Розробка технології передбачення спорідненості органічних сполук до молекулярно-біологічних мішеней для рецепторно-орієнтованого скринінгу in silico

Ключевые слова: сродство, энергия взаимодействия, функция распределения состояний, лиганд, молекулярно-биологическая мишень, виртуальный скрининг.

SUMMARY

Yakovenko O. Ya. Development of technology of organic molecules to biomolecular targets affinity estimation in field of high-throughput receptor-based in silico screening. - Manuscript.

Thesis for Philosophy Doctor (PhD) Degree in Biology, speciality 03.00.20 - biotechnology. - Institute of Molecular Biology and Genetics of National Academy of Sciences of Ukraine, Kyiv, 2008.

The problem of computational affinity estimation has grown as computers became a powerful tool of overall today science. It is hardly to find any modern investigation that is not aided with computational models.

Unfortunately demands of scientists grow even faster than microelectronic industry capabilities. A lot of computational clusters make their affords to facilitate bio- and/or pharmacology projects. But queues of projects that require intense computations just elongate.

To cope with it researches have to use less accurate but computationally faster theoretical method in their projects. That is why new, more efficient and/or faster algorithms are of great interests in modern bioscience.

One of the key features that today computers are involved in by mathematical modeling in biology is a problem of interactions estimation between biomolecular target and small organic molecule. The knowledge of such interactions permits us to interrupt certain processes in the cells and, for example, stop such devastating disease as cancer. That is why a lot of affords are concentrated on design of better models and algorithm of high-throughput virtual screening technique.

The goal of this work is the development of new, efficient methods in field of receptor-oriented high-throughput virtual screening. This kind of screening is of especial interest during last few years, when the number of spatial 3D models of biomolecular targets significantly increased.

Our view of today receptor-based virtual screening is focused on three-component minimal interaction system, which consists of ligand (small organic molecule), biomolecular targer (huge biopolymer, usually protein, with known 3D structure) and surrounding solvent environment (water solution).

There are three kind of interactions in such minimal molecular system: ligand-receptor, ligand-water and receptor-water. The last term, however, is neglected due to small size of binding site that organic ligand occupy relatively to surface of the whole biopolymer globe. Receptor-ligand and ligand-water interactions were treated with new force field and modified explicit solvent GBSA methods correspondingly.

The Kirchhoff method for empirical atomic charges computations is the base of proposed here methods for interaction energy evaluation.

Proposed method was fitted against high level theory ab initio (HF/6-311++G(d,p)) dipole and quadrupole electric moments with simplex function optimization method. Obtained charges reproduce ab initio electric moments with errors in range of 0.2-0.4D and 0.1-0.2B correspondingly with only a few outliers, where resonance electrons distribution is significantly distorted.

Accurate charges obtained with Kirchhoff method permits us to upgrade GBSA explicit solvent model. Measured regression of desolving energies computationally predicted with modified GBSA methods for more that 400 various organic compounds of various chemical classes is as good as r=0.84.

To increase accuracy of ligand-receptor interaction energy evaluation the Kirchhoff charges calculation model was used for creating a new force field YFF. This force field is obtained by joining Van-der-Waals and bonded part of well-known MMFF94 with our charge calculation scheme. New field incorporates accurate functional known for MMFF but its coulomb part is improved from 0.8-1.0 D reported in MMFF94 to 0.2-04 D of our charges.

All together interactions energies are then used to calculate affinity of small organic molecule to binding site of huge biomolecular target in water surrounding.

The molecular states distribution functions of statistical physics are used fto account of entropy change during complex formation. The additional docking code is used to place ligand in the binding site of the receptor molecule with flexible docking method.

The Hessian matrix is formed than in accordance with potential energy functional of YFF for several best ligands poses found with docking core. Hessian evaluation is performed in the position of ligands energy minimum in the receptors cavity.

The integration is performed for internal and for rotational-translational freedom degreases as well. For better correspondence of predicted inhibition constants and experimentally derived ones the additional square root is taken from Hessian matrix determinant; this modification we called quasy-arbitrary rotational-translational distribution functions model. The special treatment procedure is applied to account internal freedom degreases of small organic ligand in binding site of huge receptor molecule.

With aim of speedup we applied neither internal nor rotational-translational freedom degreases for protein globe. Although the internal freedom degreases of key amino-acides of proteins binding site is under future consideration and our method does not prohibit it.

Rotational-translational degreases of protein globe are switched of to lower ambiguous complexity of distribution functions calculations. According to our view the rotational-translational degreases of freedom of receptor molecule hardly influence binding due to huge receptors surface are and water viscosity.

The accuracy of the method was tested with the set of experimentally resolved CK2 complexes with small organic inhibitors with esperimentally derived inhibition constants. The correlation between predicted and experimentally known Ki is r=0.96 in logarithmic scale units.

Virtual screening of 100 000 in-house chemical compounds with developed program suit resulted in identification of several new inhibitors of CK2. Virtual screening focused entire database to 0.1% of its size which was tested in biological screening.

Several high potent inhibitors were identified in biological tests of focused set and one new selective chemical class of tetrahalogenoisoindols inhibitors was identified. The best affinity of identifiend inhibitors of focused set was less that 0.1M. The virtual screening experiment of 90 000 compounds took nearly 14 days of 8 processors of combinatorial chemistry department cluster.

Key words: affinity, interaction energy, states distribution functions, ligand, biomolecular target, virtual screening.

Размещено на Allbest.ru