Current virtual screening tools are fast, but reliable scoring is elusive. Here, we present the '... more Current virtual screening tools are fast, but reliable scoring is elusive. Here, we present the 'SQM/COSMO filter', a novel scoring function featuring a quantitative semiempirical quantum mechanical (SQM) description of all types of noncovalent interactions coupled with implicit COSMO solvation. We show unequivocally that it outperforms eight widely used scoring functions. The accuracy and chemical generality of the SQM/COSMO filter make it a perfect tool for late stages of virtual screening. Despite the enormous advances in method development for structure-based in silico drug design, reliable predictions of the structures (docking) and affinities (scoring) of protein–ligand (P–L) complexes still remain an unsolved task. 1 A plethora of scoring functions (SFs) have been devised by utilising experimental data for regression analyses, by constructing knowledge-based potentials, or based on physical laws. 2,3 As none of the SFs is general enough to perform equally strongly for a diverse set of P–L complexes, utilising several SFs at once (consensus scoring) holds promise. 4 Regression analysis and knowledge-based approaches to scoring are trained on a set of P–L complexes and rely on variable master equation terms. Their validity is limited to complexes similar to the training set. In principle, this problem has been overcome in physics-based methods. Because of computational cost, preference has been given to molecular mechanics (MM) methods, such as the combination of MM interaction energies with implicit solvation free energy terms (generalised Born, GB, or Poisson–Boltzmann, PB) to estimate affinities. 2 Additionally, the wide coverage of organic chemical space in the GAFF (general AMBER force field) 5 has made the parameterisation of ligands for MM straightforward. However, an explicit description of quantum mechanical (QM) effects in P–L interactions, such as charge transfer, polarisa-tion, covalent-bond formation or s-hole bonding, was missing. QM methods, which describe these effects qualitatively better than the energy functions used in MM-based SFs, were thus introduced into computational drug design. 6,7 Recent developments in QM methods and algorithms as well as the availability of a powerful computing infrastructure have paved the way to apply them for P–L complexes in numerous setups: linear scaling or efficient parallelisation of semi-empirical QM (SQM) methods, 7–10 QM/MM, 7,8,11,12 DFT-D3 on truncated P–L complexes 13 or various fragmentation methods. 11,14 Specifically, AM1, RM1, PM6 or DF–TB SQM methods have been used 7–9,12,15 as such or with empirical corrections for dispersion, hydrogen-and halogen-bonding 16 to describe the P–L noncovalent interactions. Merz et al. pioneered this area by introducing a QM-based SF (QMScore), a combination of the AM1 SQM method with an empirical dispersion (D) and the PB implicit solvent [eqn (1)]. 17 The method was useful for describing metalloprotein–ligand binding, but further corrections were needed, especially for a quantitative treatment of dispersion and hydrogen bonding. 10 Score = DE int + DDG solv + DG 0w conf À TDS (1) The above equation is a general physics-based SF. The terms are the gas-phase interaction energy (DE int), the change of solvation free energy upon complex formation (DDG solv), the change of conformational 'free' energy (DG 0w conf) and the change of entropy upon ligand binding (ÀTDS). Our approach is systematic. Using accurate calculations in small model systems as a benchmark, we developed corrections for SQM methods that provide reliable and accurate description of a wide range of noncovalent interactions including dispersion , hydrogen-and halogen-bonding. 16 Coupled with the PM6 SQM method, 18 the resulting PM6-D3H4X approach is applicable to a wide chemical space and does not require any
Carborane-based compounds are promising lead structures for development of inhibitors of carbonic... more Carborane-based compounds are promising lead structures for development of inhibitors of carbonic anhydrases (CAs). Here, we report structural and computational analysis applicable to structure-based design of carborane compounds with selectivity toward the cancer-specific CAIX isoenzyme. We determined the crystal structure of CAII in complex with 1-methylenesulfamide-1,2-dicarba-closo-dodecaborane at 1.0 Å resolution and used this structure to model the 1-methylenesulfamide-1,2-dicarba-closo-dodecaborane interactions with CAIX. A virtual glycine scan revealed the contributions of individual residues to the energy of binding of 1-methylenesulfamide-1,2-dicarba-closo-dodecaborane to CAII and CAIX, respectively.
Overactivation of NMDA receptors has been implicated in various neuropathological conditions, inc... more Overactivation of NMDA receptors has been implicated in various neuropathological conditions, including brain ischaemia, neurodegenerative disorders and epilepsy. Production of d-serine, an NMDA receptor co-agonist, from l-serine is catalyzed in vivo by the pyridoxal-5'-phosphate (PLP)-dependent enzyme serine racemase. Specific inhibition of this enzyme has been proposed as a promising strategy for treatment of neurological conditions caused by NMDA receptor dysfunction. Here we present the synthesis and activity analysis of a series of malonate-based inhibitors of mouse serine racemase (mSR). The compounds possessed IC50 values ranging from 40 ± 11 mM for 2,2-bis(hydroxymethyl)malonate down to 57 ± 1 μM for 2,2-dichloromalonate, the most effective competitive mSR inhibitor known to date. The structure-activity relationship of the whole series in the human orthologue (hSR) was interpreted using Glide docking, WaterMap analysis of hydration and quantum mechanical calculations based on the X-ray structure of the hSR/malonate complex. Docking into the hSR active site with three thermodynamically favourable water molecules was able to discern qualitatively between good and weak inhibitors. Further improvement in ranking was obtained using advanced PM6-D3H4X/COSMO semiempirical quantum mechanics-based scoring which distinguished between the compounds with IC50 better/worse than 2 mM. We have thus not only found a new potent hSR inhibitor but also worked out a computer-assisted protocol to rationalize the binding affinity which will thus aid in search for more effective SR inhibitors. Novel, potent hSR inhibitors may represent interesting research tools as well as drug candidates for treatment of diseases associated with NMDA receptor overactivation.
Proteins: Structure, Function, and Bioinformatics, 2008
We are proposing an interresidue interaction energy map (IEM)--a new tool for protein structure a... more We are proposing an interresidue interaction energy map (IEM)--a new tool for protein structure analysis and protein bioinformatics. This approach employs the sum of pair-wise interaction energies of a particular residue as a measure of its structural importance. We will show that the IEM can serve as a means for identifying key residues responsible for the stability of a protein. Our method can be compared with the interresidue contact map but has the advantage of weighting the contacts by the stabilization energy content which they bring to the protein structure. For the theoretical adjustment of the proposed method, we chose the Trp-cage mini protein as a model system to compare a spectrum of computational methods ranging from the ab initio MP2 level through the DFT method to empirical force-field methods. The IEM method correctly identifies Tryptophane 6 as the key residue in the Trp-cage. The other residues with the highest stabilizing contributions correspond to the structurally important positions in the protein. We have further tested our method on the Trp2Cage miniprotein--a P12W mutant of the Trp-cage and on two proteins from the rubredoxin family that differ in their thermostability. Our method correctly identified the thermodynamically more stable variants in both cases and therefore can also be used as a tool for the relative measurement of protein stability. Finally, we will point out the important role played by dispersion energy, which contributes significantly to the total stabilization energy and whose role in aromatic pairs is clearly dominant. Surprisingly, the dispersion energy plays an even more important role in the interaction of prolines with aromatic systems.
ABSTRACT The folded structures of RNA molecules and large ribonucleoprotein particles are stabili... more ABSTRACT The folded structures of RNA molecules and large ribonucleoprotein particles are stabilized by a wide range of base pairs that actively utilize the 2′-OH groups of ribose for base pairing. Such base pairing does not occur in DNA and is essential for functional RNAs. We report reference quantum chemical calculations of base pairing energies for a representative selection of 25 RNA base pairs utilizing the ribose moiety for base pairing, including structures with amino acceptor interactions. All base pairs are evaluated at the MP2 level with extrapolation to the complete basis set (CBS) of atomic orbitals. CCSD(T) correction terms were obtained for four base pairs. In addition, the base pairing is evaluated using the DFT-SAPT perturbational procedure along with the aug-cc-pVDZ basis set, which allows for the decomposition of the interaction energies into separate, physically meaningful, components. These calculations confirm that, compared to canonical base pairs, many RNA base pairs exhibit a modestly increased role of dispersion attraction compared to canonical base pairs. However, the effect is smaller than one would assume based on assessment of the ratio of HF and correlation components of the interaction energies. Interaction energies are further calculated using the SCS(MI)-MP2 and DFT-D methods. Finally, we estimate the effect of aqueous solvent screening on the base pairing stability using the continuum solvent approach.
ABSTRACT tThe frequent use of pesticides in crop protection leads to their accumulation in soil a... more ABSTRACT tThe frequent use of pesticides in crop protection leads to their accumulation in soil as well as to theformation of their complexes with ions present in soil. The formation of such complexes can change themobility of both the pesticides and the ions as well as their ability to penetrate biological membranes andenter biota. The problem of triazole-fungicide complexation with dications is demonstrated on the caseof zinc and cadmium ions (M). Complexation with tebuconazole (1) has been studied using electrosprayionization mass spectrometry and quantum mechanical DFT-based computations. These experimentshave revealed that the deprotonation of 1 leads to the stabilization of the complexes formed, which hasbeen confirmed by our computations. In terms of stability, two types of complexes have been described:(i) a less stable complex (approx. 125 kJ mol−1) with neutral 1 [M(1)2,3,4]2+and [MCl(1)1,2]+, and (ii) a morestable complex (approx. 220 kJ mol−1) with deprotonated 1 [M(1)n(1-H)]+(n = 0,1). The high stability ofthe complexes with deprotonated 1 is given by their stabilization via a metal interaction with the oxygenatom or phenyl ring. The structure of [M(1-H)]+is very compact when compared to the structure of parent1.
The intrinsic stability of Watson-Crick d(AT) and r(AU) hydrogen bonds was analyzed by employing ... more The intrinsic stability of Watson-Crick d(AT) and r(AU) hydrogen bonds was analyzed by employing a variety of quantum-mechanical techniques, such as energy calculations, determination of reactivity indexes, and analysis of electron density topology. The analyses were performed not only for equilibrium gas-phase geometries, but also on hundreds of conformations derived from molecular dynamics (MD) and database analysis. None of our results support the idea that r(AU) hydrogen bonds are intrinsically more stable than those of d(AT). Instead, our data are in accordance with the traditional view that the greater stability of RNA relative to DNA is attributable to a variety of effects (e.g., stacking, sugar puckering, solvation) rather than to a significant difference in the hydrogen bonding of DNA and RNA base pairs.
Current virtual screening tools are fast, but reliable scoring is elusive. Here, we present the '... more Current virtual screening tools are fast, but reliable scoring is elusive. Here, we present the 'SQM/COSMO filter', a novel scoring function featuring a quantitative semiempirical quantum mechanical (SQM) description of all types of noncovalent interactions coupled with implicit COSMO solvation. We show unequivocally that it outperforms eight widely used scoring functions. The accuracy and chemical generality of the SQM/COSMO filter make it a perfect tool for late stages of virtual screening. Despite the enormous advances in method development for structure-based in silico drug design, reliable predictions of the structures (docking) and affinities (scoring) of protein–ligand (P–L) complexes still remain an unsolved task. 1 A plethora of scoring functions (SFs) have been devised by utilising experimental data for regression analyses, by constructing knowledge-based potentials, or based on physical laws. 2,3 As none of the SFs is general enough to perform equally strongly for a diverse set of P–L complexes, utilising several SFs at once (consensus scoring) holds promise. 4 Regression analysis and knowledge-based approaches to scoring are trained on a set of P–L complexes and rely on variable master equation terms. Their validity is limited to complexes similar to the training set. In principle, this problem has been overcome in physics-based methods. Because of computational cost, preference has been given to molecular mechanics (MM) methods, such as the combination of MM interaction energies with implicit solvation free energy terms (generalised Born, GB, or Poisson–Boltzmann, PB) to estimate affinities. 2 Additionally, the wide coverage of organic chemical space in the GAFF (general AMBER force field) 5 has made the parameterisation of ligands for MM straightforward. However, an explicit description of quantum mechanical (QM) effects in P–L interactions, such as charge transfer, polarisa-tion, covalent-bond formation or s-hole bonding, was missing. QM methods, which describe these effects qualitatively better than the energy functions used in MM-based SFs, were thus introduced into computational drug design. 6,7 Recent developments in QM methods and algorithms as well as the availability of a powerful computing infrastructure have paved the way to apply them for P–L complexes in numerous setups: linear scaling or efficient parallelisation of semi-empirical QM (SQM) methods, 7–10 QM/MM, 7,8,11,12 DFT-D3 on truncated P–L complexes 13 or various fragmentation methods. 11,14 Specifically, AM1, RM1, PM6 or DF–TB SQM methods have been used 7–9,12,15 as such or with empirical corrections for dispersion, hydrogen-and halogen-bonding 16 to describe the P–L noncovalent interactions. Merz et al. pioneered this area by introducing a QM-based SF (QMScore), a combination of the AM1 SQM method with an empirical dispersion (D) and the PB implicit solvent [eqn (1)]. 17 The method was useful for describing metalloprotein–ligand binding, but further corrections were needed, especially for a quantitative treatment of dispersion and hydrogen bonding. 10 Score = DE int + DDG solv + DG 0w conf À TDS (1) The above equation is a general physics-based SF. The terms are the gas-phase interaction energy (DE int), the change of solvation free energy upon complex formation (DDG solv), the change of conformational 'free' energy (DG 0w conf) and the change of entropy upon ligand binding (ÀTDS). Our approach is systematic. Using accurate calculations in small model systems as a benchmark, we developed corrections for SQM methods that provide reliable and accurate description of a wide range of noncovalent interactions including dispersion , hydrogen-and halogen-bonding. 16 Coupled with the PM6 SQM method, 18 the resulting PM6-D3H4X approach is applicable to a wide chemical space and does not require any
Carborane-based compounds are promising lead structures for development of inhibitors of carbonic... more Carborane-based compounds are promising lead structures for development of inhibitors of carbonic anhydrases (CAs). Here, we report structural and computational analysis applicable to structure-based design of carborane compounds with selectivity toward the cancer-specific CAIX isoenzyme. We determined the crystal structure of CAII in complex with 1-methylenesulfamide-1,2-dicarba-closo-dodecaborane at 1.0 Å resolution and used this structure to model the 1-methylenesulfamide-1,2-dicarba-closo-dodecaborane interactions with CAIX. A virtual glycine scan revealed the contributions of individual residues to the energy of binding of 1-methylenesulfamide-1,2-dicarba-closo-dodecaborane to CAII and CAIX, respectively.
Overactivation of NMDA receptors has been implicated in various neuropathological conditions, inc... more Overactivation of NMDA receptors has been implicated in various neuropathological conditions, including brain ischaemia, neurodegenerative disorders and epilepsy. Production of d-serine, an NMDA receptor co-agonist, from l-serine is catalyzed in vivo by the pyridoxal-5'-phosphate (PLP)-dependent enzyme serine racemase. Specific inhibition of this enzyme has been proposed as a promising strategy for treatment of neurological conditions caused by NMDA receptor dysfunction. Here we present the synthesis and activity analysis of a series of malonate-based inhibitors of mouse serine racemase (mSR). The compounds possessed IC50 values ranging from 40 ± 11 mM for 2,2-bis(hydroxymethyl)malonate down to 57 ± 1 μM for 2,2-dichloromalonate, the most effective competitive mSR inhibitor known to date. The structure-activity relationship of the whole series in the human orthologue (hSR) was interpreted using Glide docking, WaterMap analysis of hydration and quantum mechanical calculations based on the X-ray structure of the hSR/malonate complex. Docking into the hSR active site with three thermodynamically favourable water molecules was able to discern qualitatively between good and weak inhibitors. Further improvement in ranking was obtained using advanced PM6-D3H4X/COSMO semiempirical quantum mechanics-based scoring which distinguished between the compounds with IC50 better/worse than 2 mM. We have thus not only found a new potent hSR inhibitor but also worked out a computer-assisted protocol to rationalize the binding affinity which will thus aid in search for more effective SR inhibitors. Novel, potent hSR inhibitors may represent interesting research tools as well as drug candidates for treatment of diseases associated with NMDA receptor overactivation.
Proteins: Structure, Function, and Bioinformatics, 2008
We are proposing an interresidue interaction energy map (IEM)--a new tool for protein structure a... more We are proposing an interresidue interaction energy map (IEM)--a new tool for protein structure analysis and protein bioinformatics. This approach employs the sum of pair-wise interaction energies of a particular residue as a measure of its structural importance. We will show that the IEM can serve as a means for identifying key residues responsible for the stability of a protein. Our method can be compared with the interresidue contact map but has the advantage of weighting the contacts by the stabilization energy content which they bring to the protein structure. For the theoretical adjustment of the proposed method, we chose the Trp-cage mini protein as a model system to compare a spectrum of computational methods ranging from the ab initio MP2 level through the DFT method to empirical force-field methods. The IEM method correctly identifies Tryptophane 6 as the key residue in the Trp-cage. The other residues with the highest stabilizing contributions correspond to the structurally important positions in the protein. We have further tested our method on the Trp2Cage miniprotein--a P12W mutant of the Trp-cage and on two proteins from the rubredoxin family that differ in their thermostability. Our method correctly identified the thermodynamically more stable variants in both cases and therefore can also be used as a tool for the relative measurement of protein stability. Finally, we will point out the important role played by dispersion energy, which contributes significantly to the total stabilization energy and whose role in aromatic pairs is clearly dominant. Surprisingly, the dispersion energy plays an even more important role in the interaction of prolines with aromatic systems.
ABSTRACT The folded structures of RNA molecules and large ribonucleoprotein particles are stabili... more ABSTRACT The folded structures of RNA molecules and large ribonucleoprotein particles are stabilized by a wide range of base pairs that actively utilize the 2′-OH groups of ribose for base pairing. Such base pairing does not occur in DNA and is essential for functional RNAs. We report reference quantum chemical calculations of base pairing energies for a representative selection of 25 RNA base pairs utilizing the ribose moiety for base pairing, including structures with amino acceptor interactions. All base pairs are evaluated at the MP2 level with extrapolation to the complete basis set (CBS) of atomic orbitals. CCSD(T) correction terms were obtained for four base pairs. In addition, the base pairing is evaluated using the DFT-SAPT perturbational procedure along with the aug-cc-pVDZ basis set, which allows for the decomposition of the interaction energies into separate, physically meaningful, components. These calculations confirm that, compared to canonical base pairs, many RNA base pairs exhibit a modestly increased role of dispersion attraction compared to canonical base pairs. However, the effect is smaller than one would assume based on assessment of the ratio of HF and correlation components of the interaction energies. Interaction energies are further calculated using the SCS(MI)-MP2 and DFT-D methods. Finally, we estimate the effect of aqueous solvent screening on the base pairing stability using the continuum solvent approach.
ABSTRACT tThe frequent use of pesticides in crop protection leads to their accumulation in soil a... more ABSTRACT tThe frequent use of pesticides in crop protection leads to their accumulation in soil as well as to theformation of their complexes with ions present in soil. The formation of such complexes can change themobility of both the pesticides and the ions as well as their ability to penetrate biological membranes andenter biota. The problem of triazole-fungicide complexation with dications is demonstrated on the caseof zinc and cadmium ions (M). Complexation with tebuconazole (1) has been studied using electrosprayionization mass spectrometry and quantum mechanical DFT-based computations. These experimentshave revealed that the deprotonation of 1 leads to the stabilization of the complexes formed, which hasbeen confirmed by our computations. In terms of stability, two types of complexes have been described:(i) a less stable complex (approx. 125 kJ mol−1) with neutral 1 [M(1)2,3,4]2+and [MCl(1)1,2]+, and (ii) a morestable complex (approx. 220 kJ mol−1) with deprotonated 1 [M(1)n(1-H)]+(n = 0,1). The high stability ofthe complexes with deprotonated 1 is given by their stabilization via a metal interaction with the oxygenatom or phenyl ring. The structure of [M(1-H)]+is very compact when compared to the structure of parent1.
The intrinsic stability of Watson-Crick d(AT) and r(AU) hydrogen bonds was analyzed by employing ... more The intrinsic stability of Watson-Crick d(AT) and r(AU) hydrogen bonds was analyzed by employing a variety of quantum-mechanical techniques, such as energy calculations, determination of reactivity indexes, and analysis of electron density topology. The analyses were performed not only for equilibrium gas-phase geometries, but also on hundreds of conformations derived from molecular dynamics (MD) and database analysis. None of our results support the idea that r(AU) hydrogen bonds are intrinsically more stable than those of d(AT). Instead, our data are in accordance with the traditional view that the greater stability of RNA relative to DNA is attributable to a variety of effects (e.g., stacking, sugar puckering, solvation) rather than to a significant difference in the hydrogen bonding of DNA and RNA base pairs.
Uploads
Papers by Pavel Hobza