.

The N-acetylglutamate kinase from Streptococcus mutans was expressed in Escherichia coli in soluble form and purified to homogeneity. Crystals suitable for X-ray diffraction were obtained by hanging-drop vapor diffusion method and diffracted to 2.06 Å. The crystal belonged to space group P21212, with unit cell parameters a = 57.19 Å, b =94.76 Å, c =47.58 Å. The gel filtration and initial phasing results showed that the enzyme exists as a monomer, which is different from previously reported N-acetylglutamate kinases.