The synthesis and investigation of reagents, which can modify biopolymers at the specific location is one of the most important problems in Molecular Biology. One class of reagents has the form: R1(R2)N-CH2-CH2-Cl. These molecules are DNA/RNA alkylation agents, which can be "turned on" using light. The influence of electronegativity of R1 and R2 on the reactivity of these reagents has been studied. The linear correlation between the first ionization potential and reactivity of the compounds has been shown [1]. Similar results were also obtained for more complex compounds (X-Ph-)(R-)N-CH2-CH2-Cl [16].
A computer program for optimization of MNDO semiempirical
parameters was developed [5]. This program has
been used to find MNDO parameters for biologically important metals Zn [10] and Ca [15] (Bliznyuk & Voityuk), for Na
and Mg (Voityuk) and for K, Rb,
Cs (unpublished). The parameters obtained for Zn were used for investigation of
Zn complexes with H2O,
H-bonds play an important role in structure and reactivity of biopolymers. We started our work with modification of Burshtein's scheme to include S-H...X H-bonds [3, 6]. Unfortunately, the original scheme has some inherent shortcomings. A new approach MNDO/M has been developed and used for study of small complexes with H-bonds [9, 11, 12, 14, 26]. Investigation of influence of protonation on relative stability of nucleic acid base pairs was the next step taken. It is well known that there are many charged residues in enzymes, which are the proton donors. The obtained results show that protonation affects the complementary and non-complementary base pairs differently, and therefore may have considerable importance in the molecular mechanism of replication fidelity [13, 21, 24]. Our investigations of charged complexes with H-bonds showed that in some systems the position of H atom in H-bond is unstable and may be changed by small external perturbation. We choose the model system of DNA - enzyme interaction (G-ImH). It was demonstrated that attachment of H2O can lead to stabilization or destabilization of the complex and change dramatically the energy of proton transfer reaction [23, 25]. A summary of our and others' findings on the investigation of charged complexes with H-bonds has been published as a review [33].
It is well known that many reactions in enzymes are nucleophilic substitution reactions:
Nu + X=O -> Nu-X-O -> ... where X = P or C.
The first state of the process can be stabilized by formation of H-bond X=O...HY.
During the reaction the negative charge on O atom increases and the H-bond
energy increases as well. This was demonstrated on the model reaction in active
site of t-RNA-synthase (X=P) [18] and in a more general case (X=P,C) [20]. The difference
between water and enzyme environment was also outlined.
Interestingly, this was the first (to the best of my knowledge) demonstration how strong hydrogen bonds can facilitate enzyme reactions. In later years there were numerous attempts to study strong hydrogen bonds in enzyme but no simple explanation of their mechanism was provided.
The reactions of formation and hydrolysis of amide bonds are
very important in biochemistry. We studied four different model reactions:
(1) H2(NH2)C=O (formamide) + H2O -> H2COOH + NH3
(neutral molecules)
(2) H2(NH2)C=O + OH- -> H2COO- + NH3 (base catalysis)
(3) H2(NH2)C-OH+ + H2O -> H2COOH + NH4+ (acid catalysis)
(4) H2(NH3+)C=O + H2O -> H2COOH + NH4+ (acid catalysis)
All transition states were obtained by Baker's EF method. The influence of
water environment was simulated by modified point-dipole scheme. [22]
It is well known, that the strength of H-bonds in charged
systems is proportional to the difference in proton affinities (PA) of their
components. The evaluation of PAs is very important
in predicting strength of H-bonds in enzymes, on their models. We use MNDO
method to estimate PAs of DNA base pairs and in their
complexes [19].
A very interesting approach to accurate evaluation of absolute values of PA has
been used in ref. 34. The size of the
investigated systems (histidine and lysine aminoacids) was too big to use state of the art ab initio calculations. We use a combination
of semiempirical and ab initio methods.
An excellent agreement with experiment has been obtained [34].
My graduation specialization was Inorganic Chemistry. Some calculation using INDO approach of different Pd, Fe, Al complexes have been performed. The results have been used in interpretation of experimental photoelectron X-Ray spectra, Mossbauer spectra and in evaluation of thermochemical characteristics. Only part of these results have been published [8, 17]. The interest in Inorganic Chemistry was the main reason for paper 7, which showed that parameters derived by M. Dewar for MNDO are not suitable for calculation of inorganic compounds of Sn.
One of the very interesting problems in chemistry is enzyme catalysis. Unfortunately, QM methods are not applicable to systems of such size. On the other hand, the MM methods do not allow complete characterization of chemical reactions. We developed combined semiempirical / molecular mechanical approach to study these systems. The main feature is parameterization of this method to reproduce available experimental energies of H-bonds. The obtained results were surprisingly good [32].
A new approach to modeling the effect of solvation in molecular mechanics calculations has been developed [35]. This method is comparable in accuracy with finite-difference continuum models, while requiring significantly shorter execution time. The induced charges, dipoles, and quadrupoles at the solute's atom centres are estimated employing parameterized version of reaction field theory. The effective parameterization term allows computation of an induced electrostatic multipoles in one iteration. The obtained multipoles are positioned on the atoms' centres, which permit calculation of desirable contributions to solvation energy, electrostatic field, and Coulomb interaction of atoms at the same time. The nonelectrostatic parts are assumed to be proportional to the molecular surface area. A newly developed fast and accurate numerical method for molecular surface area computation has been utilized [36,37]. Pilot studies of small molecule solvation energies, using the PARSE charges, show good accuracy with available experimental data. The additivity of electrostatic contributions of atomic centres allows fast computation of solvation term for molecules with partially fixed coordinates. This makes this model especially attractive for application to docking.
Very simple method of locating possible ligand binding site on the surface of proteins has been developed [40] and used together with the solvation model for investigating different DHFR structures [38]
The idea of lock and key type of binding is widely used as an explanation of enzyme specificity. However, as our calculations demonstrate [52] the lock and key type binding is not the only way enzymes can facilitate chemical reactions. Loose-binding is also good. It appears that loose binding maybe an evolutionally first step towards more specific key-binding, and is also very useful for drug-resistance.
One of the problems why current methods can’t estimate docking energies efficiently is that MM force-fields are simply not good enough. As our computations showed [46, 49, 50] neglect in polarization in current MM methods may lead to large (upto 50 kcal/mol) errors in electrostatic potentials. The problem is that MM parameters are optimized in small molecule calculations. In large molecules (like proteins) nearby atoms disturbed electronic densities, that leads to large errors for fixed-charge models. Many unsuccessful attempts to develop polarizable force-fields make me to believe that simplified QM methods are necessary for accurate description of electrostatics in large molecules. Towards that idea we had a look at the ability of current semiempirical methods to reproduce known binding energies [50].
We use semiempirical methods extensively to calculate some interesting reactions in collaboration with experimentalists [28] and to show how inaccurate location of transition state may lead to a wrong conclusion [30]. The ab initio investigation of hypothetical Ih molecule N20 has been performed [31]. It was shown that such molecule is stable and can be (we hope) found by experiment.
I was very fortunate to have an opportunity to participate in writing of large reviews on the subjects that were of much interest to me:
I started to write programs when I was a third year student and found it fascinating. Self-education aided me to obtain a job as system operator, and after graduation I was invited to design and teach the course - "Introduction to Computer Programming" for undergraduates of the Natural Sciences Department. So far two major projects have been more or less completed: MNDO-85 (semiempirical methods) (Fortran, 30,000 lines) and PWN program (molecular mechanics) (C, 50,000 lines); one medium-sized project (determination of point group symmetry of a molecule, adjustment of coordinates, and generation of independent set of optimization coordinates) (C, 10,000 lines); and a large number of smaller programs (Fortran, C, PL/I) have been written. I also had some experience in transferring computational chemistry codes between different operational systems / Fortran dialects. This included early versions of Hondo, Mopac and Amber. Recently, I was involved in Java programming of Computational Chemistry Portal in APAC grid. (Will write more when the project finishes, hopefully, middle of 2007).
In 1997-2003 I was working as computational chemist at Super Computer Facility of the Australian National University (ANUSF). My main responsibility was maintaining and developing Mopac program. Some of the work we done can be found in refs. [39] and [45]. Most of the work, however, could not be published.
Among other things, the work included
We also have done some interesting applications of the linear-scaling semiempirical methods: