- Record: found
- Abstract: found
- Article: not found
ProBiS-Database: Precalculated Binding Site Similarities and Local Pairwise Alignments of PDB Structures
research-article
Publication date (Electronic):
23 January 2012
Publisher:
American Chemical Society
Read this article at
There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.
Abstract
ProBiS-Database is a searchable repository of precalculated local structural alignments in proteins detected by the ProBiS algorithm in the Protein Data Bank. Identification of functionally important binding regions of the protein is facilitated by structural similarity scores mapped to the query protein structure. PDB structures that have been aligned with a query protein may be rapidly retrieved from the ProBiS-Database, which is thus able to generate hypotheses concerning the roles of uncharacterized proteins. Presented with uncharacterized protein structure, ProBiS-Database can discern relationships between such a query protein and other better known proteins in the PDB. Fast access and a user-friendly graphical interface promote easy exploration of this database of over 420 million local structural alignments. The ProBiS-Database is updated weekly and is freely available online at http://probis.cmm.ki.si/database.
Related collections
Most cited references15
- Record: found
- Abstract: found
- Article: found
ProBiS algorithm for detection of structurally similar protein binding sites by local structural alignment
Janez Konc, Dušanka Janežič (2010)
Motivation: Exploitation of locally similar 3D patterns of physicochemical properties on the surface of a protein for detection of binding sites that may lack sequence and global structural conservation. Results: An algorithm, ProBiS is described that detects structurally similar sites on protein surfaces by local surface structure alignment. It compares the query protein to members of a database of protein 3D structures and detects with sub-residue precision, structurally similar sites as patterns of physicochemical properties on the protein surface. Using an efficient maximum clique algorithm, the program identifies proteins that share local structural similarities with the query protein and generates structure-based alignments of these proteins with the query. Structural similarity scores are calculated for the query protein's surface residues, and are expressed as different colors on the query protein surface. The algorithm has been used successfully for the detection of protein–protein, protein–small ligand and protein–DNA binding sites. Availability: The software is available, as a web tool, free of charge for academic users at http://probis.cmm.ki.si Contact: dusa@cmm.ki.si Supplementary information: Supplementary data are available at Bioinformatics online.
- Record: found
- Abstract: found
- Article: not found
Detecting evolutionary relationships across existing fold space, using sequence order-independent profile-profile alignments.
Lei Xie, Philip E. Bourne (2008)
Here, a scalable, accurate, reliable, and robust protein functional site comparison algorithm is presented. The key components of the algorithm consist of a reduced representation of the protein structure and a sequence order-independent profile-profile alignment (SOIPPA). We show that SOIPPA is able to detect distant evolutionary relationships in cases where both a global sequence and structure relationship remains obscure. Results suggest evolutionary relationships across several previously evolutionary distinct protein structure superfamilies. SOIPPA, along with an increased coverage of protein fold space afforded by the structural genomics initiative, can be used to further test the notion that fold space is continuous rather than discrete.
- Record: found
- Abstract: found
- Article: found
Drug Discovery Using Chemical Systems Biology: Weak Inhibition of Multiple Kinases May Contribute to the Anti-Cancer Effect of Nelfinavir
Li Xie, Thomas Evangelidis, Lei Xie … (2011)
Introduction Tremendous effort has been directed at rational drug design where one strives to understand, and subsequently optimize, how a small molecule interacts with a single protein target and impacts a disease state. However, such approaches are less fruitful in discovering effective and safe therapeutics to treat complex diseases such as cancer. It is suggested that the inhibition or activation of a single specific target may fail owing to the inherent robustness of the underlying biological networks causing the disease state [1], [2], [3], [4], [5], [6]. The goal then is to perturb multiple relevant targets. Perturbation may be achievable through the use of drug cocktails, or possibly through a single drug that has the appropriate polypharmacological effect [1], [2], [4], [6], [7], [8], [9], [10], [11]. To rationally design such a drug is a very complex problem that begins by identifying the targets to which that drug binds. Here we address a much simpler problem, that is, to take a drug that is already believed to show this effect and attempt to explain why it might be so. Nevertheless, we must still begin by identifying the multiple targets to which it binds. To this end, we have developed an off-target pipeline to identify protein-drug interaction profiles on a structural proteome-wide scale. The off-target pipeline integrates our previous chemical systems biology approach [12], [13], [14] with algorithms that accurately estimate binding affinity. We then use the target list predicted from the off-target pipeline to suggest physiological outcomes from the associated biological networks and determine how well these outcomes map to what is observed clinically. The extension to our previous approach presented here is to better estimate the binding affinity in forming a protein-ligand complex, as both experimental and theoretical studies suggest that even weak binding to multiple targets may have profound impact on the overall biological system [15], [16], [17]. Available computational tools that quantitatively study protein-ligand interactions are based predominantly on protein-ligand docking and free energy calculations for the protein-ligand complex [18], [19]. A formidable task then is to include protein flexibility into the binding affinity calculation since errors in scoring mainly result from the use of rigid protein conformations [20]. The modeling of protein flexibility requires computationally intensive molecular dynamic (MD) simulations. However, it is impractical to apply MD simulation to the whole structural proteome. Our approach pre-filters the structural proteome to find the most likely cases to apply MD. Specifically, we undertake a human structural proteome-wide ligand binding site comparison using previously developed algorithms [21], [22], [23] and add intensive binding free energy calculations, based on protein-ligand docking, MD simulation and MM/GBSA free energy calculations. We apply this strategy to explore the molecular mechanism for the observed anti-cancer effect of Nelfinavir, a human immunodeficiency virus (HIV) protease inhibitor. Recently, Nelfinavir has been repurposed for cancer treatment [24], [25], [26]. However, its molecular targets remain unknown. The majority of published data indicates that the drug suppresses the Akt signaling pathway [27]. In human, the Akt family includes the serine/threonine protein kinases Akt1, Akt2 and Akt3. These proteins are involved in cell survival, protein synthesis and glucose metabolism and are considered markers for many types of cancer [28], [29], [30]. Akt3 is also known to be stimulated by platelet-derived growth factor (PDGF), insulin and insulin-like growth factor 1 (IGF1) [31]. Thus inhibition of the Akt pathway may also cause insulin resistance and diabetes, a phenomenon observed as a side effect of treatment by HIV protease inhibitors. Currently, there is no experimental evidence to suggest that Nelfinavir binds directly to members of the Akt family, rather it has been suggested that the drug acts upstream of the Akt signaling pathway [32]. Using our structural proteome-wide off-target pipeline, we find that multiple members of the protein kinase-like superfamily as off-targets of Nelfinavir. Most of these protein kinases are found upstream of the Akt, MAPK, JNK, NF-κB, mTOR and focal adhesion pathways. We hypothesize that this weak but broad spectrum protein kinase inhibition by Nelfinavir contributes to the therapeutic effect against different types of cancer. Our hypothesis is supported by kinase activity assays and consistent with other existing experimental and clinical observations. This suggests that the next challenges are specifically to optimize Nelfinavir as a targeted polypharmacology agent, and more generally, to determine whether our computational protocol can be applied to other systems. Results Putative off-targets of Nelfinavir The steps in our off-target pipeline are shown in Figure 1. In the first step, the Nelfinavir binding pocket in the HIV protease dimer structure (PDB Id: 1OHR) was used to search against 5,985 PDB structures of human proteins or homologs of human proteins using the SMAP software (see Supporting Information (SI) and methods for details), which is based on a sensitive and robust ligand binding site comparison algorithm [21], [22], [23]. Hits are considered significant if the SMAP p-value 1.0e-2). Although more analyses are required to determine if Nelfinavir binds to other proteins that indirectly regulate EGFR pathways, the data reported here at least suggest that the pleotropic effect of Nelfinavir comes from the direct inhibition of a variety of protein kinases. A fundamental question raised by this work is whether weak binding of a drug to multiple targets can cumulatively cause strong phenotypic changes? Existing studies of biological networks have shown that the malfunction of multiple nodes more likely causes the system to fail than the removal of a single node as a result of diversity, redundancy and system control of the biological network. Multiple node failures have been called “fail-on” [50], and used to explain neurological disorders [51] and cancer [52], [53] in recent genome-wide studies. Addressing the fail-on phenomenon would require a polypharmacological effect. The therapeutic efficacy of multiple protein kinase inhibitors suggested here has already been demonstrated by less specific protein kinase inhibitors which attack tumors through multiple mechanisms and are used in more than one type of cancer therapy [54]. For example, Sunitinib is the first cancer drug simultaneously approved for two different cancer treatments, namely, renal cell carcinoma and imatinib-resistant gastrointestinal stromal tumor. A protein kinase assay against 113 different kinases shows that Sunitinib can bind to 73 additional kinases apart from its primary target [55]. In another example, moderate micomolar RAF inhibitor PLX4720 is potent in inhibiting downstream signaling and proliferation of the cell harboring BRAF, and in treating melanoma cell lines [56]. In contrast, Sorafenib that was developed as a potent nanomolar RAF inhibitor failed in the clinical trial due to its low anti-melonoma efficacy. Araujo et al. demonstrated the synergistic effect of multiple low-dose inhibition of upstream processes on the attenuation of downstream signals in the EGFR signaling pathway [57], and suggested that low-dose combination therapy may reduce drug side effects and resistances in the treatment of cancer [58]. Nelfinavir is a potential lead compound in the design of the next generation of anti-cancer drugs. As indicated by the MM/GBSA binding free energies for different protein kinases, the binding affinity of Nelfinavir is weaker than for the original inhibitors. Entropy changes during binding contribute significantly to the differences in binding affinity since Nelfinavir consists of more rotatable bonds and is more flexible than many small molecule protein kinase inhibitors. Covalent bonds could be added to the Nelfinavir structure to reduce the degree of freedom and increase the specificity and binding affinity. On the other hand, it can be hypothesized that the weak binding of Nelfinavir to multiple protein kinases helps avoid severe side effects, but still impacts the system enough to have a positive effect. That is, weak inhibition of multiple protein kinases may be just enough to return the system to a normal state, as suggest by dynamic analysis of model systems [57], [58]. There are a number of unmet computational challenges in exploiting the concept of multiple weak interactions and designing selective polypharmacology therapeutics, from target identification to lead optimization. Computational techniques that are able to identify optimal combination targets and their inhibition windows in cellular networks have been developed, but their scope is still limited [59], [60], [61]. It is well accepted that an optimal lead should balance binding potency and molecular size [62]. A highly potent lead compound usually leads to a drug candidate with high molecular weight, which is often linked to a higher risk of failure in drug development [63]. Analysis of the binding affinity of marketed drugs and natural products indicates that therapeutic efficacy is not necessarily associated with high binding affinity [63]. Moreover, drug-target interactions in vivo are different from those in vitro. An increasing body of evidence suggests that the drug-target residence time, a measurement of the lifetime of the drug-target complex, better correlates to drug efficacy than does the binding affinity [64], [65]. This suggests that lead optimization should focus on the drug-target residence time instead of binding affinity. Although methodologies have been proposed for multi-target screening based on binding affinity [66], there are simply no computational tools available for the efficient and accurate a priori estimation of the drug-target residence time from molecular structures. A detailed understanding of the effect of multiple interactions on the biological network requires innovative systems biology approaches. The qualitative description of the biological network presented here is limited in its predictive power, considering the highly dynamic nature of signal transduction pathways. A mathematical modeling approach will be more powerful than the static approach as we have demonstrated recently in a study of CETP inhibitors [67]. Existing mathematical modeling methods such as ordinary differential equations, Petri nets, and pi-calculus require a large number of kinetics parameters to simulate the dynamic behavior of the biological system [68]. In practice many of these parameters may not be available. Thus the network model has to be reduced. The qualitative properties derived from off-target binding network may help to develop restrained but functional dynamic models that are suitable for parameter optimization and mathematical modeling. In conclusion, by integrating methods from structural bioinformatics, molecular modeling and network analysis, we propose that the observed anti-cancer effects of the HIV protease inhibitor Nelfinavir derive from weak binding to multiple protein kinases that are mostly upstream of the PI3K/Akt pathway. Our computational approach, enhanced from previous work with the use of MD simulation and MM/GBSA free energy calculations, is supported by kinase activity assays and existing experimental and clinical evidence. This type of approach has the potential to be generalized as a form of rational polypharmacological drug design. Materials and Methods Overview of structural proteome-wide off-target pipeline The structural proteome-wide off-target pipeline is outlined in Figure 1. Firstly, the Nelfinavir binding pocket in the HIV protease (PDB Id: 1OHR) was used to search against 5,985 PDB structures of human proteins or homologous of human proteins using the SMAP software [21], [22], [23]. Secondly, the binding poses and affinities of Nelfinavir to these putative off-targets are estimated using two docking methods, Surflex [33] and eHiTs [34]. If the docking score indicates severe structural clashes between Nelfinavir and the predicted binding pocket, the protein is removed from the off-target list. Finally, the remaining putative off-targets are subject to MD simulation, MM/GBSA calculation, network reconstruction, and kinase activity assay. Ligand binding site similarity search 5,985 PDB structures that are homologous to human proteins (sequence identify >30%, alignment coverage larger than 90%) are searched against the HIV-1 protease dimer (PDB id: 1OHR) using SMAP, which can be downloaded from http://funsite.sdsc.edu. The detailed algorithms implemented in SMAP are presented elsewhere [21], [22], [23]. In brief, proteins are represented using Cα atoms only and characterized by a geometric potential [21]. Then two proteins are aligned to identify similar local binding sites using the Sequence Order Independent Profile-Profile Alignment (SOIPPA) algorithm [22]. The statistical significance of the binding site similarity is estimated using an extreme value distribution model [23]. Reverse docking of the human structural proteome The binding affinity of Nelfinavir to the putative off-targets with SMAP p-value less than 1.0e-3 are estimated by two docking methods, Surflex [33] and eHiTs [34]. First, the complex structure of HIV-1 protease with Nelfinavir is superimposed onto these proteins according to the SMAP alignment. The superimposed structure of Nelfinavir is used as the starting conformation for docking. The binding pose of Nelfinavir in these statistically significant off-targets is locally optimized and scored starting from the starting conformation using Surflex 2.1 (default setting) and eHits 6.2 (the fastest setting). The docking score is normalized using the protocol described in reference [13]. MD simulation and MM/GBSA binding free energy calculation MM/GBSA [69], [70] was developed for free energy calculations and has been used to estimate the binding affinity for several protein or DNA systems [71], [72], [73], [74]. Here we perform ensemble average MM/GBSA binding free energy calculation on the snapshots from the MD simulation to compare binding affinity of Nelfinavir with that of the co-crystallized ligands. MD simulation on the complex structures of predicted protein kinase off-targets and their inhibitors including Nelfinavir Explicit solvent molecular dynamics simulations were performed with NAMD [75] on the structures of the Nelfinavir-protein kinase complexes and co-crystallized ligand-protein kinase complexes. The starting structure for Nelfinavir in each protein kinase is the lowest energy conformation obtained through Autodock Vina [76]. These complex structures are embedded in rectangular boxes of TIP3P water [77] molecules to mimic the solvent environment. The smallest distances between the edge of the boxes and the atoms of the complex structures are adjusted to be at least 10 Å. Ions are added to neutralize these systems and satisfy the salt concentrations. The salt concentration is obtained from individual experimental condition for each protein kinase. The long distance cut-off for both van der Waals interactions and electrostatic interactions is set as 14 Å. A switching function is used to truncate the van der Waals energy smoothly at the cut-off distance. The Particle Mesh Ewald (PME) [78] method is applied to treat the long range electrostatic interactions. All covalent bonds involving hydrogen atoms are constrained by the SHAKE algorithm [79]. In order to simulate the NPT ensemble (system with a fixed pressure P, temperature T, and number of atoms N), the Langevin piston Nose-Hoover method [80], [81] in NAMD together with the periodic boundary conditions is used to maintain a constant pressure and temperature for these systems. Systems are first minimized by a five-stage minimization protocol. Hydrogen atoms are optimized in the first stage, keeping all other atoms fixed. Then water molecules and side chain atoms are relaxed in the second and third stage, respectively. All atoms are optimized in the fourth stage, with position restraints on backbone atoms of proteins and ligands. Minimization is completed by an additional 25,000 steps, without any restraints, to remove bad contacts. All minimizations are preformed with the conjugate gradient energy minimization method [82] in NAMD. The optimized systems are then gradually heated from 0 K to 50K, 100 K, 150 K, 200 K, 250 K, and experimental temperature (about 298 K) with position restraints on the backbone atoms. The structures are equilibrated at each temperature for 250 ps with a 1.0 fs time step. The force constant of restraints is 4.0 kcal/mol·Å-2. After the systems are heated to the experimental temperature, position restraints are removed in the following 120 ps simulation by gradually reducing the force constant. Subsequently 8 ns NPT MD simulations are carried out on these systems with 1.0 fs time step at the experimental temperature. 200 snapshots are extracted from the last 2 ns simulations with 10.0 ps time intervals to generate representative configurations for the MM/GBSA binding free energy calculation. MM/GBSA calculation The binding free energy can be calculated through the following equation: (1) where Gcomplex , Greceptor , Gligand are the free energies of the complex, receptor and ligand respectively. The free energy of each molecular on the right hand side can be considered as the sum of molecular mechanical energy in gas phase, solvation energy and entropy term, as shown in the following formula: (2) EMM is calculated by the molecular mechanics method with standard force field parm9 in AMBER9 package [83], [84]. The electrostatic contribution to the solvation free energy is determined by the Generalized Born (GB) model [85], [86], [87], [88], a widely used continuum solvent model. The “OBC” model with modified Bondi radii (mbondi2) [89], [90], [91] in AMBER9 is applied to calculate this part of energy. This model is newer than the original version of the GB model and provides a significant improvement and is recommended for both proteins and nucleic acids. The interior dielectric constant of the molecule of interest is set as 1.0 and the exterior or solvent dielectric constant is set as 78.5. The non-polar contribution to the solvation free energy is proportional to the solvent-accessible surface area [89], [92]. The surface area is calculated by the LCPO model [93] and the surface tension used to calculate the non-polar part is taken as 0.0072 kcal/mol·Å-2. The entropic term is the most time-intensive part of the MM/GBSA calculation but is found to be indistinguishable among different conformational states and contributes less than the other two terms in many application for estimating relative binding free energies [69], [94], [95]. The entropy change associated with ligand binding is estimated by normal mode analysis [96] in AMBER9. For each system, the MM/GBSA calculation is carried out on the 200 snapshots extracted from the last 2ns of the MD simulation. Protein kinase activity assay HTRF® TranscreenerTM ADP Assays were performed on EGFR, ErbB2 and ErbB4 Akt1, Akt2 and Akt3 by GenScript (New Jersey, U.S.A). Nelfinavir Mesylate was purchased from Toronto Research Chemicals (North York, Canada). The compound is diluted to a 10 mM concentration with acetone and stored at -20°C. Inhibition of Nelfinavir at 20 µM was tested on EGFR, ErbB-2, ErbB-4 and Akt (Akt1, Akt2, Akt3). Supporting Information Figure S1 Structural root mean square deviations (RMSDs) for receptor backbone atoms and ligand non-hydrogen atoms as a function of simulation time. (DOC) Click here for additional data file. Figure S2 Comparison between binding poses of predicted Nelfinavir and co-crystallized inhibitors after MD simulation for the four receptor tyrosine protein kinases, EGFR, IGF-1R, FGFR and EPHB4. (DOC) Click here for additional data file. Figure S3 Comparison between binding poses of predicted Nelfinavir and co-crystallized inhibitors after MD simulation for FAK, Akt2, CDK2, Abl, ARK and PDK1. (DOC) Click here for additional data file. Figure S4 Structure of Nelfinavir. (DOC) Click here for additional data file. Table S1 SMAP p-values, Surflex docking scores and eHiTs docking scores for putative off-targets of Nelfinavir. (DOC) Click here for additional data file. Table S2 Comparison of normalized docking score (NDS) of Nelfinavir and co-crystallized ligands to the predicted PK off-targets. (DOC) Click here for additional data file. Table S3 Comparison of Autodock Vina energies normalized of Nelfinavir, Saquinavir and Indinavir when bound to EGFR, ErbB2 and ErbB4. (DOC) Click here for additional data file.
Author and article information
Journal
Journal ID (nlm-ta): J Chem Inf Model
Journal ID (publisher-id): ci
Journal ID (coden): jcisd8
Title:
Journal of Chemical Information and Modeling
Publisher:
American
Chemical Society
ISSN
(Print):
1549-9596
ISSN
(Electronic):
1549-960X
Publication date
(Electronic):
23
January
2012
Publication date
(Print):
27
February
2012
Volume: 52
Issue: 2
Pages: 604-612
Affiliations
Author notes
Article
DOI: 10.1021/ci2005687
PMC ID: 3287116
PubMed ID: 22268964
SO-VID: a95c3e27-7b3a-4bfd-b6fd-44ae50b1e95d
Copyright © Copyright © 2012 American Chemical Society
License:
This is an open-access article distributed under the ACS AuthorChoice Terms & Conditions. Any use of this article, must conform to the terms of that license which are available at http://pubs.acs.org.
History
Date
received
: 02
December
2011
Date: 07
February
2012
Date: 27
February
2012
Date: 23
January
2012
Categories
Subject:
Article
Custom metadata
document-id-old-9 ci2005687
document-id-new-14 ci-2011-005687
ccc-price
ScienceOpen disciplines: Computational chemistry & Modeling
Data availability:
ScienceOpen disciplines: Computational chemistry & Modeling