JP2005018447A - Method for searcing acceptor-ligand stable complex structure - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To develop a method to search an acceptor-ligand stable complex structure using a computer which realizes not only normal virtual screening but also the search of an acceptor to be stably bonded with a ligand by developiong a method to automatically search the stable acceptor-ligand stable complex structure only by giving the three dimensional structure of the ligand and then acceptor. <P>SOLUTION: The position of a bonding pocket in an acceptor is presumed, the bond mode of a ligand to the bond pocket is replaced with a simplified pharmacophore model and the various conformations of the ligand preliminarily generated by a computing chemical method are continuously compared to see whether or not the conformations suit this model, so that the structure of the ligand supposed to form when the acceptor and the ligand forms the stable complex can be presumed in a short period of time together with the bond position. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

【００５４】
【発明の効果】
本発明で開示された新規な探索方法を用いることにより、結合部位が特定されていない受容体に対してさえも、短時間でリガンドとの安定複合体構造を特定することが可能となった。したがって、既知の受容体に対する新規リガンドの開発のみならず、既知のリガンドが他の受容体と安定複合体を形成するかどうかを検討することにより、新たな薬理活性物質を見出し、あるいは副作用や毒性をもたらす可能性を事前に推定することができるため、創薬研究のコストと時間を大幅に改善することが可能となった。
【図面の簡単な説明】
【図１】処理のフローチャートを示す図。
【図２】第１の工程で発生させた多様なリガンド配座の例を示す図。
【図３】第２の工程で得られたダミー原子と受容体の分子モデルを示す図。実線はビオチン結合タンパク質、球はビオチン、小球は結合候補ポケットを示すダミー原子の中心であり、結合ポケットの大きさを Ｃｏｎｎｏｌｌｙ の解析表面（黒色の空間）で示す。
【図４】第３、第４の工程で得られたリガンド原子が存在し得るファーマコフォア領域と排除体積を示す球を示す図。
【図５】第５の工程で得られた複合体構造の候補（棒球モデル）の例を示す図。
【図６】第６の工程で得られた最安定複合体の構造を示す図。
【図７】第６の工程で得られた最安定複合体中のリガンド ｂｉｏｔｉｎ （棒球表示）と、Ｘ結晶構造解析で決定されたリガンド（棒表示）の重ね合わせを示す図。
【図８】アセチルコリンエステラーゼ（ＰＤＢ エントリー １ＡＣＬ）について得られた最安定複合体中のリガンド Ｄｅｃａｍｅｔｈｏｎｉｕｍ （棒球表示）と、Ｘ結晶構造解析で決定されたリガンド（棒表示）の重ね合わせを示す図。
【図９】レチノール結合タンパク質（ＰＤＢ エントリー １ＡＱＢ）について得られた最安定複合体中のリガンド Ｒｅｔｉｎｏｌ （棒球表示）と、Ｘ結晶構造解析で決定されたリガンド（棒表示）の重ね合わせを示す図。
【図１０】グルタミン酸受容体サブユニット２（ＰＤＢ エントリー １ＦＴＭ）について得られた最安定複合体中のリガンド（ＡＭＰＡ， 棒球表示）と、Ｘ結晶構造解析で決定されたリガンド（棒表示）の重ね合わせを示す図。
【図１１】ＨＩＶ−１逆転写酵素（ＰＤＢ エントリー １ＨＮＶ）について得られた最安定複合体中のリガンド ＴＩＢＯ Ｒ８６１８３ （棒球表示）と、Ｘ結晶構造解析で決定されたリガンド（棒表示）の重ね合わせを示す図。
【図１２】核内受容体（ＰＤＢ エントリー ３ＥＲＴ）について得られた最安定複合体中のリガンド４−ｈｙｄｒｏｘｙｔａｍｏｘｉｆｅｎ （ＯＨＴ， 棒球表示）と、Ｘ結晶構造解析で決定されたリガンド（棒表示）の重ね合わせを示す図。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method using a computer for searching for a stable complex structure of a receptor and a ligand, which can be used for molecular design of physiologically active substances such as pharmaceuticals and agricultural chemicals. The present invention also relates to a program for causing an information processing apparatus to execute such a method.
[0002]
[Prior art]
Many drugs and physiologically active substances express their pharmacological activity or toxicity by binding to a biopolymer (for example, protein, nucleic acid and complex containing them) serving as a receptor at a specific site. Compounds such as relatively low molecular weight drugs, physiologically active substances, and inhibitors that bind to such receptors are generally called ligands, but there are many receptors whose ligands have not been clarified. . For such a receptor, efficiently finding a ligand capable of forming a stable complex with the receptor and exhibiting a medicinal effect is an important basis in drug discovery research.
[0003]
With regard to the three-dimensional structure of biopolymers that are candidates for receptors, the structure can be determined in a shorter period of time than before due to advances in analytical instruments such as X-rays and NMR and software for analyzing the results. And registered in a database such as a protein data bank (see, for example, Non-Patent Document 1). On the other hand, deciphering the entire base sequence of a gene has become a reality, and due to advances in molecular modeling technology using computers, the predicted amino acids of biological macromolecules whose experimental three-dimensional structure has not been determined experimentally A certain amount of three-dimensional structure can be estimated from the arrangement or the like.
[0004]
As a method using a computer to search for a ligand for the receptor, the atomic coordinates of the receptor and the ligand are input, an approximate binding site in the receptor is designated, and the computer is used to connect the ligand to the receptor. By finding a reasonable binding mode and applying this calculation to multiple ligands in succession, it is possible to find a ligand that may bind or to screen out compounds that do not bind. Widely known. These calculation methods are generally called docking studies, and are also called virtual screening when a large number of ligands are compared at high speed at the expense of a certain degree of accuracy. Research methods for designing ligand molecules based on the three-dimensional structure of such receptors are also collectively referred to as structure-dependent drug discovery.
[0005]
In order to apply these methods to real molecular systems, the development of an algorithm that satisfies the two criteria of efficient search of coordinate space (i) and evaluation criterion (ii) for evaluating the validity of the binding mode Are essential, and many methods have been proposed for each of them (for example, see Non-Patent Document 2 or Patent Document 1). Some of them are disclosed or marketed as software packages (Tripos's FlexX (trade name), Accelrys' C2-LigandFit (trade name), Schrodinger's Glide (trade name), etc.).
[0006]
However, considering not only the complex conformation of the ligand molecule but also the degree of freedom including the relative position where the molecule can exist in the receptor, the degree of freedom of the space to be searched in (i) is enormous. It is not realistic to conduct a complete exhaustive search even with the current computational capability, and there has not yet been a complete evaluation function for quantifying the criteria of (ii) .
[0007]
In general, docking studies presume the position of the binding pocket and then search within that limited space to estimate a stable complex structure, but if the binding pocket is unknown, There is no known method that can automatically determine what conformation of a ligand exists at a position. In addition, since the structure of the receptor is usually much larger than that of the ligand, not only the degree of freedom of the structure of the ligand but also the position where the ligand can bind in the receptor should be searched. However, it is impossible to carry out with high accuracy and within a realistic time only by applying an algorithm used in conventional structure-dependent drug discovery.
[0008]
There is INVDOCK (for example, refer to Patent Document 2) as a method for searching for a specific ligand by expressing its conformation as a vector and matching a vector created from the shape of a depression of protein or nucleic acid. However, in general, the ligand does not fit exactly one-to-one in the receptor binding pocket, but in most cases only a part of the ligand overlaps a part of the ligand. For this reason, this approach can be applied when the degree of coincidence is high, but in more general cases where some of the ligand's functional groups are in the proper spatial arrangement on the receptor surface, including the depression. Was not suitable.
[0009]
Furthermore, the initial spatial arrangement is determined with a vector representing the ligand, energy optimization is performed, and a commercially available program based on the same algorithm that optimizes the structure after matching the shape of the ligand and the binding pocket is made by Accelrys. C2 · LigandFit ((trade name) (for example, see Non-Patent Document 3). However, the correlation between the binding pocket and the shape of the ligand is originally low, and if the chemical characteristics of the ligand are ignored, it is very In addition, since it is necessary to perform structure optimization from an initial structure having a low degree of fitness, a lot of useless calculations are required.
[0010]
On the other hand, although it is different from the general concept of structure-dependent drug discovery, there is a possibility of binding to the same receptor based on one or more known ligands without knowing the conformation of the corresponding receptor. Methods for finding new ligands were called pharmacophore model search and receptor surface model search. The pharmacophore model search extracts parts (features) that are considered to have chemical properties that strongly interact with the receptor, such as hydrogen-bonded substituents and ionic atoms, in the ligand. A simple molecular model (pharmacophore) consisting of types and distances between them is defined, and molecules that match the model are searched for. It is a technique that can find a ligand that can bind to the target. The receptor surface model is a method for estimating the shape and electrostatic characteristics of the surface covering the whole molecule from the molecular structure of the ligand, and searching for a ligand that can exist stably on the surface. With regard to these methods, for example, in pharmacophore search, CATLYST (trade name) manufactured by Accelrys (for example, refer to Non-Patent Document 4), UNITY (trade name) manufactured by Tripos (for example, refer to Non-Patent Document 5), For example, MOE (trade name) manufactured by Chemical Computing Group (see, for example, Non-Patent Document 6), and C2-Receptor (trade name) manufactured by Accelrys (for example, see Non-Patent Document 7 or 8) are used as receptor surface models. It is commercially available.
[0011]
However, both the pharmacophore model search and the receptor surface model search are methods that use a three-dimensional quantitative structure-activity relationship to find new ligands having properties similar to those of known ligands. It did not deal directly with the interaction.
[0012]
[Patent Document 1] Japanese Patent No. 2621842
[Patent Document 2] US Pat. No. 6,519,611
[Non-Patent Document 1] Provided by Osaka University, Protein Data Bank Japan, [online], [Search June 20, 2003], Internet <URL: http: // pdbj. protein. osaka-u. ac. jp />
[Non-patent Document 2] DesJarlais R. L. , Sheridan R., et al. P. , Dixon J. S. Kuntz I .; D. , Venkataraghavan R .; , J. et al. Med. Chem. Vol. 29, p2149-2153 (1986)
[Non-Patent Document 3] Peters, K. et al. P. , Fauck, J .; Frommel, C.I. , Journal of Molecular Biology, vol. 256, p201-213 (1996)
[Non-Patent Document 4] Barnum, D. et al. Greene, J .; , Smallie, A .; , Sprague, P.M. , J. et al. Chem. Inf. Comput. Sci. vol. 36, p563-571 (1996)
[Non-Patent Document 5] Martin Y. et al. C. , J. et al. Med. Chem. , Vol. 35, p2145-2154 (1992)
[Non-patent Document 6] Provided by Chemical Computing Group, Distributed Computing using MOE, [online], [searched on June 24, 2003], Internet <URL: http: // www. chemcomp. com />
[Non-patent Document 7] Hahn, M .; , J. et al. Med. Chem. , Vol. 38, p2080-90 (1995)
[Non-patent Document 8] Hahn, M .; Rogers, D.C. , J. et al. Med. Chem. , Vol. 38, p2091-2102 (1995)
[0013]
[Problems to be solved by the invention]
The present invention is a method for automatically searching for a stable complex structure simply by giving a three-dimensional structure of a ligand and a receptor. In addition to screening a ligand for a known receptor, the ligand binds stably. It is an object of the present invention to provide a method for searching a receptor-ligand stable complex structure using a computer, which can also perform reverse screening for searching for a possible receptor.
[0014]
[Means for Solving the Problems]
In order to solve the above-mentioned problems, the present inventors have intensively studied, and as a result, replaced the position estimation of the binding pocket in the receptor and the binding mode of the ligand to the binding pocket with a simplified pharmacophore model, This would be the case when the receptor and ligand form a stable complex by continuously comparing the various conformations of the ligand generated by computational chemistry methods to fit this model. We have succeeded in developing a method that makes it possible to estimate the structure of a ligand, including its binding position, in a short time. In addition, we found that the structure of the complex can be determined by applying molecular mechanics calculation to the structure estimated here and refining the structure and comparing the energy rank of each structure. The present invention has been completed.
[0015]
That is, the present invention is as follows.
(1) A method for searching for a stable complex structure of a receptor and a ligand, which comprises the following steps:
A first step of comprehensively generating possible conformations of the ligand;
In the three-dimensional structural model of the receptor, a binding pocket is exhaustively searched, and a space in which a sphere having a radius of 0.5 to 4.0 mm can exist in the obtained binding pocket is obtained. A second step of placing dummy atoms based on the coordinates of the space
A third step of setting a pharmacophore in the three-dimensional structural model of the receptor by clustering the dummy atoms with reference to the coordinates of the dummy atoms;
A fourth step of setting, for the constituent atoms of the receptor present around the binding pocket searched in the second step, a spherical excluded volume region having a radius equal to or larger than the van der Waals radius of the atom;
A combination that matches each conformation of the ligand generated in the first step with the pharmacophore set in the third step is selected, and the combination of the ligand is further selected from the combination. A fifth step of selecting a combination in which the constituent atoms do not exist in the region set in the fourth step; and
A sixth step of obtaining an energetically stable combination by molecular mechanics calculation for the combination found in the fifth step.
(2) In the third step, the value of the electrostatic interaction that the dummy atom receives from the receptor is further calculated, the dummy atom is classified based on the value, and the dummy atom is clustered for each classification. In the fifth step, the ligand conformation and the pharmacophore are superposed by further matching the characteristics of the constituent atoms of the ligand with the characteristics of the classified dummy atoms. Method (1).
(3) For each of a specific receptor and a plurality of ligands capable of forming a complex with the receptor, a complex in which the receptor and the ligand are stable is obtained by the method (1) or (2). A method of searching for a ligand that forms a stable complex with a specific receptor by determining whether or not to form.
(4) For each of a plurality of receptors capable of forming a complex with a specific ligand, the ligand and the receptor form a stable complex by the method (1) or (2). A method of searching for a receptor that forms a stable complex with a specific ligand by determining whether or not to do so.
(5) The method according to any one of (1) to (4), wherein the receptor is a protein.
(6) A program for causing an information processing apparatus to execute processing for searching for a stable complex structure of a receptor and a ligand, and for executing the following procedure:
A first procedure for comprehensively generating possible conformations of the ligand;
In the three-dimensional structural model of the receptor, a binding pocket is exhaustively searched, and a space in which a sphere having a radius of 0.5 to 4.0 mm can exist in the obtained binding pocket is obtained. A second procedure for arranging dummy atoms based on the coordinates of the defined space;
A third procedure for setting a pharmacophore in a three-dimensional structural model of a receptor by clustering the dummy atoms with reference to the coordinates of the dummy atoms;
A fourth procedure for setting a spherical excluded volume region having a radius equal to or larger than the van der Waals radius of the atoms of the acceptor constituent atoms existing around the binding pocket searched by the second procedure;
A combination that overlaps and matches each conformation of the ligand generated by the first procedure with the pharmacophore set in the third procedure is selected. A fifth procedure for selecting a combination in which the constituent atoms do not exist in the region set in the fourth procedure; and
A sixth procedure for obtaining an energetically stable combination by molecular mechanics calculation for the combination found by the fifth procedure.
(7) The third procedure further calculates the value of the electrostatic interaction that the dummy atom receives from the receptor, classifies the dummy atom based on the value, and performs clustering of the dummy atom for each classification. And the fifth procedure further includes matching the ligand conformation and the pharmacophore by associating the characteristics of the constituent atoms of the ligand with the characteristics of the classified dummy atoms. (6) The program according to (6), further comprising a procedure for performing the matching.
(8) The program according to (6) or (7), wherein the receptor is a protein.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the present invention will be described in detail.
[0017]
In the present invention, the receptor means a high molecular compound such as protein, nucleic acid, sugar, etc., preferably a protein. The receptors may be natural ones, modified ones thereof, or artificial ones synthesized by imitating them. It is preferable that the receptor has a known three-dimensional structure or can easily estimate the three-dimensional structure from the structure of a receptor having a known three-dimensional structure. When a protein is used as a receptor, the three-dimensional structure of the protein can be obtained from, for example, Protein Data Bank Japan (http://pdbj.protein.osaka-u.ac.jp/).
[0018]
In addition, a ligand is generally a small molecule compared to a receptor, and is a compound that specifically binds to a specific site in a receptor to express physiological activity or toxicity, regardless of whether it is a natural product or a chemically synthesized substance. It is a concept that includes the whole. Ligands include, for example, drugs or enzyme substrates or inhibitors. In general, a ligand sometimes refers to a substance that binds to an active site of a receptor and exerts a pharmacological action, but in the present invention, the ligand may be any substance that can form a stable complex with a receptor. It does not necessarily have to bind to the active site and exert pharmacological action.
[0019]
The method of the present invention will be described step by step with reference to the flowchart of FIG. In the first step, first, the conformation that the ligand can take is comprehensively generated. The ligand may be used by inputting a chemical formula, or may be selected from a compound database including a plurality of ligands.
[0020]
Various known methods can be used as a method for efficiently generating various conformations that can be taken by the ligand. For example, for a rotatable bond in a ligand, a general systematic search that sequentially rotates at a constant angle to generate all structures comprehensively, dihedral angles and atomic coordinates are changed randomly, and then molecular mechanics Stochastic search for structural optimization by calculation (Ferguson, DM, Raber, DJ, J. Am. Chem. Soc. Vol. 111, p 4371-4378, 1989) Simulated annealing (Nilges, M., Clore, GM, Gronenborn) sampling various conformations that would be obtained by long-term observation by repeating structural optimization by molecular dynamics and molecular mechanics calculations , AM, FEBS Lett., Vol. 239, p12 -136, 1988), and the like. In addition, a commercial program in which these methods are extended, for example, the Information Import function in MOE (trade name) manufactured by Chemical Computing Group of Canada, or the polling method (Smelli, A) installed in CATALYST (trade name) manufactured by Accelrys of the United States. , Teig, S.L., and Towbin, P., J. Comp.Chem., Vol.16, p171-187, 1995), etc., to search various conformational spaces in a short time be able to.
[0021]
In the second step, the binding pocket is first exhaustively searched in the three-dimensional structural model of the receptor. Here, the binding pocket refers to an inner hole or a depression that is observed on the surface of a biopolymer molecule and has a size that allows a ligand to bind. For example, the size of the binding pocket is preferably such that four or more spheres having a radius of 0.5 to 4.0 mm can exist, and more preferably seven or more.
[0022]
Further, a space in which a sphere having a radius of 0.5 to 4.0 can exist in the binding pocket is searched, and dummy atoms are arranged based on the coordinates of the space. The space to be searched is preferably a space where a sphere with a radius of 1.0 to 3.0 mm can exist, and more preferably a space where a sphere with a radius of 1.2 to 2.0 mm can exist. In addition, the dummy atom has a size, mass, etc. in order to use it as a standard for processing such as movement and superposition of molecules to relative positions in space when handling molecular structures on a computer. It is a virtual atom that is arbitrarily set and generally has no potential energy. The dummy atoms are set based on the coordinates of the space, but can be arbitrarily set in the space, and a plurality of dummy atoms may be set.
[0023]
As a method for searching for a space in which a binding pocket and a sphere having a radius of 0.5 to 4.0 mm can exist in the receptor, for example, a coordinate space in which the receptor exists is divided into a three-dimensional grid (grid). A geometric method for detecting lattice points where no body atoms are present can be used. Specifically, various software such as the well-known and simple Accelrys C2-LigandFit (trade name) can be obtained and used.
[0024]
However, this method may have different results depending on the relative coordinates of the biopolymer and the grid, and a method that is less susceptible to the influence of the coordinates is more preferable. As a method which is not easily influenced by coordinates, for example, a method by Edelsbrunner et al. (Edelsbrunner, H., Facello, M., Fu, R., Liang, J., “Measuring Proteins and VOIDs in Proteins. International Conference on Systems Science., P256-264, 1995)) and Alpha Site Finder installed in MOE (trade name) manufactured by Chemical Computing Group based on this algorithm.
And the like.
[0025]
In the third step, clustering is performed on the dummy atoms arranged in the second step, and a pharmacophore is set based on the obtained cluster. Clustering can be performed based on the coordinates of the dummy atoms. For example, clustering can be performed by bringing together dummy atoms within a distance of 2.0 cm or less, preferably 1.5 cm or less.
[0026]
In this step, the electrostatic interaction energy that each dummy atom will receive from the acceptor is calculated by molecular mechanics or quantum mechanics, preferably molecular mechanics, and the value of the obtained electrostatic interaction is It is good also as a standard of clustering with a position. For example, the electrostatic interaction energy can be calculated as follows. In other words, assuming a dummy atom with a virtual charge of +1, the interaction energy with an atom having a charge of q existing at a distance r can be regarded as q / εr, where the dielectric constant is ε (r). it can. Here, ε is distance dependent in consideration of the electrostatic shielding effect in the protein solution. This calculation is performed for all atoms around the bond pocket, and the sum is the electrostatic interaction energy received by the dummy atom. From the calculated value of the electrostatic interaction, when a ligand atom exists at the coordinates of the dummy atom, it is estimated whether or not it has a donor property or an acceptor property with respect to a hydrogen bond. Dummy atoms can be classified into characteristics such as ionic), acceptor (anionic), and aromatic (hydrophobic), and none of them can be clustered for each category. . Similarly to the above, clustering for each classification can be performed by bringing together dummy atoms within a distance of, for example, 2.0 mm or less, preferably 1.5 mm or less.
[0027]
The pharmacophore can be set by combining the clusters obtained by the clustering in each binding pocket. In addition, a representative feature (a cluster having characteristics that are considered to interact strongly with the receptor among the above characteristics) is selected from the clusters, and a pharmacophore is set by combining these features. It may be simplified. When simplifying the model, if the number of features is set to be 4 or more, preferably 7 or more, a specific combination of a pharmacophore and a ligand in the fifth step Search is possible.
[0028]
In order to form a complex in which the receptor and the ligand are stable, the ligand has an atom corresponding to at least a part of the characteristic at the position (pharmacophore) defined above, and the other atom is the receptor. It is essential to be able to take a conformation that does not overlap. In order to determine whether or not this condition is satisfied, in the fourth step, a spherical exclusion set to a radius greater than the van der Waals radius of the atom, centered on the atomic coordinates of the acceptor molecule present around the binding pocket Set the volume area. The problem here is the accuracy of the three-dimensional structure of a given receptor. The error in coordinates obtained by techniques such as X-ray and NMR is 1 to 3Å for normal proteins. Therefore, in setting this excluded volume region, it is necessary to set conditions in consideration of this error. In general, the three-dimensional structure that can be determined by techniques such as X-rays is only the coordinates of heavy atoms other than hydrogen, and is bonded to carbon, oxygen, nitrogen, sulfur atoms, etc. that constitute most amino acids and nucleic acids. The position of the hydrogen atom is determined by a method such as molecular mechanics calculation. The typical bond length between carbon and hydrogen is about 1.1 １．１, but the position of the hydrogen atom can easily move within a certain range due to the presence of the ligand.
[0029]
If the exclusion volume is set based on the coordinates of all atoms around the binding pocket, there is a risk that the structure to be originally bonded may be excluded due to the accuracy of the acceptor structure that is the above assumption. is there. Therefore, in the present invention, for example, only the atoms that are more than 2Å, preferably more than 3Å, of the dummy atoms set in the second step are considered, and a radius equal to or greater than the van der Waals radius is set for these atoms. A spherical excluded volume may be set. The radius equal to or greater than the van der Waals radius is preferably a radius of 100 to 200%, more preferably a radius of 100 to 150% of the van der Waals radius. Moreover, it is preferable to set the excluded volume in consideration of only atoms other than hydrogen.
[0030]
Further, in the fourth step, the excluded volume region is set based on atoms other than hydrogen atoms present in the range of 2 cm or more, preferably 3 mm away from the dummy atoms, and within 15 mm, preferably within 12 cm. It is possible to shorten the calculation time and improve the search accuracy.
[0031]
In the fifth step, first, a combination that matches the conformation of the ligand generated in the first step with the pharmacophore model obtained in the third step is selected. Here, the “matching combination” means, for example, a combination in which the root mean square distance (RMSD) between the dummy atom and the corresponding constituent atom of the ligand is 2.0 or less, preferably 1.2 or less.
[0032]
Subsequently, for the selected combination, a combination that does not include the ligand constituent atom in the exclusion region set in the fourth step is searched. That is, in the fifth step, any atom in the ligand exists at the position of the dummy atom in the pharmacophore defined in the third step, and any of the atoms in the exclusion region defined in the fourth step. This is a process of searching for the arrangement of the ligand and the receptor so that no atoms exist.
[0033]
In addition, when each dummy atom is classified by electrostatic interaction and clustered for each feature in the third step, characteristics of the constituent atoms of the ligand (cationicity, anionicity, etc.) in the fifth step. ) And the characteristics of the dummy atom feature, the ligand conformation and pharmacophore can be superimposed. For example, when the value of the electrostatic interaction of the dummy atom is positive, it is more stable that an atom having a negative charge exists at that position. Therefore, in the ligand, an atom having a negative partial charge is present. There is a high possibility of being in this position. Such atoms are considered to be hydrogen bond acceptors or anions. On the contrary, when the value of the electrostatic interaction of the dummy atom is negative, it is considered that a donor or a cation is likely to exist at that position.
[0034]
In the fifth step, the pharmacophore that superimposes with the ligand conformation is set so that the number of features in the third step is 4 or more, preferably 7 or more. In other words, a more stable combination of the pharmacophore and the ligand can be searched. If a stable combination of pharmacophore and ligand cannot be obtained when using a pharmacophore with the above number of features, automatically search for ligands by reducing the number of matching features automatically. Thus, it is possible to preferentially search for a combination of a highly specific ligand and pharmacophore.
[0035]
The three-dimensional structure of the ligand generated in the first step is strictly a conformation generated outside the receptor as a single ligand, and therefore exactly matches the structure in the complex with the receptor. There is no guarantee. Therefore, in the present invention, the first to fifth steps first estimate a highly probable complex structure that has good feature matching and does not overlap with the receptor, and then the sixth step molecular mechanics calculation or quantum calculation. By performing structural refinement based on dynamic calculation, the three-dimensional structure that the ligand can take in the receptor is estimated.
[0036]
In the structure refinement in the sixth step, it is preferable to use molecular mechanics calculation from the viewpoint of calculation speed. Since the accuracy in molecular mechanics calculation depends on the force field to be used, it is necessary to use a force field having parameters corresponding to both the biopolymer and the low-molecular ligand in the docking calculation. Commonly available force fields include CVFF (P. Douber-Osghorpe et al, Proteins: Structure. Func. Genet., Vol. 4, p31, 1988), CFF (M. J. Hwang et al., J. Am. Chem. Soc., Vol 116, p2515, 1994), OPLS-AA (W. L. Jorgensen et al, J. Am. Chem. Soc., Vol. 117, p11225-12136, 1996), MMFF94, MMFF T.A. Halgren, J. Comp.Chem.vol.17, No. 5 & 6, p490, 1996) and the like. Further, as a molecular mechanics calculation engine including these parameters, Discover (registered trademark, Accelrys Inc. (USA)), CHARMM (WD Cornell et al, J. Am. Chem. Soc., Vol. 117,). p5179, 1995), MOE-Minimizer (registered trademark, Chemical Computing Group Inc. (registered trademark)), Direct Force Field (registered trademark, Aeon Technology, Inc. (registered trademark)), and the like. is not.
[0037]
Furthermore, in the present invention, the first to sixth steps are continuously performed for each of a specific receptor and a plurality of ligands capable of forming a complex with the receptor. Thus, it is possible to search for a ligand that forms a stable complex with respect to a specific receptor. Thereby, it is possible to discover candidate compounds for novel physiologically active substances.
[0038]
In the present invention, the first step to the sixth step are continuously performed on a specific ligand and each of a plurality of receptors capable of forming a complex with the ligand. It is possible to search for a receptor that forms a stable complex with a specific ligand. As a result, it is possible to clarify the action mechanism of the drug or to estimate the physiological activity of the ligand, such as side effects and toxicity, from the known function of the receptor to which the ligand can bind.
[0039]
The present invention also provides a program for causing an information processing apparatus to execute the first to sixth steps. Such a program is, for example, a program in which a procedure for executing each of the above steps is described in an appropriate computer language. The computer language is not particularly limited as long as it is a language usually used for scientific calculation, such as C, C ++, FORTRAN, JAVA (registered trademark), SVL. In addition, the program of the present invention may be a consistent program that describes all processes, but a function that implements each function or a function that realizes the function by combining published programs that have a function to implement a part of the process. It is also possible to collect scripts that call. With respect to a file containing sufficient data for reproducing the three-dimensional structure of a molecule on a computer, such as coordinates of receptors and ligands, element names, etc., the above first to sixth information processing apparatuses such as a computer are used. By carrying out the step, it is possible to search for a stable complex structure of the receptor and the ligand.
[0040]
【Example】
In order to show the effectiveness of the method of the present invention, a program capable of continuously executing the steps of the present invention is created, and a receptor and a ligand are separated from a complex having a known three-dimensional structure. By applying this program, it was confirmed that a known ligand-receptor complex structure can be reproduced in a short time.
[0041]
Here, a program was created using the language SVL and functions installed in the computational chemistry system MOE (trade name) (2002.03 version) manufactured by Chemical Computing Group (Canada). In the molecular mechanics calculation, Merck force fields (MMFF94s) having parameters for a wide range of ligands as well as biopolymers and having high structure reproduction accuracy of the complex were used. For parameters that are still insufficient, a rule-dependent force field that complements this Merck force field can be used to automatically calculate any intermolecular interaction.
[0042]
Among the complexes registered in the protein data bank (http://pdbj.protein.osaka-u.ac.jp/), streptavidin (entry 2rtd), acetylcholinesterase (1acl), retinol-binding protein (1aqb) , Glutamate receptor subunit 2 (1 ftm), HIV-1 reverse transcriptase (1 hnv), and nuclear receptor (3ert) are stable with known ligands known to bind to these receptors Whether or not a complex was formed was confirmed using the above program.
[0043]
Hereinafter, the procedure will be described in detail by taking streptavidin as an example. The three-dimensional structure registered in the protein data bank is determined by X-ray and does not include the coordinates of hydrogen atoms. Before applying the method of the present invention, this structure was read into MOE, and after adding a hydrogen atom, the position of the hydrogen atom was determined by molecular mechanics calculation. At this time, in order to maintain the structure obtained by the experiment, the coordinates of atoms other than hydrogen were fixed, and only the position of the hydrogen atom was optimized by molecular mechanics calculation.
[0044]
In the first step, the conformation_import function mounted on the MOE (trade name) was used to comprehensively generate the conformations that biotin can take (FIG. 2). In the second step, a sphere with a radius of 1.4 to 1.8 mm is entered using the MOE-Alpha Site Finder function, which is an extension of the method of Edelsbrunner et al. (Edelsbrunner, H., et al., 1995). The possible coordinates were determined, and dummy atoms having no energy were arranged at all positions (FIG. 3).
[0045]
In the third step, all the distances between the dummy atoms were measured, and all the distances within 2.5 距離 were collected. In this system, six groups were obtained.
[0046]
In the third step, the value of the electrostatic potential that would be received from the acceptor when each dummy atom was assumed to have a charge of +1 was calculated as the Coulomb force. Within each cluster, dummy atoms are classified into three types: those whose potential value exceeds +15 kcal / mol, those that are less than −15 kcal / mol, and those in the middle. We clustered dummy atoms within the distance and set pharmacophore based on it.
[0047]
In these regions, for example, a region having a potential smaller than −15 kcal / mol tends to stabilize the energy of the complex because a positively charged atom exists in this region in the ligand. That is, it is considered that this region has a high possibility that a hydrogen bonding acceptor or an anion exists. Thus, hydrogen bonding acceptor (anion) region (Acc | Ani) of −15 kcal / mol or less, hydrogen bonding donor (cation) region (Don | Cat) of +15 kcal / mol or more, Aromatic (hydrophobic) regions (Aro | Hyd), which are not any of them, were classified into three types. A ligand having more matching properties at these relative positions is likely to form a complex stably with the receptor. In the present invention, a pharmacophore model of a ligand that may exist stably in the structure of the receptor was estimated from the structure of the receptor.
[0048]
In the fourth step, as the excluded volume region, the coordinates of atoms other than the hydrogen atom of the acceptor that are at a distance of 4.5 以 or more and 8 か ら or less from all the dummy atoms are detected, and the radius is calculated from each coordinate. A spherical region of 0 設定 was set (FIG. 4).
[0049]
In the fifth step, the conformation generated in the first step was superposed on the pharmacophore model obtained in the third step, and an arrangement that could become a stable complex was found. Here, when Acc or Don is present in the pharmacophore model, they are superposed in consideration of the characteristics (hydrogen bonding) of the ligand atoms, and further, in the third step, the atoms are the ligand atoms. Only those that did not exist were selected. One of the resulting combinations of ligand and pharmacophore is shown in FIG. It can be seen that 7 out of the 10 features that make up the pharmacophore match the ligand atoms.
[0050]
In the sixth step, the structures of the ligands that are candidates for stable complexes obtained as described above were sequentially placed in the receptor, and the structure was optimized by molecular mechanics calculation. The smaller the total of the internal energy of the ligand and the interaction energy of the ligand and the receptor, the more stable the complex, so the structures were rearranged in order from the lowest total energy.
[0051]
As a result, it was confirmed that a very stable complex was formed by placing the ligand in one of the six binding pockets (FIG. 6). Furthermore, it was confirmed that the experimental data was correctly reproduced by superimposing this structure on the structure determined by X-ray crystal structure analysis (FIG. 7). Table 1 shows the results of calculating the root mean square distance (RMSD) between corresponding atoms in the structure of the ligand in the complex determined by the method of the present invention and the structure determined by X-ray crystal structure analysis. Indicated. The root mean square distance was 1.218, and it was found that the structures of both were very close.
[0052]
Combinations of other ligands and receptors (Decamethonium and acetylcholinesterase, Retinol and retinol binding protein, AMPA ((S) -alpha-amino-3-hydroxy-5-methyl-4-isoxolepropionic acid) and glutamate receptor subunit 2 , TIBO R86183 and HIV-1 reverse transcriptase, OHT and nuclear receptor), a stable complex structure was determined by the method of the present invention, and X-ray crystal structure analysis data and root mean square distance were calculated. The results are also shown in Table 1. Also for these combinations, the RMSD was in the range of 0.744 to 1.279, and it was proved that the accuracy of the method of the present invention was high. For reference, the interaction energy of these complexes is also shown.
[0053]
[Table 1]

[0054]
【The invention's effect】
By using the novel search method disclosed in the present invention, it has become possible to specify a stable complex structure with a ligand in a short time even for a receptor whose binding site is not specified. Therefore, not only the development of new ligands for known receptors but also the discovery of new pharmacologically active substances, side effects and toxicity by investigating whether known ligands form stable complexes with other receptors. Therefore, the cost and time of drug discovery research can be greatly improved.
[Brief description of the drawings]
FIG. 1 shows a flowchart of processing.
FIG. 2 is a diagram showing examples of various ligand conformations generated in the first step.
FIG. 3 is a diagram showing a molecular model of a dummy atom and a receptor obtained in the second step. The solid line indicates the biotin-binding protein, the sphere indicates the biotin, the small sphere indicates the center of the dummy atom indicating the binding candidate pocket, and the size of the binding pocket is indicated by the Connolly analysis surface (black space).
FIG. 4 is a diagram showing a pharmacophore region in which ligand atoms obtained in the third and fourth steps may exist and a sphere showing an excluded volume.
FIG. 5 is a diagram showing an example of a complex structure candidate (bar ball model) obtained in the fifth step.
FIG. 6 shows the structure of the most stable complex obtained in the sixth step.
FIG. 7 is a diagram showing the superposition of the ligand biotin (represented by a bar) in the most stable complex obtained in the sixth step and the ligand (represented by a bar) determined by X crystal structure analysis.
FIG. 8 is a diagram showing a superposition of the ligand Decamethonium (bar display) in the most stable complex obtained for acetylcholinesterase (PDB entry 1ACL) and the ligand (bar display) determined by X crystal structure analysis.
FIG. 9 is a diagram showing a superposition of the ligand Retinol (bar display) in the most stable complex obtained for retinol-binding protein (PDB entry 1AQB) and the ligand (bar display) determined by X crystal structure analysis.
FIG. 10: Overlay of the ligand (AMPA, indicated by barball) in the most stable complex obtained for glutamate receptor subunit 2 (PDB entry 1FTM) and the ligand determined by X crystal structure analysis (displayed by bar) FIG.
FIG. 11: Overlay of ligand TIBO R86183 (bar display) in the most stable complex obtained for HIV-1 reverse transcriptase (PDB entry 1HNV) and ligand determined by X crystal structure analysis (bar display) FIG.
FIG. 12: Overlay of ligand 4-hydroxytamoxifen (OHT, barball display) and ligand (bar display) determined by X crystal structure analysis in the most stable complex obtained for nuclear receptor (PDB entry 3ERT) FIG.

Claims (8)

A method for searching for a stable complex structure of a receptor and a ligand, comprising the following steps:
A first step of comprehensively generating possible conformations of the ligand;
In the three-dimensional structural model of the receptor, a binding pocket is exhaustively searched, and a space in which a sphere having a radius of 0.5 to 4.0 mm can exist in the obtained binding pocket is obtained. A second step of placing dummy atoms based on the coordinates of the space
A third step of setting a pharmacophore in the three-dimensional structural model of the receptor by clustering the dummy atoms with reference to the coordinates of the dummy atoms;
A fourth step of setting, for the constituent atoms of the receptor present around the binding pocket searched in the second step, a spherical excluded volume region having a radius equal to or larger than the van der Waals radius of the atom;
A combination that matches each conformation of the ligand generated in the first step with the pharmacophore set in the third step is selected, and the combination of the ligand is further selected from the combination. A fifth step of selecting a combination in which the constituent atoms do not exist in the region set in the fourth step; and
A sixth step of obtaining an energetically stable combination by molecular mechanics calculation for the combination found in the fifth step. In the third step, the value of the electrostatic interaction that the dummy atom receives from the receptor is further calculated, the dummy atom is classified based on the value, and the dummy atom is clustered for each classification; and The superposition of the ligand conformation and the pharmacophore is performed in the fifth step by further matching the characteristics of the constituent atoms of the ligand with the characteristics of the classified dummy atoms. The method described in 1. Whether the receptor and the ligand form a stable complex by the method according to claim 1 or 2 for a specific receptor and each of a plurality of ligands capable of forming a complex with the receptor. A method of searching for a ligand that forms a stable complex for a specific receptor by determining whether or not. Whether the ligand and the receptor form a stable complex by the method according to claim 1 or 2 for a specific ligand and each of a plurality of receptors capable of forming a complex with the ligand. A method of searching for a receptor that forms a stable complex with a specific ligand. The method according to any one of claims 1 to 4, wherein the receptor is a protein. A program for causing an information processing apparatus to execute processing for searching for a stable complex structure of a receptor and a ligand, and for executing the following procedure:
A first procedure for comprehensively generating possible conformations of the ligand;
In the three-dimensional structural model of the receptor, a binding pocket is exhaustively searched, and a space in which a sphere having a radius of 0.5 to 4.0 mm can exist in the obtained binding pocket is obtained. A second procedure for arranging dummy atoms based on the coordinates of the defined space;
A third procedure for setting a pharmacophore in a three-dimensional structural model of a receptor by clustering the dummy atoms with reference to the coordinates of the dummy atoms;
A fourth procedure for setting a spherical excluded volume region having a radius equal to or larger than the van der Waals radius of the atoms of the acceptor constituent atoms existing around the binding pocket searched by the second procedure;
A combination that overlaps and matches each conformation of the ligand generated by the first procedure with the pharmacophore set in the third procedure is selected. A fifth procedure for selecting a combination in which the constituent atoms do not exist in the region set in the fourth procedure; and
A sixth procedure for obtaining an energetically stable combination by molecular mechanics calculation for the combination found by the fifth procedure. The third procedure further includes a procedure of calculating a value of the electrostatic interaction received by the dummy atom from the receptor, classifying the dummy atom based on the value, and performing clustering of the dummy atom for each classification. And the fifth procedure further superimposes the ligand conformation and the pharmacophore by associating the characteristics of the constituent atoms of the ligand with the characteristics of the classified dummy atoms. The program according to claim 6, further comprising a procedure. The program according to claim 6 or 7, wherein the receptor is a protein.

Images