MOTIVATION: B-cell epitope is a small area on the surface of an antigen that binds to an antibody. Accurately locating epitopes is of critical importance for vaccine development. Compared with wet-lab methods, computational methods have strong potential for efficient and large-scale epitope prediction for antigen candidates at much lower cost. However, it is still not clear which features are good determinants for accurate epitope prediction, leading to the unsatisfactory performance of existing prediction methods. METHOD AND RESULTS: We propose a much more accurate B-cell epitope prediction method. Our method uses a new feature B factor (obtained from X-ray crystallography), combined with other basic physicochemical, statistical, evolutionary and structural features of each residue. These basic features are extended by a sequence window and a structure window. All these features are then learned by a two-stage random forest model to identify clusters of antigenic residues and to remove isolated outliers. Tested on a dataset of 55 epitopes from 45 tertiary structures, we prove that our method significantly outperforms all three existing structure-based epitope predictors. Following comprehensive analysis, it is found that features such as B factor, relative accessible surface area and protrusion index play an important role in characterizing B-cell epitopes. Our detailed case studies on an HIV antigen and an influenza antigen confirm that our second stage learning is effective for clustering true antigenic residues and for eliminating self-made prediction errors introduced by the first-stage learning. AVAILABILITY AND IMPLEMENTATION: Source codes are available on request.

BACKGROUND: Binding free energy and binding hot spots at protein-protein interfaces are two important research areas for understanding protein interactions. Computational methods have been developed previously for accurate prediction of binding free energy change upon mutation for interfacial residues. However, a large number of interrupted and unimportant atomic contacts are used in the training phase which caused accuracy loss. RESULTS: This work proposes a new method, βACVASA, to predict the change of binding free energy after alanine mutations. βACVASA integrates accessible surface area (ASA) and our newly defined β contacts together into an atomic contact vector (ACV). A β contact between two atoms is a direct contact without being interrupted by any other atom between them. A β contact's potential contribution to protein binding is also supposed to be inversely proportional to its ASA to follow the water exclusion hypothesis of binding hot spots. Tested on a dataset of 396 alanine mutations, our method is found to be superior in classification performance to many other methods, including Robetta, FoldX, HotPOINT, an ACV method of β contacts without ASA integration, and ACVASA methods (similar to βACVASA but based on distance-cutoff contacts). Based on our data analysis and results, we can draw conclusions that: (i) our method is powerful in the prediction of binding free energy change after alanine mutation; (ii) β contacts are better than distance-cutoff contacts for modeling the well-organized protein-binding interfaces; (iii) β contacts usually are only a small fraction number of the distance-based contacts; and (iv) water exclusion is a necessary condition for a residue to become a binding hot spot. CONCLUSIONS: βACVASA is designed using the advantages of both β contacts and water exclusion. It is an excellent tool to predict binding free energy changes and binding hot spots after alanine mutation.

BACKGROUND: Distinction between true protein interactions and crystal packing contacts is important for structural bioinformatics studies to respond to the need of accurate classification of the rapidly increasing protein structures. There are many unannotated crystal contacts and there also exist false annotations in this rapidly expanding volume of data. Previous tools have been proposed to address this problem. However, challenging issues still remain, such as low performance when the training and test data contain mixed interfaces having diverse sizes of contact areas. METHODS AND RESULTS: B factor is a measure to quantify the vibrational motion of an atom, a more relevant feature than interface size to characterize protein binding. We propose to use three features related to B factor for the classification between biological interfaces and crystal packing contacts. The first feature is the sum of the normalized B factors of the interfacial atoms in the contact area, the second is the average of the interfacial B factor per residue in the chain, and the third is the average number of interfacial atoms with a negative normalized B factor per residue in the chain. We investigate the distribution properties of these basic features and a compound feature on four datasets of biological binding and crystal packing, and on a protein binding-only dataset with known binding affinity. We also compare the cross-dataset classification performance of these features with existing methods and with a widely-used and the most effective feature interface area. The results demonstrate that our features outperform the interface area approach and the existing prediction methods remarkably for many tests on all of these datasets. CONCLUSIONS: The proposed B factor related features are more effective than interface area to distinguish crystal packing from biological binding interfaces. Our computational methods have a potential for large-scale and accurate identification of biological interactions from the experimentally determined structural data stored at PDB which may have diverse interface sizes.

BACKGROUND: Human triosephosphate isomerase (HsTIM) deficiency is a genetic disease caused often by the pathogenic mutation E104D. This mutation, located at the side of an abnormally large cluster of water in the inter-subunit interface, reduces the thermostability of the enzyme. Why and how these water molecules are directly related to the excessive thermolability of the mutant have not been investigated in structural biology. RESULTS: This work compares the structure of the E104D mutant with its wild type counterparts. It is found that the water topology in the dimer interface of HsTIM is atypical, having a "wet-core-dry-rim" distribution with 16 water molecules tightly packed in a small deep region surrounded by 22 residues including GLU104. These water molecules are co-conserved with their surrounding residues in non-archaeal TIMs (dimers) but not conserved across archaeal TIMs (tetramers), indicating their importance in preserving the overall quaternary structure. As the structural permutation induced by the mutation is not significant, we hypothesize that the excessive thermolability of the E104D mutant is attributed to the easy propagation of atoms' flexibility from the surface into the core via the large cluster of water. It is indeed found that the B factor increment in the wet region is higher than other regions, and, more importantly, the B factor increment in the wet region is maintained in the deeply buried core. Molecular dynamics simulations revealed that for the mutant structure at normal temperature, a clear increase of the root-mean-square deviation is observed for the wet region contacting with the large cluster of interfacial water. Such increase is not observed for other interfacial regions or the whole protein. This clearly suggests that, in the E104D mutant, the large water cluster is responsible for the subunit interface flexibility and overall thermolability, and it ultimately leads to the deficiency of this enzyme. CONCLUSIONS: Our study reveals that a large cluster of water buried in protein interfaces is fragile and high-maintenance, closely related to the structure, function and evolution of the whole protein.

MOTIVATION: Worldwide and substantial mortality caused by the 2009 H1N1 influenza A has stimulated a new surge of research on H1N1 viruses. An epitope conservation has been learned in the HA1 protein that allows antibodies to cross-neutralize both 1918 and 2009 H1N1. However, few works have thoroughly studied the binding hot spots in those two antigen-antibody interfaces which are responsible for the antibody cross-neutralization. RESULTS: We apply predictive methods to identify binding hot spots at the epitope sites of the HA1 proteins and at the paratope sites of the 2D1 antibody. We find that the six mutations at the HA1's epitope from 1918 to 2009 should not harm its binding to 2D1. Instead, the change of binding free energy on the whole exhibits an increased tendency after these mutations, making the binding stronger. This is consistent with the observation that the 1918 H1N1 neutralizing antibody can cross-react with 2009 H1N1. We identified three distinguished hot spot residues, including Lys(166), common between the two epitopes. These common hot spots again can explain why 2D1 cross-reacted. We believe that these hot spot residues are mutation candidates which may help H1N1 viruses to evade the immune system. We also identified eight residues at the paratope site of 2D1, five from its heavy chain and three from its light chain, that are predicted to be energetically important in the HA1 recognition. The identification of these hot spot residues and their structural analysis are potentially useful to fight against H1N1 viruses. CONTACT: [email protected] AVAILABILITY: Z-score is available at http://155.69.2.25/liuqian/indexz.py SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

This paper studies efficient mining of negative correlations that pace in collaboration. A collaborating negative correlation is a negative correlation between two sets of variables rather than traditionally between a pair of variables. It signifies a synchronized value rise or fall of all variables within one set whenever all variables in the other set go jointly at the opposite trend. The time complexity is exponential in mining. The high efficiency of our algorithm is attributed to two factors: (i) the transformation of the original data into a bipartite graph database, and (ii) the mining of transpose closures from a wide transactional database. Applying to a Yeast gene expression data, we evaluate, by using Pearson's correlation coefficient and P-value, the biological relevance of collaborating negative correlations as an example among many real-life domains.

BACKGROUND: A protein binding hot spot is a small cluster of residues tightly packed at the center of the interface between two interacting proteins. Though a hot spot constitutes a small fraction of the interface, it is vital to the stability of protein complexes. Recently, there are a series of hypotheses proposed to characterize binding hot spots, including the pioneering O-ring theory, the insightful 'coupling' and 'hot region' principle, and our 'double water exclusion' (DWE) hypothesis. As the perspective changes from the O-ring theory to the DWE hypothesis, we examine the physicochemical properties of the binding hot spots under the new hypothesis and compare with those under the O-ring theory. RESULTS: The requirements for a cluster of residues to form a hot spot under the DWE hypothesis can be mathematically satisfied by a biclique subgraph if a vertex is used to represent a residue, an edge to indicate a close distance between two residues, and a bipartite graph to represent a pair of interacting proteins. We term these hot spots as DWE bicliques. We identified DWE bicliques from crystal packing contacts, obligate and non-obligate interactions. Our comparative study revealed that there are abundant unique bicliques to the biological interactions, indicating specific biological binding behaviors in contrast to crystal packing. The two sub-types of biological interactions also have their own signature bicliques. In our analysis on residue compositions and residue pairing preferences in DWE bicliques, the focus was on interaction-preferred residues (ipRs) and interaction-preferred residue pairs (ipRPs). It is observed that hydrophobic residues are heavily involved in the ipRs and ipRPs of the obligate interactions; and that aromatic residues are in favor in the ipRs and ipRPs of the biological interactions, especially in those of the non-obligate interactions. In contrast, the ipRs and ipRPs in crystal packing are dominated by hydrophilic residues, and most of the anti-ipRs of crystal packing are the ipRs of the obligate or non-obligate interactions. CONCLUSIONS: These ipRs and ipRPs in our DWE bicliques describe a diverse binding features among the three types of interactions. They also highlight the specific binding behaviors of the biological interactions, sharply differing from the artifact interfaces in the crystal packing. It can be noted that DWE bicliques, especially the unique bicliques, can capture deep insights into the binding characteristics of protein interfaces.

We introduce a new topological concept called k-partite protein cliques to study protein interaction (PPI) networks.In particular, we examine functional coherence of proteins in k-partite protein cliques. A k-partite protein clique is a k-partite maximal clique comprising two or more nonoverlapping protein subsets between any two of which full interactions are exhibited. In the detection of PPI's k-partite maximal cliques, we propose to transform PPI networks into induced K-partite graphs with proteins as vertices where edges only exist among the graph's partites. Then, we present a k-partite maximal clique mining (MaCMik) algorithm to enumerate k-partite maximal cliques from K-partite graphs. Our MaCMik algorithm is applied to a yeast PPI network. We observe that there does exist interesting and unusually high functional coherence in k-partite proteincliques-most proteins in k-partite protein cliques, especially those in the same partites, share the same functions. Therefore, the idea of k-partite protein cliques suggests a novel approach to characterizing PPI networks, and may help function prediction for unknown proteins.

MOTIVATION: The O-ring theory reveals that the binding hot spot at a protein interface is surrounded by a ring of residues that are energetically less important than the residues in the hot spot. As this ring of residues is served to occlude water molecules from the hot spot, the O-ring theory is also called 'water exclusion' hypothesis. We propose a 'double water exclusion' hypothesis to refine the O-ring theory by assuming the hot spot itself is water-free. To computationally model a water-free hot spot, we use a biclique pattern that is defined as two maximal groups of residues from two chains in a protein complex holding the property that every residue contacts with all residues in the other group. METHODS AND RESULTS: Given a chain pair A and B of a protein complex from the Protein Data Bank (PDB), we calculate the interatomic distance of all possible pairs of atoms between A and B. We then represent A and B as a bipartite graph based on these distance information. Maximal biclique subgraphs are subsequently identified from all of the bipartite graphs to locate biclique patterns at the interfaces. We address two properties of biclique patterns: a non-redundant occurrence in PDB, and a correspondence with hot spots when the solvent-accessible surface area (SASA) of a biclique pattern in the complex form is small. A total of 1293 biclique patterns are discovered which have a non-redundant occurrence of at least five, and which each have a minimum two and four residues at the two sides. Through extensive queries to the HotSprint and ASEdb databases, we verified that biclique patterns are rich of true hot residues. Our algorithm and results provide a new way to identify hot spots by examining proteins' structural data. AVAILABILITY: The biclique mining algorithm is available at http://www.ntu.edu.sg/home/jyli/dwe.html. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.