Bioinformatics Advance Access originally published online on May 30, 2007
Bioinformatics 2007 23(15):1919-1926; doi:10.1093/bioinformatics/btm277
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Exploring sequence-structure relationships in the tyrosine kinome space: functional classification of the binding specificity mechanisms for cancer therapeutics
Department of Pharmaceutical Chemistry, School of Pharmacy, Center for Bioinformatics, The University of Kansas, 2030 Becker Drive, Lawrence, KS 66047-1620
*To whom correspondence should be addressed.
 |
ABSTRACT |
|---|
 TOP ABSTRACT 1 INTRODUCTION2 SYSTEMS AND METHODS3 RESULTS AND DISCUSSION4 CONCLUSIONSREFERENCES |
|---|
Motivation: Evolutionary and structural conservation patterns shared by more than 500 of identified protein kinases have led to complex sequence-structure relationships of cross-reactivity for kinase inhibitors. Understanding the molecular basis of binding specificity for protein kinases family, which is the central problem in discovery of cancer therapeutics, remains challenging as the inhibitor selectivity is not readily interpreted from chemical proteomics studies, neither it is easily discernable directly from sequence or structure information. We present an integrated view of sequence-structure-binding relationships in the tyrosine kinome space in which evolutionary analysis of the kinases binding sites is combined with computational proteomics profiling of the inhibitor–protein interactions. This approach provides a functional classification of the binding specificity mechanisms for cancer agents targeting protein tyrosine kinases.
Results: The proposed functional classification of the kinase binding specificities explores mechanisms in which structural plasticity of the tyrosine kinases and sequence variation of the binding-site residues are linked with conformational preferences of the inhibitors in achieving effective drug binding. The molecular basis of binding specificity for tyrosine kinases may be largely driven by conformational adaptability of the inhibitors to an ensemble of structurally different conformational states of the enzyme, rather than being determined by their phylogenetic proximity in the kinome space or differences in the interactions with the variable binding-site residues. This approach provides a fruitful functional linkage between structural bioinformatics analysis and disease by unraveling the molecular basis of kinase selectivity for the prominent kinase drugs (Imatinib, Dasatinib and Erlotinib) which is consistent with structural and proteomics experiments.
Contact: verk{at}ku.edu or gverkhiv{at}ucsd.edu
Supplementary information: Supplementary data are available at Bioinformatics online.
|
1 INTRODUCTION |
|---|
|
TOPABSTRACT1 INTRODUCTION2 SYSTEMS AND METHODS3 RESULTS AND DISCUSSION4 CONCLUSIONSREFERENCES |
|---|
The human genome advances have accelerated a comprehensive analysis of protein kinases in human and model organisms and a detailed comparison of protein phosphorylation in normal and disease states (Hubbard, 2002; Hubbard and Till, 2000; Madhusudan and Ganesan, 2004; Sridhar et al., 2000). Protein tyrosine kinases, which are responsible for development of many cancers (Hubbard, 2002; Hubbard and Till, 2000), have emerged as a major class of drug targets in recent years (Madhusudhan and Ganesan, 2004; Sridhar, et al., 2000). Evolutionary and structural conservation patterns shared by more than 500 of identified protein kinases (Kostich et al., 2002; Krupa and Srinivasan, 2002; Manning et al., 2002) have led to complex sequence-structure relationships of cross-reactivity for kinase inhibitors. While alternative cellular targets have begun to be unraveled using proteomics technologies, these discoveries are still constrained by the difficulty of building and running large numbers of consistent activity assays. Understanding the molecular basis of protein kinases binding specificity presents a fundamental and highly challenging biological problem as the binding specificity is not readily interpreted from chemical proteomics studies, neither it is easily discernable directly from sequence and structural information (Bain et al., 2003; Sawyer et al., 2004; Sawyer et al., 2003).
The biological activities of the kinase inhibitors and their selectivity profiles with protein kinases are only weakly related with the phylogenetic classification of protein kinases inferred from sequence alignments of the kinase domain (Vieth et al., 2005; Vulpetti and Bosotti, 2004). Imatinib mesylate (Gleevec), which is an effective direct inhibitor of ABL, KIT and PDGFR tyrosine kinases (Deininger et al., 2005; Druker, 2004; Lydon and Duker, 2004) exemplifies inherent complexity in inferring binding specificity from sequence homology of kinase domains. According to the phylogenetic annotation of the protein kinase catalytic domain, PDGFR and KIT kinases, specifically targeted by Imatinib with high affinity, are more divergent from the ABL tyrosine kinase than SRC kinases, which are not inhibited by Imatinib (Deininger et al., 2005; Wong and Witte, 2004). Evolutionary conservation of the kinase domain co-exists with a considerable conformational plasticity of protein kinases involved in regulating catalytic activity of this protein superfamily (Huse and Kuriyan, 2002; Levinson et al., 2006; Nager et al., 2002; Schindler et al., 2000; Wisniewski et al., 2002). The remarkable variability of the ABL kinase conformational states, which include active, inactive, intermediate conformations and inactive–like conformations, has confirmed that diverse structures of the ABL activation loop are natural protein conformations and dynamic equilibrium between multiple conformational states plays an important critical role in molecular recognition of ABL (Levinson et al., 2006; Tokarski et al., 2006; Young et al., 2006). Indeed, the molecular basis for a high selectivity of Imatinib at inhibiting ABL, KIT and PDGFR tyrosine kinases is primarily determined by specific recognition of the unique inactive conformation shared by all these kinases. Furthermore, the crystal structure of the ABL kinase with Dasatinib, an alternative BCR-ABL cancer therapeutic agent which has demonstrated a broad range of tyrosine kinases activities, has provided further evidence that the inhibitor activity against various tyrosine kinases is largely driven by the recognition requirements to recognize multiple conformational states of the enzyme (Levinson et al., 2006; Tokarski et al., 2006).
Another example of complexities associated with deriving functional profiles from sequence conservation of kinases domains is Erlotinib (OSI-774, Tarceva), an approved drug for treatment of non-small cell lung cancer (NSCLC) and a high affinity, selective kinase inhibitor against EGFR receptor with only weak affinity against ABL, SRC and LCK kinases (Brown and Shephard et al., 2005; Kwak et al., 2005; Stamos et al., 2002; Wood et al., 2004). Crystal structures of the EGFR kinase domain which have been determined previously in a free form and in the complex with Erlotinib (Stamos et al., 2002) have revealed an active conformation of the enzyme. However, a structurally different inactive conformational state of EGFR kinase domain has been found in the complex with the drug Lapatinib (Wood et al., 2004), which resembles closely that of inactive Src family kinases. Strikingly, the wild-type EGFR kinase domain assumes an autoinhibited, Src-like inactive state and the original crystal structure of the EGFR kinase domain (Stamos et al., 2002) may have been observed in the active conformation largely due to specific crystallization conditions. The recently solved crystal structure of the EGFR kinase domain in complex with AMP-PNP have also produced inactive SRC-like conformation, suggesting that active conformation of the EGFR kinase domain may be a consequence of mimicry by the crystal of an intrinsic activation mechanism that destabilize the otherwise thermodynamically favorable inactive conformation of the EGFR receptor (Zhang et al., 2006). These experiments have suggested that Erlotinib-binding affinity and selectivity with the EGFR receptor is achieved by a complex mechanism which may involve recognition of multiple conformational states of EGFR, rather than binding with the single active state of EGFR. The growing body of structural and biochemical information about protein kinases recognition points to a diversity and complexity of the binding mechanisms. These mechanisms appear to be far more complex than simple functional annotations of the kinases binding specificities with known cancer drugs which can be typically derived from sequence conservation profiles of the catalytic domain and the binding sites.
The emerging diversity of the tyrosine kinase crystal structures have broadened the realm of computational models, which may be employed for functional classification of the binding specificities and explaining the activity profiles of kinase inhibitors. A considerable structural repertoire of the protein kinase conformational states involved in regulating molecular recognition of the enzyme with the inhibitors and interacting proteins has provided support to a modern ‘conformational selection’ paradigm of biomolecular binding based on the energy landscape theory (Kumar et al., 2000; Levy et al., 2004; Verkhivker et al., 2003; Shoemaker et al., 2000), which implies the presence of an ensemble of the multiple conformational states for both interacting molecules. We have recently introduced a computational approach for in silico profiling of the kinase inhibitors, which is based on the ‘conformational selection’ binding model and combines evolutionary analysis with computational proteomics profiling of biomolecular interactions (Verkhivker, 2006; Verkhivker, 2007; Verkhivker, et al., 2006). In this work, we present an integrated view of sequence-structure-binding relationships in the tyrosine kinome space which allows a functional classification of the binding specificity mechanisms for cancer agents targeting protein tyrosine kinases.
|
2 SYSTEMS AND METHODS |
|---|
|
TOPABSTRACT1 INTRODUCTION2 SYSTEMS AND METHODS3 RESULTS AND DISCUSSION4 CONCLUSIONSREFERENCES |
|---|
2.1 Phylogenetic analysis of kinases
Phylogenetic dendrogram of the human protein tyrosine kinase family is inferred from the sequence conservation profile of the kinase catalytic domains, which were aligned using CLUSTAL W (Thompson et al., 1994). Phylogenetic trees for whole-domain alignments and binding site subalignments were created with PHYLIP package of programs (version 3.6) (Felsenstein, 2002). The robustness of the nodes in the trees generated by Neighbor–Joining method (Saitou and Nei, 1987) and the PamDayhoff distance matrix has been tested with 300 500 bootstrap replicates using the bootstrap algorithm available in CLUSTAL W (Saitou and Nei, 1987). Confidence limits have been tested by bootstrapping the trees 1000 times.
Phylogenetic trees based on evolutionary conservation profiles of the binding site residues have also been constructed using Neighbor–Joining method (Saitou and Nei, 1987) within the Pfaat protein family alignment annotation tool (Johnson et al., 2003). Binding-site residues were defined as those with a side-chain heavy atom 5Å from the ligand. We included C
atoms in the side-chain definition so that glycine residues that forms a part of the binding pocket were not excluded. Residues with side chains making a crystallographically defined water-mediated hydrogen bond were also considered part of the binding pocket.
The residues lining the binding cleft of protein kinases are assumed to serve as better evolutionary indicators of ligand-binding function and specificity than whole-domain sequences. For each solved structure in the alignment, residues with side chains within 4.5Å of their respective ligands were identified, and the columns of the multiple alignment that contained these binding-site residues were extracted. This composite set of 26 critically important contact alignment positions were compiled from all presently available tyrosine crystal structures (referred to as the binding site subalignment) and was then used for phylogenetic clustering. We have constructed the phylogenetic dendrogram based on the evolutionary conservation profile of the binding-site residues, defined from the crystal structures of Imatinib mesylate complex with the ABL kinase and complex of Erlotinib with the EGFR receptor.
2.2 Structural classification of protein tyrosine kinases
All currently available crystal structures of protein tyrosine kinases in the Protein Data Bank (PDB) (Berman et al., 2000) are used to categorize the structural space into panels of active and inactive kinase conformations. The following protein structures, employed in this study, provide the extensive coverage of tyrosine kinase family and include : 6 ABL (1fpu, 1iep, 1m52, 1opj, 1opk, 1opl); 1 ACK1 (1u4d); 3 KIT (1pkg, 1t45, 1t46); 1 ZAP70 (1u59); 1 BTIK (1k2p); 1 ITK (1sm2); 1CSK (1byg); 2 SYK (1xbb, 1xbc); 6 EGFR (1m14, 1m17, 1xkk, 2gs2, 2gs6, 2gs7); 4 IGF1R (1m7n, 1k3a, 1jqh, 1p4o); 2 EphA2 (1mqb, 1jpa); 1 FAK (1mp8); 4 FGFR1 (2fgi, 1fgk, 1fgi, 1agw); 2 HGFR (1r0p, 1r1w); 1 FLT3 (1rjb); 3 HCK (1ad5, 1qcf; 2hck); 4 INSR (1irk, 1ir3, 1i44, 1p14); 5 LCK (1qpc, 1qpd, 1qpe, 1qpj, 3lck); 1 MUSK (1luf); 4 SRC (1fmk, 1ksw, 2ptk, 2src); 1 TIE2 (1fvr); 1 KDR/VEGFR (1vr2).
The following 20 structures belong to the category of inactive kinase conformations : 3 ABL (1fpu, 1iep, 1opj); 2 KIT (1t45, 1t46); 1 BTIK (1k2p); 1CSK (1byg); 2 IGF1R (1m7n, 1k3a); 2 EphA2 (1mqb, 1jpa); 1 FAK (1mp8); 1 FGFR1 (1fgk); 2 EGFR (1xkk, 2gs7); 1 HGFR (1r1w); 1FLT3 (1rjb); 1 HCK (1ad5); 1 INSR (1irk); 1 MUSK (1luf) ; 1 SRC (1fmk); 1 TIE2 (1fvr).
The superposition of crystal structures from the protein tyrosine kinase family into a common reference frame is based on similarity of C atoms for a common set of residues defining the ATP binding site.
2.3 Monte Carlo binding simulations: energy model
The molecular recognition energetic model used in this study includes intramolecular energy terms, given by torsional and non-bonded contributions of the DREIDING force field (Mayo et al., 1990), and the intermolecular energy contributions calculated using the AMBER force field (Cornell et al., 1995) to describe protein–protein interactions combined with an implicit solvation model (Stouten et al., 1993). The dispersion-repulsion and electrostatic terms have been modified and include a soft core component that was originally developed in free energy simulations to remove the singularity in the potentials and improve numerical stability of the simulations (Beutler et al., 1994). Softening the AMBER force field can enhance sampling of the conformational space while retaining adequate description of the binding energy landscape.
A solvation term was added to the interaction potential to account for the free energy of interactions between the explicitly modeled atoms of the protein–protein system and the implicitly modeled solvent. The term was derived by considering the transfer of atom, from an environment where it is completely surrounded by solvent, to an environment in which it has explicit atomic neighbors. The solvent-accessible surface is created by rolling a spherical water probe with a radius of 1.4Å over the Van der Waals surface of the molecule. The center of the probe traces the solvent-accessible surface.
2.4 Monte Carlo binding simulations: simulated tempering dynamics
We have carried out equilibrium simulations with the ensembles of protein kinase conformations using parallel simulated tempering dynamics (Hansmann, 1997; Sugita and Okamoto, 1999; Verkhivker et al., 2003) with 50 replicas of the ligand–protein system attributed, respectively to 50 different temperature levels that are uniformly distributed in the range between 5300 and 300 K. In simulations with ensembles of multiple protein conformations, protein conformations are linearly assigned to each temperature level, that implies a consecutive assignment of protein conformations starting from the highest temperature level and allows each protein conformation from the ensemble to be assigned to a certain temperature level at least once. Starting with the highest temperature, every pair of adjacent temperature configurations is tested for swapping until the final lowest value of temperature is reached. This process of swapping configurations is repeated 50 times after each simulation cycle for all replicas where the exchange of conformations presents an improved global update which increases thermalization of the system and allows regions with a small density of states to be sampled accurately.
Monte Carlo moves are performed simultaneously and independently for each replica at the corresponding temperature level. After each simulation cycle, that is completed for all replicas, exchange of configurations for every pair of adjacent replicas at neighboring temperatures is introduced. The mth and nth replicas, described by a common Hamiltonian H(X1, ... ,Xm, ... ,Xn, ... ,XN), are associated with the inverse temperatures ßm and ßn, and the corresponding conformations Xm and Xn. The exchange of conformations between adjacent replicas m and n is accepted or rejected according to Metropolis criterion with the probability
|
| (1) |
|
| (2) |
At equilibrium, the fraction of time that the ligand–protein system spends at a protein conformation 
= i to time spent at a protein conformation = j is determined by the Boltzmann distribution
|
| (3) |
2.5 Binding free energy calculations
Binding free energies are computed using the molecular mechanics AMBER force field (Cornell et al., 1995) and the solvation energy term based on continuum generalized Born and solvent accessible surface area (GB/SA) solvation model (Mohamadi et al., 1990; Still et al., 1990).
The binding free energy of the ligand–protein complex can be written as follows:
|
| (4) |
|
| (5) |
|
| (6) |
In the GB/SA model, the Gcavity and Gvdw contributions are combined together via evaluating solvent-accessible surface areas:
|
| (7) |
|
| (8) |
Gpol is the polar solvation energy which is computed using the GB/SA solvation model. Ssolute is the vibrational entropy of the molecule. EMM is the molecular mechanical energy of the molecule summing up the electrostatic Ees interactions, van der Waals contributions Evdw and the internal strain energy Eint.
|
3 RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACT1 INTRODUCTION2 SYSTEMS AND METHODS3 RESULTS AND DISCUSSION4 CONCLUSIONSREFERENCES |
|---|
We have previously found that evolutionary classification of binding specificity based on the binding-site residues is rather sensitive to the binding-site definition and may not have a significant relationship with the respective function of these kinases to bind a particular inhibitor, unless binding-site residues are defined from the crystal structure with the respective ligand (Fig. 1) (Verkhivker, 2006). Indeed, when the binding-site residues are determined from the crystal structure of Imatinib mesylate with the ABL kinase, the phylogenetic dendrogram and the evolutionary conservation profile of the ABL kinase binding-site residues result in a closer proximity of ABL, CKIT and PDGFR kinases, whereas ABL and SRC kinases emerged as less similar (Fig. 1). Phylogenetic classification of the protein tyrosine kinase family and subalignments of the binding site identify both conserved and variable residue positions (Figs 1 and 2). This analysis points to a subset of variable residues (Q767, L768, M768, P770, F771, G772, L820 and T830) among ABL, SRC and EGFR kinases, suggesting their potential functional role in governing selective EGFR inhibition of Erlotinib. Interestingly, the active site conservation patterns suggest that binding specificity to a particular inhibitor may be provided by a somewhat different constellation of variable amino acids recruited to accomplish specific binding interactions (Figs 1 and 2). The phylogenetic analysis based on the binding-site residues obtained from the crystal structure of EGFR with Erlotinib reproduces a spectrum of Erlotinib activities and identifies EGFR along with the ABL and SRC kinases as functionally relevant for binding kinase targets (Figs 1 and 2). However, evolutionary analysis of the kinase binding-site residues cannot be readily translated into the distinct specificity phenotype observed for these kinase inhibitors. Although the evolutionary and structural role of the conserved Thr gate-keeper residue (Thr-315 in the ABL–Imatinib complex and Thr-766 in the EGFR–Erlotinib complex) shared between EGFR and ABL and SRC kinase families is well recognized; the binding specificities of Imatinib, Dasatinib and Erlotinib against ABL, SRC, LCK and EGFR kinases can vary dramatically.
|
|
We suggest that a functional linkage between protein conformational diversity and sequence plasticity of the kinase binding site may be important to understand and rationalize binding specificity profiles of these cancer agents. Structural and energetic aspects of Dasatinib and Erlotinib binding with the conformational states of the tyrosine kinases are studied using computational profiling of the inhibitor–protein interactions with an ensemble of inactive and active kinase crystal structures. The working hypothesis is that binding specificity mechanisms may be driven by conformational adaptability of the inhibitor to structurally different conformational states of the enzyme. The crystal structure of Imatinib from the complex with the specific inactive form of ABL is incompatible with any of the alternative conformational states of the enzyme. Consequently, the high affinity binding of Imatinib with the ABL kinase is assured by the conformationally specific inhibitor recognition of the ABL inactive form (Verkhivker, 2006). While Imatinib binding is highly sensitive to the activation state of the enzyme, binding promiscuity of Dasatinib in inhibiting a spectrum of tyrosine kinases may be largely determined by a considerable adaptability of the inhibitor binding mode to structurally diverse conformational states of ABL, CKIT, SRC and EGFR kinases with high binding affinity (Fig. 3). Interestingly, the conformationally permissive binding of Dasatinib with the multiple conformational states of various tyrosine kinases can be ensured by using only small thermal variations of the dominant binding mode. This is reflected in a high degree of structural similarity between the crystal structure and the predicted inhibitor conformations, whereby Dasatinib fluctuates only within 1.5 Å from the crystallographic conformation in order to favorably accommodate the multitude of kinase conformational states (Fig. 3). Simulations of Erlotinib binding reveal considerably larger thermal fluctuations of the inhibitor and more stringent requirements to specific binding (panels C,D in Fig. 2). Nevertheless, the predicted bound conformations of Dasatinib and Erlotinib (panels C,D in Fig. 2) in complexes with both active and inactive conformational states of ABL, SRC, LCK and EGFR tyrosine kinases conform to the crystallographic binding mode of the inhibitors and accurately reproduce the key interactions formed in the active site.
|
Importantly, we have discovered that both active and inactive forms of EGFR can favorably interact with Erlotinib, and functionally relevant fluctuations of the inhibitor around the crystallographic conformation may accommodate to structurally different conformational states of the EGFR receptor (Fig. 4). These results reconcile the available experimental data, strongly indicating that binding specificity of Erlotinib with EGFR may be achieved by recognizing multiple conformational states of EGFR with high affinity, which is different from Imatinib, where specific binding to ABL is determined by recognizing a unique inactive conformational state of the enzyme. The results also suggest that Erlotinib specificity w. r. t. ABL and SRC kinases may be determined by the binding mechanism in which conformational adaptability of the inhibitor is not sufficient to adjust to the conformational variations seen in the inactive forms of ABL, CKIT and SRC kinases (Figs 2 and 4). The discovered structural and energetic differences in binding of Imatinib, Dasatinib, Erlotinib with an array of conformational states of protein kinases are consistent with the proteomics data and thereby may have pharmacological relevance in acquiring a specific signature of potent activities. A comparative sequence-structure analysis of the tyrosine kinase binding specificities with Dasatinib and Erlotinib reveals an excellent agreement between computational and experimental data (Figs 2, 3 and 4), thereby providing a useful insight into evolutionary and structural determinants of the binding specificity scenarios.
|
Accordingly, a functional classification of the binding specificity mechanisms observed for the tyrosine kinases may include specific recognition only with the unique inactive form of the target (highly selective Imatinib binding with ABL); specific recognition with the multiple conformational states of the target (selective Erlotinib inhibition of EGFR) and specific recognition of multiple conformational states of multiple targets from a kinase family (promiscuous Dasatinib binding with ABL, CKIT, SRC, LCK). The proposed functional annotation of the kinase binding mechanisms with the prominent cancer drugs points to a remarkable sequence-structure versatility of tyrosine kinases in achieving binding function, which extends beyond simple phylogenetic relationships of the binding-site residues.
|
4 CONCLUSIONS |
|---|
|
TOPABSTRACT1 INTRODUCTION2 SYSTEMS AND METHODS3 RESULTS AND DISCUSSION4 CONCLUSIONSREFERENCES |
|---|
The proposed functional classification of the kinase binding specificities explores mechanisms in which structural plasticity of the tyrosine kinases and sequence variation of the binding-site residues are linked with conformational preferences of the inhibitors in achieving effective drug binding. The molecular basis of binding specificity for tyrosine kinases may be largely driven by conformational adaptability of the inhibitors to an ensemble of structurally different conformational states of the enzyme, rather than being determined by their phylogenetic proximity in the kinome space or differences in the interactions with the variable binding-site residues. The insights into mechanisms of sequence-structure relationships in the kinome space and molecular basis of conformationally tolerant binding with the tyrosine kinases may have implications for understanding drug sensitivity. The presented computational approach may be employed to characterize activity signatures of small molecules against potential kinase targets and can also assist in rational redesign of the existing inhibitors to engineer novel activities and specificities.
Conflict of Interest: none declared.
|
FOOTNOTES |
|---|
Associate Editor: Anna Tramontano
Received on April 5, 2007; revised on April 5, 2007; accepted on May 16, 2007
|
REFERENCES |
|---|
|
TOPABSTRACT1 INTRODUCTION2 SYSTEMS AND METHODS3 RESULTS AND DISCUSSION4 CONCLUSIONSREFERENCES |
|---|
Bain J, et al. The specificities of protein kinase inhibitors: an update. Biochem. J. (2003) 371:199–204.[CrossRef][Web of Science][Medline]
Berman HM, et al. The protein data bank. Nucleic Acids Res. (2000) 28:235–242.
Beutler TC, et al. Avoiding singularities and numerical instabilities in free energy calculations based on molecular simulations. Chem. Phys. Lett. (1994) 222:529–539.[CrossRef][Web of Science]
Brown ER, Shepherd FA. Erlotinib in the treatment of non-small cell lung cancer. Expert Rev. Anticancer Ther. (2005) 5:767–775.[CrossRef][Web of Science][Medline]
Cornell WD, et al. A second generation force field for simulation of proteins, nucleic acids, and organic molecules. J. Am. Chem. Soc. (1995) 117:5179–5197.[CrossRef][Web of Science]
Deininger M, et al. The development of imatinib as a therapeutic agent for chronic myeloid leukemia. Blood (2005) 105:2640–2653.
Druker BJ. Imatinib as a paradigm of targeted therapies. Adv. Cancer Res. (2004) 91:1–30.[CrossRef][Web of Science][Medline]
Felsenstein J. PHYLIP: Phylogeny Inference Package, Version 3.6 (2002) Seattle, WA: University of Washington.
Hansmann UHE. Parallel tempering algorithm for conformational studies of biological molecules. Chem. Phys. Lett. (1997) 281:140–150.[CrossRef][Web of Science]
Hubbard SR, Till JH. Protein tyrosine kinase structure and function. Annu. Rev. Biochem. (2000) 69:73–98.
Hubbard SR. Protein tyrosine kinases: autoregulation and small-molecule inhibition. Curr. Opin. Struct. Biol. (2002) 12:735–741.[CrossRef][Web of Science][Medline]
Huse M, Kuriyan J. The conformational plasticity of protein kinases. Cell (2002) 109:226–282.
Johnson JM, et al. Protein family annotation in a multiple alignment viewer. Bioinformatics (2003) 19:544–545.
Kostich M, et al. Human members of the eukaryotic protein kinase family. Genome Biol. 3 (2002) 1–12. research0043.
Krupa A, Srinivasan N. The repertoire of protein kinases encoded in the draft version of the human genome: atypical variations and uncommon domain combinations. Genome Biol. (2002) 3. research0066.1–0066.
Kumar S, et al. Folding and binding cascades: dynamic landscapes and population shifts. Protein Sci. (2000) 9:10–19.[Web of Science][Medline]
Kwak EL, et al. Irreversible inhibitors of the EGF receptor may circumvent acquired resistance to gefitinib. Proc. Natl Acad. Sci. USA (2005) 102:7665–7670.
Levinson NM, et al. A SRC-like inactive conformation in the ABL tyrosine kinase domain. PLoS Biol. (2006) 4:753–767.[Web of Science]
Levy Y, et al. Protein folding topology determines binding mechanism. Proc. Natl Acad. Sci. USA (2004) 101:511–516.
Lydon NB, Duker BJ. Lessons learned from the development of imatinib. Leukemia Res. (2004) 28(Suppl. 1):29–38.[CrossRef]
Madhusudan S, Ganesan TS. Tyrosine kinase inhibitors in cancer therapy. Clin. Biochem. (2004) 37:618–635.[CrossRef][Web of Science][Medline]
Manning G, et al. The protein kinase complement of the human genome. Science (2002) 298:1912–1934.
Mayo SL, et al. DREIDING: a generic force field for molecular simulation. J. Phys. Chem. (1990) 94:8897–8909.[CrossRef][Web of Science]
Mohamadi F, et al. MacroModel-an integrated software system for modeling organic and bioorganic molecules using molecular mechanics. J. Comput. Chem. (1990) 11:440–467.[CrossRef][Web of Science]
Nagar B, et al. Crystal structures of the kinase domain of c-Abl in complex with the small molecule inhibitors PD173955 and imatinib (STI-571). Cancer Res. (2002) 62:4236–4243.
Saitou N, Nei M. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. (1987) 4:406–425.[Abstract]
Sawyer T, et al. Src inhibitors: genomics to therapeutics. Expert Opin. Investig. Drugs (2003) 10:1327–1344.[CrossRef]
Sawyer T, et al. Novel oncogenic protein kinase inhibitors for cancer therapy. Curr. Med. Chem. Anticancer Agents (2004) 4:449–455.[CrossRef][Medline]
Schindler T, et al. Structural mechanism for STI-571 inhibition of abelson tyrosine kinase. Science (2000) 289:1938–1942.
Shoemaker BA, et al. Speeding molecular recognition by using the folding funnel: the fly-casting mechanism. Proc. Natl Acad. Sci. USA (2000) 97:8868–8873.
Sridhar R, et al. Protein kinases as therapeutic targets. Pharm. Res (2000) 17:1345–1353.[CrossRef][Web of Science][Medline]
Stamos J, et al. Structure of the epidermal growth factor receptor kinase domain alone and in complex with a 4-anilinoquinazoline inhibitor. J. Biol. Chem. (2002) 277:46265–46272.
Still WC, et al. Semianalytical treatment of solvation for molecular mechanics and dynamics. J. Am. Chem. Soc. (1990) 112:6127–6129.[CrossRef][Web of Science]
Stouten PFW, et al. An effective solvation term based on atomic occupancies for use in protein simulations. Mol. Simulat. (1993) 10:97–120.[CrossRef]
Sugita Y, Okamoto Y. Replica-exchange molecular dynamics method for protein folding. Chem. Phys. Lett. (1999) 314:141–151.[CrossRef][Web of Science]
Thompson JD, et al. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. (1994) 22:4673–4680.
Tokarski JS, et al. The structure of Dasatinib (BMS-354825) bound to activated ABL kinase domain elucidates its inhibitory activity against imatinib-resistant ABL mutants. Cancer Res. (2006) 66:5790–5797.
Verkhivker GM. Imprint of evolutionary conservation and protein structure variation on the binding function of protein tyrosine kinases. Bioinformatics (2006) 22:1846–1854.
Verkhivker GM. Computational proteomics of biomolecular interactions in the sequence and structure space of the tyrosine kinome: deciphering the molecular basis of the kinase inhibitors selectivity. Proteins (2007) 66:912–929.[CrossRef][Web of Science][Medline]
Verkhivker GM, et al. Simulating disorder—order transitions in molecular recognition of unstructured proteins : where folding meets binding. Proc. Natl Acad. Sci. USA (2003) 100:5148–5153.
Verkhivker GM, et al. In silico profiling of tyrosine kinases binding specificity and drug resistance using Monte Carlo simulations with the ensembles of protein kinase crystal structures. Biopolymers (2006) 85:333–348.[CrossRef][Web of Science]
Vieth M, et al. Kinomics: characterizing the therapeutically validated kinase space. Drug Discov. Today (2005) 10:839–846.[CrossRef][Web of Science][Medline]
Vulpetti A, Bosotti R. Sequence and structural analysis of kinase ATP pocket residues. Il Farmaco (2004) 59:759–765.[CrossRef][Medline]
Wisniewski D, et al. Characterization of potent inhibitors of the Bcr-Abl and the c-Kit receptor tyrosine kinases. Cancer Res. (2002) 62:4244–4255.
Wong S, Witte ON. The BCR-ABL story: bench to bedside and back. Annu. Rev. Immunol. (2004) 22:247–306.[CrossRef][Web of Science][Medline]
Wood ER, et al. A unique structure for epidermal growth factor receptor bound to GW572016 (Lapatinib): relationships among protein conformation, inhibitor off-rate, and receptor activity in tumor cells. Cancer Res. (2004) 64:6652–6659.
Young MA, et al. Structure of the kinase domain of an imatinib-resistant Abl mutant in complex with the Aurora kinase inhibitor VX-680. Cancer Res. (2006) 6:1007–1014.
Zhang X, et al. An allosteric mechanism for activation of the kinase domain of epidermal growth factor receptor. Cell (2006) 125:1137–1149.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||