JP5142179B2 - Prediction method of interaction sites in biopolymers - Google Patents

Description

  The present invention relates to a method and a program for predicting an interaction point at which a biopolymer interacts or forms a complex with an external factor such as a small molecule or a polymer.

The post-genome era is now called, and as a result of a huge number of gene analyses, many biopolymers have been identified or predicted to exist. On the other hand, the field of life science such as functional materials focusing on the functions of biopolymers, fuels and new materials is also attracting attention.
Most of these biopolymers and non-natural polymers simulating them do not perform their functions alone, but only after interacting with or forming complexes with metal atoms, low molecules, or similar or different polymers. Demonstrate the function. In addition, interaction with an external factor may cause a change in the higher-order structure of the biopolymer, and as a result, the function may be expressed by interaction with its own remote structure. Originally, the higher-order structure of a biopolymer itself is maintained by interaction with external factors (for example, water) or complex formation, and further by formation of a macrobiomolecule system called a biological device.
However, in most of these biopolymers, external factors involved in functional expression through interaction or complex formation are unknown, and even when external factors are known, interaction sites with biopolymers or complex formation In most cases, the site has not been identified.
In order to effectively, selectively, and safely exert the function of the biopolymer, it is necessary to identify the interaction site or complex formation site in the biopolymer and analyze it at the atomic level. Such a procedure is not only a search for natural external factors but also a rapid and most logical search for new factors and structural modifications to known factors.

Structure-based drug discovery requires the three-dimensional structure of the target biopolymer, strictly speaking, the atomic coordinates of the stable three-dimensional structure, which currently has X-ray crystal structure analysis, nuclear magnetic resonance (NMR), or scientific evidence. Obtained by molecular modeling such as homology modeling.
However, the three-dimensional structure of biopolymers does not have a clear “key and keyhole” shape expressed in textbooks, etc., and what position is the interaction site or complex formation site from the appearance I cannot judge at all.
At present, by using the X-ray crystal structure analysis technique, a complex formation site of a ligand (that is, a ligand binding site) and a complex formation site such as a polymer can be specified. In most cases, it is impossible to identify a ligand-binding site or a complex-forming site such as a polymer.
Other technologies include chemical methods such as point mutation and radiolabeling, and spectroscopic methods such as NMR and fluorescence, all of which require enormous experimentation and effort, but are successful. However, it is possible to obtain only a result that can estimate a very broad area.
On the other hand, few algorithms are known for predicting a ligand-binding site or a complex-forming site with a polymer.
For example, with regard to algorithms for predicting ligand binding sites, these are all a set of cubes (ie, grids) at regular intervals, using probe atoms and probe molecules, and all or part of the biopolymer that has undergone the necessary treatment. A method of dividing and arranging the probe on each lattice point to estimate the interaction energy (that is, grid-based) (for example, see Non-Patent Document 1 and Non-Patent Document 2) or a non-grid-based method (non-patent) Reference 3).

These methods using probe atoms and probe molecules have many problems described below.
That is, in these methods, (1) in order to estimate the interaction energy, component data such as molecules other than biopolymers and metal atoms must be removed from the data of the provided biopolymer system. In addition to ignoring the energy contribution from the removed components, extra work is involved and special working knowledge is required. (2) Probe atoms and probe molecules are sized (ie (It has an atomic radius or a molecular radius), and a narrow space cannot be searched, and the complex formation site may be missed. (3) After reading the coordinates of the biopolymer, or on the whole lattice point The interaction energy must be calculated by scanning the surface of the molecule with probe atoms or molecules, resulting in extra work ( ) When probe molecules are used, the problem of orientation of probe molecules with respect to the internal coordinate system of biopolymers arises. (5) At present, only the force field calculation level can be used to estimate the interaction energy. At the level, the interaction energy greatly contributed by the electron correlation interaction is greatly underestimated, and the complex formation site may be missed. For (5), this problem is significant, especially since the ligand binding site is often in a hydrophobic environment. This is the result of Non-Patent Document 1 and Non-Patent Document 2 that the prediction accuracy is improved by using a hydrophobic probe molecule, and conversely, when the probe molecule is used, the contribution of electrostatic interaction is The large interaction energy is greatly underestimated, leading to missed complex formation sites and is not an essential solution. Furthermore, in the grid-based algorithm, when there is an energy minimum point that rapidly stabilizes around the lattice point, a complex formation site may be missed, and as described in Non-Patent Document 3, even with the same grid distance, The grid base position is shifted depending on the starting point for creating the grid, resulting in a difference in energy value. As a result, there is a further disadvantage inherent to the grid base in that the reproducibility of the predicted position is lost.

In addition, regarding a complex formation site with a polymer, the currently known algorithm is essentially only a method of searching and estimating a hydrophobic surface of a biopolymer.
As for a method for overcoming the above problems, an effective method is not known, although it is currently being studied as a national project by investing a huge amount of money from various directions all over the world.
Structure-based drug discovery, such as docking simulation and virtual screening, is a technique that is used after the ligand binding site has been identified. Needless to say, these techniques cannot be applied to molecules. On the other hand, if it is possible to accurately predict the ligand binding site or the site where the complex is formed with the polymer, various subsequent applications are expected.

The 26th Symposium on Information Chemistry Abstract J14, 2003 Bioinformatics, 21, 1908-1916, 2005 32nd Structure-Activity Relationship Symposium Abstract K07, 2004

  The present invention searches for an internal remote structure in a biopolymer or an interaction site with an external factor such as a metal atom, a small molecule or a polymer, and has a chemical group, species or molecular species having an affinity for the interaction site. The method of predicting is provided.

The present inventor has a completely different viewpoint from the existing method of handling the entire structure of the biopolymer, that is, the entire structure of the biopolymer is made into a basic fragment that constitutes the biopolymer and is minimized to the extent that the chemical properties are not impaired. And restructuring the entire structure from the basic fragment, thereby assigning the energy properties of highly complex biopolymers to the simplified energy properties of the basic fragments. Settled.
That is, the present inventor creates a database of interaction point information of biopolymer constituent atoms and calculates the position (coordinate) of the interaction point from the atomic coordinate information of any biopolymer constituent atom with reference to the database. By doing so, the inventors succeeded in developing a method for predicting the interaction site of the biopolymer.

That is, the present invention
A method for predicting an interaction site in a biopolymer,
The first step of preparing atomic coordinate information of biopolymers,
A second step of identifying coordinates of the interaction point candidate of the biopolymer based on the interaction point information database, wherein the interaction point information database identifies the coordinates of the interaction point of the biopolymer constituent unit Identifying the position of a candidate interaction point by applying the information to atomic coordinate information of a given biopolymer constituent unit,
Of the interaction point candidates identified in the second step, the third step of removing the interaction point candidates involved in the interaction with the constituent atoms of the biopolymer,
The above method is provided.

The present invention also provides:
After the third step, a fourth step of further removing interaction point candidates involved in the interaction between the surface of the biopolymer and the solvent molecule,
A method for predicting an interaction site in a biopolymer is provided.

Furthermore, the present invention provides
A computer program for predicting an interaction site in a biopolymer,
First step to read atomic coordinate information of biopolymer,
A procedure for identifying coordinates of interaction point candidates of the biopolymer with reference to an interaction point information database, the interaction point information database identifying coordinates of interaction points of biopolymer constituent units A second procedure for identifying a position of a candidate interaction point by applying the information to atomic coordinate information of a given biopolymer constituent unit,
For the interaction point candidates identified in the second step,
Determining whether or not a constituent atom of the biopolymer exists within a predetermined distance from the interaction point candidate, and when present within the predetermined distance, a third procedure for removing the interaction point candidate;
A computer program for causing a computer to execute a procedure for displaying a predicted position of an interaction point on a display is provided.

The present invention also provides:
A computer program for predicting an interaction site in a biopolymer,
First step to read atomic coordinate information of biopolymer,
A procedure for identifying coordinates of interaction point candidates of the biopolymer with reference to an interaction point information database, the interaction point information database identifying coordinates of interaction points of biopolymer constituent units A second procedure for identifying a position of a candidate interaction point by applying the information to atomic coordinate information of a given biopolymer constituent unit,
For the interaction point candidates identified in the second step,
A third procedure for determining whether or not a constituent atom of a biopolymer exists within a predetermined distance from the interaction point candidate, and removing the interaction point candidate when present within a predetermined distance;
A fourth procedure of removing each interaction point candidate when the interaction point candidate exists outside the solvent contact surface or van der Waals surface of the biopolymer.
A computer program for causing a computer to execute a procedure for displaying a predicted position of an interaction point on a display is provided.

  According to the present invention, a complex formation site of an external factor can be specified automatically and accurately only from the entire structure or partial structure of an arbitrary biopolymer system. In addition, by providing the physicochemical information of the site, it is possible to identify the shape and detailed chemical requirement of external factors that can form a complex at the complex-forming site.

Hereinafter, the present invention will be described in detail. Please refer to FIG. 1 showing the process of the present invention if necessary.
In the present specification, the “biopolymer” includes a high molecular compound generally present in the living body such as protein, nucleic acid, sugar and the like. The biopolymers need not be natural, and may be modified from them or artificially synthesized by imitating them. Particularly preferred biopolymers in the present invention are proteins. Biomolecules may be water molecules, small molecules such as ligands, other molecules such as metal atoms, components of biological membranes, or the like in the living body or bound as artifacts. When included, the biopolymer and the other molecules are collectively represented by the term biopolymer system.
In the present specification, the “interaction point” means a minimum point of interaction energy generated with a biopolymer constituent atom or a local region around it, for example, van der Waals interaction, electrostatic Energy minimum points such as interaction are listed. In addition, the entire region that includes the interaction point and is recognized to interact with the external factor in the biopolymer is represented as an interaction site. External factors refer to atoms or molecules that have structural, functional, or physical effects on biopolymers or are involved in information transmission, such as metal atoms, coenzymes, activators, suppressors, etc. Cofactors, low molecular weight compounds, biopolymers such as nucleic acids and sugars, or water molecules. In particular, the term ligand or substrate is used when the external factor is a low molecular weight compound specific for a biopolymer. The complex formation site means a specific interaction site to which a ligand, cofactor or biopolymer specific for the biopolymer binds.

In the present specification, the “interaction form” means the type of elements (in this case, atoms and atomic groups having different chemical bonding environments) involved in a certain interaction and the positional relationship between the elements. For example, typical hydrogen bonds, salt bridges, CH-π interactions, and the like are given as typical interaction forms.
Further, in the present specification, an “interaction form type” is a classification of individual atoms and atomic groups constituting a certain interaction form classified according to the chemical bonding environment and the type of adjacent heavy atoms. For example, one of the interaction forms described above is an interaction form type that forms a classical hydrogen bond, which includes an acidic proton (or an atomic group containing it) and a basic heteroatom (or this) that accepts it. As an interaction form type that constitutes a salt bridge that is another interaction form, a carboxyl anion and an ammonium ion, and an interaction form type that constitutes a CH-π interaction, which is another interaction form. Includes a hydrogen atom (or an atomic group containing them) bonded to a neutral carbon and a π-bonded carbon atom (or an atomic group containing them). Each interaction form type is defined based on the type of atom and atomic group, based on its chemical bonding environment and the type of adjacent heavy atom, so that the interaction point involving the atom and atomic group also depends on the interaction form. It will be specific to the type.

1. First Step In the first step of the present invention, atomic coordinate information of a biopolymer is prepared. In this case, it may be handled as one body including other molecules bonded to the biopolymer. In this specification, the atomic coordinate information may be in any format as long as it is information for specifying the position of the constituent atom of the biopolymer, but is preferably specified in a three-dimensional coordinate system. This is determined, for example, by a technique such as X-ray crystal structure analysis or NMR analysis. As atomic coordinate information that can be used in the present invention, for example, “Protein Data Bank (http: //pdbj.protein.osaka- u.ac.jp/) ”.
Normally, hydrogen atoms are not added to the three-dimensional structure at the time of reading the atomic coordinates of the biopolymer system. This is because the atomic coordinates regarding the hydrogen atom of the biopolymer are generally not experimentally obtained. In general, for hydrogen atoms, the dihedral angle is automatically calculated from the distance of literature values, the bond angle from the valence of the heavy atom to be bonded, and the bond information on the adjacent heavy atom, and the hydrogen atom can rotate. If it is above, it can be added in a trans conformation. Thus, it is easy for those skilled in the art to calculate the atomic coordinates of the hydrogen atom contained in an amino acid residue from the structure of the amino acid residue, and various programs that automatically perform this calculation are also available. Exists (HyperChem ™ Professional 7.51, Hypercube, Inc., 1115 NW 4th Street, Gainesville, Florida 32601, USA). For example, Homology Modeling for HyperChem, Revision A3 (Institute of Molecular Function; Saitama, Japan, http://www.molfunction.com), an initial coordinate evaluation program for hydrogen atoms (http://www.molfunction.com/algorithm.htm) ) Can be used, but any method or program may be adopted. As will be described later, the interaction point related to a hydrogen atom is defined from the heavy atom to which the hydrogen atom is bonded, and in consideration of the distance from the hydrogen atom, a plurality of interaction points can be defined at a certain rotation angle. For example, the hydrogen atoms may be arranged in any direction.

2. Contents of the second step (1) interaction point information database In the second step of the present invention, the position of the interaction point of the biopolymer is specified based on the interaction point information database. The interaction point information database refers to the mutual relationship between structural units of biopolymers (for example, 20 types of amino acids for proteins, 5 types of nucleobases and 2 types of sugars for nucleic acids, and typical monosaccharides for polysaccharides). It means a database where action point information has been collected in advance and made available for reference.
The interaction point information database includes at least “position information” for specifying the position of an interaction point with respect to a structural unit (for example, each of twenty kinds of amino acids) of a biopolymer. Furthermore, it is preferable to include “physicochemical information” of each interaction point.
Here, “positional information” means information necessary to determine the position of the interaction point of the biopolymer, and is created in any format as long as the position of the interaction point in the three-dimensional space can be specified. May be. For example, by specifying three atoms (alpha position, beta position, and gamma position) in the biopolymer structural unit, and specifying the distance, bond angle, and dihedral angle to them, the coordinates of the interaction point are unambiguous. As an example, the information in the form of a matrix can be given.

In addition, the “physicochemical information” of the interaction point means information for expressing the property of the interaction point derived from the atom, which is obtained based on the physical or chemical property of the biopolymer constituent atom. . This information is not related to the determination of the position of the interaction point.
In the present invention, for example, as “physicochemical information”, whether the interaction energy of the interaction form including the atom that generates the interaction point is based on the contribution of the electron correlation interaction energy (dominance of the electron correlation term). ) Or the contribution of electrostatic interaction energy as the main component (dominance of electrostatic term), and when the contribution of electrostatic interaction energy is the main component, the atom generating the interaction point is positively charged (Ie, electron accepting) or negative charge (ie, electron donating), and can be recorded in the database as C, P, N, etc., respectively.
In another embodiment, if the Mariken atom charge of the atom that generates the interaction point is added to the database as information, the charge of the atom that generates the interaction point dominated by the electron correlation term is close to zero. Using this as an index, the chemical requirement on the heavy atom of the potential interaction partner can be determined.

The total interaction energy, E _tot , that contributes to the stabilization (or destabilization) of a certain interaction form is the electron correlation, in particular the energy contribution of the dispersion force (ie the electron correlation interaction energy, E _corr ; Can be defined from the energy contribution of the electrostatic interaction (ie, the electrostatic interaction energy, E _es ; hereinafter simply referred to as the electrostatic term) As long as a theoretical calculation that can handle electron correlation is used, a person skilled in the art can perform the calculation by an arbitrary method. For example, when the Meller-preset second-order correlation perturbation theory is used as the theoretical calculation that can handle the electron correlation for the complex calculation, it can be estimated from Equation 1. Further, in the case of using other theories need only replace the interaction energy obtained the E _MP2 of formula 1 from該理theory. However, in practicing the present invention, it is sufficient to know the relative strength rather than the absolute value.
In this specification, the case where the electron correlation interaction energy contributes to the stabilization of more than a majority of the total interaction energy in a certain interaction form is governed by the electron correlation term, and conversely the electrostatic interaction energy contributes. This is expressed as electrostatic term control.

Formula 1

As described above, as the physicochemical information of the interaction point, only three types of information whether the atom involved in the interaction is positively charged, negatively charged, or otherwise are useful as information. It is also possible to classify in more detail.
For example, the hydrogen and carbon atoms in the alanine, leucine, valine, isoleucine, and threonine side chain terminal methyl groups are generally in different environments on their respective residues, but are equivalent in chemical environment, and therefore Since the interaction forms that involve them and the physicochemical properties of the interactions in the interaction forms (eg, interaction energy, interaction distance and angle) are equivalent, one type, ie one interaction form type Can be handled as
Here, the equivalent chemical environment means that the chemical bonding environment and adjacent heavy atoms are the same, and does not mean strict chemical equivalence. For example, in the case of using NMR or infrared absorption (IR) reflecting the chemical bonding environment, if there is no shielding effect by remote atoms or atomic groups, in proton NMR, the proton energy of each terminal methyl group The absorption positions cannot be distinguished from each other. Similarly, in carbon NMR, the energy absorption positions of the carbon atoms of each methyl group should be close enough to be difficult to distinguish from each other. In the IR as well, these infrared absorption positions should overlap. Thus, those skilled in the art can set the interaction form type found in a protein molecule based on the chemical binding environment. However, the interaction form type can be variously set as long as the chemical properties are not greatly impaired, and the interaction form type used in the present invention is not limited to these.

Table 1 shows examples of interaction morphology types for protein molecules. For proteins, the interaction form type can be specified from the following table and registered as information of each atom on the database.
Table 1 Abbreviation notation of interaction form type

In the above table, for example, “methyl, methylene proton” means that the chemical bonding environment is equivalent for three hydrogen atoms on the primary carbon (methyl group) and two hydrogen atoms on the secondary carbon (methylene group). It means an atomic group (or atom) of amino acids constituting a protein.
By performing complex analysis using a model compound for each interaction form type, position information necessary for creating an interaction point information database described later can be collected.
Furthermore, it is also possible to collect information about the actual interaction energy value, or whether the interaction is energetically stabilized or destabilized. This is useful when examining the interaction between the polymer and the ligand, for example, when performing molecular superposition of interaction points.

(2) Creation of interaction point information database The positional information and physicochemical information necessary for the creation of the interaction point information database are obtained by methods known in the art, for example, biopolymer constituent units (for example, for protein molecules). Between two bodies between low-molecular-weight model compounds corresponding to constituent fragments contained in 20 types of amino acids, 5 types of nucleic acid bases and 5 types of nucleobases and 2 types of sugars, and 1 type of polysaccharides) It can be obtained by comprehensively performing a high-level quantum chemical calculation of an inter-body complex model, that is, a theoretical calculation at a level that has a sufficient amount of basis functions and can handle electron correlation in the complex calculation.
Here, the fragment means, for example, twenty kinds of amino acid side chains and main chain amides including a protonated state in a protein molecule, and further, a small number of further divided atomic groups constituting these amino acid side chains. . The low molecular weight model compound corresponding to the fragment can be modeled, for example, with the alanine side chain as a methane molecule, and the model compound of the terminal methyl group of leucine, isoleucine, valine or threonine should be a methane molecule as well. (See FIG. 2). The model compound may be an interaction form type itself found in a protein or may include the type as a partial structure. Thus, the electronic state of the entire biopolymer can be reconstructed from the fragment or the model compound corresponding to the fragment divided into the minimum unit structure that does not significantly impair the chemical properties. However, various fragmentation methods can be used as long as the chemical properties are not significantly impaired, and the model compounds used in the present invention are not limited to these.

In addition, it is preferable to use a compound having a structure with as high a symmetry as possible as a model compound, and it is preferable to create an initial structure of a complex model between two bodies in consideration of symmetry (FIG. 3). By considering symmetry, the interaction analysis is simplified and the actual computational resources are overwhelmingly small. Furthermore, since there is a high possibility that a highly symmetrical geometry is an energy minimum or maximum point, there is an advantage that an energy minimum point can be efficiently searched.
Based on the calculated results, as parameters of each atom of the biopolymer corresponding to the fragment, “physicochemical information on distance, angle and interaction form between heavy atoms involved in the interaction between model compounds” ”(Eg, interaction energy, electron correlation interaction energy, electrostatic interaction energy, Mulliken atomic charge, etc.) is collected. For example, in the following documents, methylammonium ion and benzene, methylammonium ion and acetic acid, benzene and methane, benzene and ammonia, benzene and water, toluene and ammonium ion, p-cresol and ammonium ion, methylindole and ammonium ion, toluene This is a case where interaction analysis was actually performed using guanidinium ion, p-cresol and guanidinium ion, methylindole and guanidinium ion.
(1) Gallivan, et.al .; J. Am. Chem. Soc, 2000, 122 870-874
(2) Suzuki, et.al .; J. Am. Chem. Soc, 2000, 122 3746-3753
(3) Suzuki, et.al .; J. Am Am. Chem. Soc, 2000, 122 11450-11458
(4) Tarukeshwar, et.al .; J. Am. Chem. Soc, 2000, 123 3323-3331
(5) Minoux, et.al .; J. Am. Chem. Soc, 1999, 121 10366-10372

  In the creation of the database used in the practice of the present invention, for example, methane as the model compound of the alanine side chain, methane thiol as the model compound of the cysteine side chain, acetate anion as the model compound of the aspartate anion side chain, phenylalanine side chain Toluene as a model compound, 5-methyl-3H-imidazole ion as a model compound of protonated histidine side chain, methylammonium ion as a model compound of lysine side chain terminal ammonium group, dimethylthioether as a model compound of methionine side chain terminal methyl group, Acetamide as a model compound of asparagine side chain, N-methylguanidinium ion as a model compound of arginine side chain guanidinium group, methanol, tryptophan side chain as a model compound of serine side chain As a model compound 3-methyl -1H- indole, p- cresol as a model compound of tyrosine side chains, also can be used N- methylacetamide as a model compound in the main chain peptides.

As described above, a method for searching for an energy minimum point for an atomic group, that is, a point that can be an interaction point is known in the art. It is also possible for those skilled in the art to obtain the energy minimum point similarly for the side chain of each amino acid residue using the model compound as described above (Tsuzuki, et.al .; J. Am. Chem. Soc, 2000 , 122 3746-3753, Tsuzuki, see et.al .; J. Am Am. Chem. Soc, 2000, 122 11450-11458), in fact well known to the amino acid residues There are also many classical interaction points that are involved in so-called hydrogen bonds and salt bridges, which have a relatively large interaction energy. As described above, the information on the interaction point for each amino acid can be collected by the person who implements the invention, arbitrarily selecting from known information, or by performing the above-described calculation. In the invention, an interaction point information database containing information collected by either method can be used.
In addition, it is desirable to adopt the distance and angle between heavy atoms as the positional information for specifying the interaction point. However, when a hydrogen atom is involved in the interaction, the distance from the hydrogen atom is also taken into consideration. It may be defined from the bonded heavy atom or from the hydrogen atom. In addition, when the hydrogen atom is rotatable, the interaction point may be defined as a single or a plurality of constant rotation angles from the bonded heavy atoms. Even when the interaction partner is a hydrogen atom, it is preferable to adopt the distance and angle to the heavy atom to which the hydrogen atom is bonded. By doing this, the position of the predicted interaction point is ideally calculated as the most stable position of the counterpart heavy atom, so the shape of the external factor having an affinity for the predicted complex formation site Complementary chemical requirements are estimated and visualized based on heavy atoms.

The number of interaction points included in each amino acid residue database is preferably at least 1 or more, for example, 5 or more, for example 5 to 40, but scientifically, each is an interaction point. As long as it is judged, there is no restriction on the upper limit and the lower limit of the number, and those who carry out the invention can freely set them. For example, a situation where an interaction point for a specific amino acid residue is not set is allowed. Furthermore, for example, when it is intended to predict only the interaction point dominated by the electron correlation term, it is naturally possible to select in advance the interaction points included in the database according to the situation. Each interaction point on the database is collected as a candidate that can be considered as an interaction point of a given biopolymer. Therefore, although it is conceivable that the number of interaction points of the biopolymer that is finally predicted (that is, remains until the end as a candidate) increases if the number of interaction points registered in the database is large, the present invention is It is essentially unrelated to the question of whether it is feasible.
In the interaction point information database, the position information (that is, the distance and angle in the interaction point information) is Z- using the internal position information of the biopolymer (for example, the name of an atom given in PDB format). A database is created in a relative coordinate system such as a matrix, and preferably, an interaction form type and physicochemical information (for example, a Maliken atomic charge) are added thereto. By making the interaction point information in a database in this way, the atomic coordinates of the biopolymer system are read, and at the same time, the interaction points on the biopolymer three-dimensional structure are referred to by referring to the position information in the interaction point information database. Coordinates can be calculated almost instantaneously.

FIG. 4 shows an example of an interaction point information database in which interaction point information of interaction points with respect to an alanine residue (11 positions in this example) is converted into a database in the form of a Z-matrix. FIG. 5 shows the “position” and “physicochemical information” of the 11 interaction points in relation to the alanine residue.
In the matrix of FIG. 4, each of 11 rows represents information of one interaction point. By applying the values of (distance, bond angle, dihedral angle) in 2-4 columns with the alpha, beta, and gamma positions specified in columns 5-7 as the reference, the position of the interaction point is It is uniquely identified. That is, the position of the interaction point of the alanine residue can be uniquely determined by the information of 2 to 7 columns, and the information of 2 to 7 columns corresponds to “position information”. In column 9, the malken atomic charge is added to the database as “physicochemical information”, and in FIG. 5, the interaction points are displayed in different colors according to the value of the malicen atomic charge. Moreover, the interaction form type is described in 8 columns.
As described above, “positional information” and “physicochemical information” are stored in a database in advance so that they can be referred to, so that each alanine residue constituting a given biopolymer can interact with each other. The position and nature of the points can be determined instantaneously and visualized if necessary. For example, with regard to proteins, if there is an interaction point information database for 20 types of amino acids, interaction points for all amino acid residues contained in a given protein can be instantly determined and displayed as an image. .

(3) Identification of the position of the interaction point candidate As described above, in this step, the interaction point information database is referred to from the atomic coordinates of the input biopolymer and based on the position information contained therein. The coordinates of the interaction point with respect to the constituent atoms are specified. The determination of the coordinates of the interaction point performed in this step can be easily performed by a known method simply by applying the position information stored in the database to the atomic coordinates of the structural unit of the biopolymer, that is, mechanically. . At that time, constituent atoms contained in the entire biopolymer may be targeted.For example, when there is a target region or when other molecules are bonded, a part of the region is designated and processed. May be.

An example of an embodiment of the second step of the present invention is shown below.
Enter the atomic coordinates of a biopolymer (eg, protein). Select the target range if necessary. Select the constituent atoms of the biopolymer. Accessing the interaction point information database, the position information corresponding to the constituent atom is matched with the name of the atom (for example, the name of the atom given in PDB format) and the name of the constituent unit (for example, the amino acid name) to which it belongs. get. The position of the interaction point candidate of the biopolymer is determined based on the position information. If other constituent atoms remain, select the next constituent atom. If it does not remain, the process ends.

This process allows the interaction point to be determined from the coordinate information without considering the characteristics of the entire biopolymer (for example, the distribution of the hydrophobic surface region and the shape of the surface irregularities in the case of proteins) as in the conventional technique. Candidate positions can be specified, and three-dimensional display can be performed together with the structure of the biopolymer if necessary. In addition, when the interaction point information database includes physicochemical information, the position of the interaction point and the physicochemical information are obtained by visualizing the physicochemical information by, for example, character notation, color coding, etc. Can be referenced simultaneously.
In particular, when the physicochemical information of the interaction point candidate includes a maliken atom charge, the obtained interaction point candidate is displayed as a sphere of about 2 to 3 mm, so that the hydrophobic region or hydrophilicity on the surface of the biopolymer is displayed. The distribution of the sex region is computerized in the same way as mapping the electrostatic potential on the surface of the biopolymer van der Waals. Furthermore, by applying clipping of three-dimensional coordinates, the electrostatic state inside the biopolymer can be confirmed while slicing.
Note that the interaction points calculated in this step may be expressed as “calculated interaction points”, “calculated interaction points” or “interaction point candidates” in the following description in accordance with the text. However, the interaction points calculated in this step are the ones that comprehensively and mechanically indicate the interaction points of the constituent atoms as candidates for the interaction points of the biopolymer (eg protein) itself. However, this does not mean that the biopolymer is actually involved in the interaction with a ligand or the like.

3. Third Step In the third step of the present invention, among the interaction point candidates identified in the second step, the interaction points involved in the interaction with the constituent atoms of the biopolymer are removed.
In the present specification, the internal structure of the biopolymer system means a term only for distinguishing the biopolymer surface from the inside thereof. This distinction is made because constituent atoms buried inside biopolymers are stabilized by interaction with surrounding constituent atoms, while constituent atoms protruding on the surface of biopolymers are solvent molecules, etc. This is due to the great difference that the energy stabilization (solvation energy) is also obtained by the interaction.
The interaction point that is determined to contribute to the interaction with the constituent atoms is already involved in the stabilization of the internal structure of the biopolymer, and is considered to have no room for other purposes such as ligand binding. Therefore, in this step, it is determined whether each interaction point calculated in the second step contributes to the interaction with the constituent atoms of the biopolymer according to a certain standard, and is determined to contribute. Things are removed from the calculated interaction point.
Examples of the criterion for determining whether or not to contribute to the interaction with the constituent atom include “distance between the interaction point candidate and the constituent atom of the biopolymer”. That is, it is possible to exclude an interaction point that satisfies a certain distance condition (hereinafter also referred to as an interaction distance condition) with a constituent atom of the biopolymer. That is, in this step, preferably, for each interaction point candidate specified in the second step, the interaction point candidate is removed when a constituent atom of the biopolymer exists within the predetermined distance.

  The interaction distance condition can be set by a person skilled in the art according to the interatomic energy curve from the average distance of the effective interaction distance from the non-bonding interaction equilibrium distance of the heavy atoms constituting the biopolymer, for example, In the case where the interaction condition is applied only to the heavy atoms in the biopolymer system, it is possible to set, for example, in the range of 2 to 4 cm, and among these, the non-bonding interaction of the biopolymer constituent heavy atoms It is preferably set from 3 to 4 mm which is a general effective distance of 3.5 mm, particularly preferably 3.5 mm. It is possible to apply the interaction condition to the hydrogen atom (see FIG. 6). The interaction distance condition that can be adopted in this case is 2 to 2 obtained by subtracting the van der Waals radius of the hydrogen atom from the effective interaction distance. It is preferable to set from 3 mm, and 2.5 mm is particularly preferable. In addition, it is not necessary to apply the same distance condition to all interaction point candidates. For example, when the physicochemical properties of the interaction point candidates are known, the distances are individually determined for the interaction points having the properties. Conditions may be set.

The average distance of the effective interaction distance from the non-bonding interaction equilibrium distance of heavy atoms will be specifically described below. For example, van der Waals interactions (which account for the majority of biopolymer internal interactions), orbital interactions, and charge transfer interactions, which are short-range interactions, are atomic van der Waals involved in these non-binding interactions. There exists an interaction stabilization region (interaction effective distance from the equilibrium distance) that depends on the van der Waals radius of the atom, with the interaction energy minimum point (equilibrium distance) near the radius. Since the van der Waals radii of carbon, nitrogen, and oxygen atoms, which are the main constituent heavy atoms of protein molecules, are almost equal, the effective distances of short-range interactions involving these heavy atoms inevitably become almost the same. The distance can be used as an average distance. Electrostatic interaction, dipole interaction, etc. are interactions that have an effective distance up to a long distance (10 cm or more), but the long distance interaction is strong within the densely packed biopolymer. It is rare to be canceled out by interaction, and it is considered that the effects of these long-range interactions are almost eliminated due to the high dielectric constant on the surface of biopolymers. In the case of interactions, only short distances are considered.
A person skilled in the art can calculate the distance between each interaction point candidate and a constituent atom of the biopolymer by any method. Since the coordinates of the constituent atoms and interaction points of the biopolymer are specified, it is clear that the distance between these two arbitrary points can be easily calculated.
The distance between two points may be calculated comprehensively for all combinations of constituent atoms and calculated interaction points of the biopolymer, but it is also possible to improve the calculation speed by limiting the spatial range, for example. is there. As an example, in the case of dealing with a three-dimensional structure of a biopolymer, it is generally performed in this technical field to simplify a calculation by dividing a three-dimensional space with a grid.

For example, the following method can be adopted. Divide the entire biopolymer system or a specific area with a grid at regular intervals, and record whether each cube constituting the grid (hereinafter referred to as a cell) contains constituent atoms of the biopolymer system. If the cells of the biopolymer system that satisfy the interaction distance condition are included in the same cell as the cell containing the interaction point or in a cell adjacent to the cell, the distance between those atoms and the calculated interaction point is calculated. Then, the method to do can be adopted.
It should be noted that the grid spacing can be arbitrarily determined by those implementing the present invention in relation to the interaction distance condition to be adopted, but in general, any value above the interaction distance condition may be used. it is obvious. Further, in this specification, “adjacent cells” mean 26 cells that share a certain cell, side, or vertex with a specific cell as the center.
By adopting the above method, it is possible to perform calculation at an overwhelmingly higher speed than the exhaustive calculation for all combinations of interaction point candidates and biopolymer constituent atoms.

An example of the embodiment of the third step of the present invention is shown below.
Select the range that needs to be calculated and divide it in a grid. Specify one interaction point candidate. It is examined whether or not a biopolymer atom exists in a cell including the interaction point candidate and a cell adjacent to the cell. If it does not exist, specify the next interaction point candidate. If it exists, the distance between the interaction point candidate and the biopolymer atom is calculated, and it is confirmed whether or not the interaction distance condition is satisfied. If there is an atom that satisfies the mutual distance condition, the interaction point candidate is deleted. If there is nothing to satisfy, specify the next candidate interaction point. If no interaction point candidates remain, the process ends.

4. Fourth Step In the fourth step of the present invention, among the interaction point candidates remaining after the third step, the interaction points involved in the interaction between the surface of the biopolymer system and the solvent molecules are removed. When performing the fourth step of the present invention, the fourth step may be performed first for the interaction point candidate specified in the second step, and then the third step may be performed. It is.
Biopolymers usually interact with surrounding water molecules and are solvated in solution. In other words, at normal temperature, the solvent molecules are in equilibrium with the biopolymer surface while oscillating, and most of the interaction points calculated on the biopolymer surface are involved in solvation with the solvent molecules. it is conceivable that. Therefore, the interaction point candidates can be narrowed down by removing the interaction point candidates involved in the interaction with the solvent molecule.
The interaction point involved in the interaction between the surface of the biopolymer and the solvent molecule means the interaction position that contributes to the stabilization of the biopolymer by the solvation energy (ie, the biopolymer is soluble in the solvent). To do. Criteria for determining whether or not a candidate interaction point is an interaction point involved in interaction with a solvent molecule can be arbitrarily set by those skilled in the art based on scientific evidence. It can be determined on the basis of the distance condition between the molecule surface constituent atoms. For example, among the crystal water obtained by X-ray crystallographic analysis of the biopolymer, what is said to be predominantly hydrogen-bonded to the biopolymer is at a distance of 2.0 to 2.5 cm from the hydrogen atom, This corresponds to the sum of the van der Waals radius of the water molecule and the van der Waals radius of the atoms constituting the surface of the biopolymer, that is, the solvent contact surface.
Examples of molecular surfaces that can be employed include solvent contact surfaces and van der Waals surfaces. That is, in the fourth step of the present invention, preferably, interaction point candidates existing outside the solvent contact surface or van der Waals surface of the biopolymer are removed.

Here, the van der Waals surface means a surface formed by the van der Waals radius of each atom constituting the molecule. The solvent contact surface is generally defined by the locus of the center of the sphere when the sphere having the van der Waals radius of the solvent molecule (1.4 Å is generally used for water molecules) freely moves on the van der Waals surface. In the present specification, the radius of the sphere does not necessarily need to be the van der Waals radius of water molecules, and it depends on a desired degree of narrowing down when the present invention is carried out. The person who implements the invention can arbitrarily set. For example, if you want to keep the outer interaction point, or if you want to search for a region that forms a complex with an external factor such as a polymer and a convex part on the surface of the molecule, set the van der Waals radius of the solvent. The corresponding distance condition can be increased (for example, 1.5 to 2.5 mm). Further, if the interaction point related to solvation with water molecules is to be removed, a radius that is about 20-30% smaller than the van der Waals radius of water, for example, 1.0 kg, should be used. In order to search for a region that forms a complex with a small dent, in order to distinguish from other interaction points calculated on the protein surface, the distance condition can be reduced (for example, 0.5 to 1). 0.0Å), it is not necessary to carry out this process in the first place.
That is, in this step, it is preferable that a person implementing the invention can select which of the solvent contact surface and the van der Waals surface is applied here. For example, if you do not want to miss the interaction point such as the substrate binding site in the shallow depression on the biopolymer surface, you can use the solvent contact surface and know that there is a ligand binding site in the cavity inside the protein The inner van der Waals surface may be used to remove all candidate interaction points outside the protein surface.

  A person skilled in the art can determine whether each interaction point candidate exists outside the solvent contact surface or van der Waals surface of the biopolymer by any method. For example, the determination may be made using surface coordinates obtained by the Connolly et al. Molecular Surface Package (http://www.biohedron.com/). Further, for example, it is possible to cover the biopolymer with a cube, divide it with a grid, determine the boundary cell on the molecular surface, and determine the distance based on whether it is inside or outside (FIG. 7). reference). Here, the border cell may be determined by any method known to those skilled in the art. See, for example, FIG. That is, in the fourth step, the biopolymer is divided by a grid, the boundary cell of the solvent contact surface or van der Waals surface is determined, and the interaction point candidates included outside the boundary cell are removed. It is possible.

An example of an embodiment of the fourth step of the present invention is shown below.
Determine the border cell. It can be determined by the following method.
(1) When van der Waals surface is used
(i) The entire biopolymer system is divided by a grid with a constant interval. The smaller the grid interval, the higher the accuracy. However, in consideration of computer resources, about 1.5 mm is sufficient.
(ii) Record whether or not each of the cubes constituting the grid (hereinafter referred to as cells) contains constituent atoms of the biopolymer.
(iii) A row of cells arranged in a line in each of the x, y, and z directions is scanned from both ends, and in each scan, a fan of atomic species included in the cell in which the constituent atoms of the biopolymer first appear Flags all cells that are outside by the distance of the Delwals radius.
(iv) The xy plane, the yz plane, and the zx plane are scanned again from both ends, and all the boundary cells are determined by using the flagged cell that appears first as the boundary cell.

(2) When using a solvent contact surface
(i) The entire biopolymer system is divided by a grid with a constant interval. Here, the grid interval can be a value of a desired distance condition, for example, 1.4 mm.
(ii) Record whether each cell constituting the grid contains a constituent atom of the biopolymer.
(iii) A row of cells arranged in a line in each of the x, y, and z directions is scanned from both ends, and in each scan, a fan of atomic species included in the cell in which the constituent atoms of the biopolymer first appear Flags all adjacent cells of the cell that is outside by the distance of the Delwars radius.
(iv) The xy plane, the yz plane, and the zx plane are scanned again from both ends, and all the boundary cells are determined by using the flagged cell that appears first as the boundary cell.

After the boundary cell is determined by the above method (1) or (2), the interaction point candidates existing outside the boundary cell are deleted.

  As described above, in the third step and the fourth step of the present invention, whether the calculated interaction point has already been used for stabilizing the internal structure of the biopolymer, and regarding the surface of the biopolymer system, the solvent molecule It was made possible to determine whether or not it is involved in the interaction. In other words, when interaction points are not involved in any of these, it is very likely that these interaction points are involved in interaction with other external factors or interaction with their own three-dimensional structure unknown region. When interaction points that have not been used can be displayed together with biopolymers on the three-dimensional structure visualization program, as shown in the examples below, a complex of the site where these interaction points exist and the actual external factors Not only does the formation site completely match, but also the shape and interaction physicochemical information obtained by three-dimensionally connecting multiple interaction points at the predicted interaction site. It was found that the shape and even chemical requirements of potential external factors can be accurately predicted.

  The present invention can be applied not only to proteins but also to other biopolymers such as nucleic acids and sugars. In that case, as described above, the atomic group constituting each biopolymer (for example, five types of nucleobases and two types of sugars, which are the structural units of nucleic acids, and a typical monosaccharide is preferable for polysaccharides). It is sufficient to collect “location information” of interaction points and, if necessary, “physicochemical information” according to the method shown in this specification, and to create a database similar to that shown in this specification so that it can be referred to. It is. For example, for each base, the energy minimum can be calculated using substituted purine, substituted pyrimidine, D-ribose and D-2-deoxyribose as model compounds. Since nucleic acids and sugars are overwhelmingly hydrophilic compared to proteins, it is necessary to use reaction field model calculations that take into account the solvation effect in model calculations. Although there is a difference that strong stabilizing interactions such as electric interactions and dipolar interactions hardly contribute to these biopolymers (Gallivan, et.al .; J. Am. Chem. Soc, 2000, 122 870-874), the essence of the interaction itself does not change, and it goes without saying that all molecules are targeted in this method using the interaction point data.

5. Utilization of Interaction Point When the present invention is implemented, the interaction point prediction can be completed in only a few seconds even with a relatively large biopolymer system or a general personal computer. In addition, since a hydrogen atom is automatically added in a normal program, no special processing such as editing biopolymer data provided from a protein data bank is required. However, when it is desired to calculate the interaction point with higher accuracy, structure optimization may be performed on the automatically added hydrogen atom or side chain.
In the present invention, the predicted interaction points are preferably displayed on a display or other medium. The display is preferably three-dimensionally displayed together with the biopolymer. If the interaction point information database contains physicochemical information, categorize the information, for example, whether the interaction energy at the interaction point is dominated by electronic correlation terms or electrostatic terms. Can be displayed in different colors so that they can be distinguished. Here, the color classification is not limited as long as it can be visually discriminated, and includes display widely distinguished by colors, patterns, and the like. As described above, by displaying the predicted interaction point position and its physicochemical information on the computer screen together with the biopolymer, the interaction point information can be easily recognized. For the display of biopolymers and interaction points, all programs such as HyperChem, RasMol, Swiss-PDB Viewer etc. can be used as long as they can recognize the coordinate format (PDB format) provided by Protein Data Bank. It is possible to use.

The interaction point obtained by the present invention is specified as a “point” at the heavy atom position of the molecule or atom that is the interaction partner, and in some cases physicochemical information of the interaction point is added, so vaguely predicting only the region Unlike methods, the most important information for molecular superposition technology that accurately predicts the shape (scaffold) and chemical requirements (pharmacophore) of the interaction partner and forms the core of structure-based drug discovery It is possible to provide scientifically meaningful “point” information.
The predicted interaction point is a potential interaction point in the target biopolymer, and a site where a plurality of interaction points are distributed is a potential interaction site. In other words, the interaction points predicted once and locally can be assumed to be positions where relatively small small molecules such as metal atoms, water molecules, and phosphate molecules that compose a biopolymer system interact. The site where the site of action is distributed is presumed to be a complex formation site of low molecular weight compounds such as ligands and coenzymes, a complex formation site of cofactors and other biopolymers, or an interaction site with its own remote structure it can. Furthermore, when the present invention is applied to a complex with an external factor such as a ligand, if an interaction point is calculated around the ligand, this is a guideline for structural modification for improving the affinity of the ligand. Is useful as a guide for structural requirements related to ligand agonism and antagonism control. In addition, the side chains and main chain conformations of biopolymers are useful for predicting the induction conformity effect associated with complex formation. In addition, in the whole or partial modeling of a biopolymer, it is useful for evaluating the stability of the biopolymer, such as searching for a stable conformation using interaction point information.

In the present invention, each interaction point in the predicted interaction site is used as a superposition point on the biopolymer side, and superposition with an arbitrary low molecular weight compound, whereby the above interaction point or mutual point is achieved. It is possible to estimate chemical groups, chemical species, molecular species, etc. that contribute to the stabilization of the action site. As a result, it is possible to estimate whether any molecule can form a complex at the interaction site, and the interaction with individual interaction points can be determined from the structure of single or multiple small molecule compounds that have been successfully superimposed. It is possible to estimate chemical groups and the like that are effective.
Any method, algorithm, etc. known in the art may be used to perform superposition. For example, superposition using the least square method can be mentioned.
In the predicted interaction site, it is optimal to use, for example, 3 to 5 interaction points for superimposition in consideration of the computer processing power. However, all the predicted interaction points or at least one or more interaction points are used. Any number of interaction points may be used.

Moreover, it is possible to perform more efficient calculation by adding an interaction form type to each heavy atom of the compound used for superposition in advance.
Since the interaction mode type on the low molecular weight compound side used for superposition is only used for superposition with the superposition point on the biopolymer side, it is not always necessary to assign it to all heavy atoms. You can just assign them to atoms. In addition, if the heavy atom is in a chemical bond environment equivalent to the interaction form type found in the protein, the same physicochemical information can be used, and it corresponds to the chemical bond environment found in halogen atoms and other pharmaceutical compounds. The interaction form type can be added, and the physicochemical information of each combination of the interaction forms can be added to the matrix data described in Table 1 and below.
The low molecular weight compound to be superimposed is preferably given as a three-dimensional structure having a stable conformation, but may be any two-dimensional structure or three-dimensional structure as long as coordinates are given. At this time, the low molecular weight compound may be any compound such as a commercially available compound, an original compound, or a virtual compound, and may be a simple substance or may include a plurality of molecules.

The superposition is performed for all combinations resulting from the number of atoms added with the number of interaction points selected as the superposition point on the biopolymer system side and the interaction form type of the low molecular weight compound. By referring to data in matrix format and considering only possible combinations of interaction form types (combinations that do not destabilize) at all superposition points, chemically superimposing superposition is possible. Can be logically excluded. Thereby, a huge number of unnecessary calculations can be logically omitted, and as a result, a highly accurate and high-speed docking simulation can be realized.
As described in the examples below, when the docking simulation was performed with an arbitrary low molecular weight compound by using the whole or a part of the interaction point as a superposition point on the biopolymer side, the most stable complex structure was high. It was found that it can be obtained with accuracy. Furthermore, it has been found that even a guideline on dynamic behavior of the internal structure of the biopolymer and enhancement of affinity of external factors can be obtained. In this way, it is related to the recognition of external factors such as ligands or remote structures of its own in any biopolymer system, and predicts the interaction point that holds the physicochemical information necessary for recognition based on the chemical basis. It became possible to use this.

Further, in the present invention, a low-function molecule such as a drug candidate compound is designed directly at an interaction site of a biopolymer using the interaction point information of the predicted single or plural interaction points. De Novo design is also possible.
In the process of creating the database of interaction point information, if the interaction point is defined as the most stable position of the partner heavy atom, considering possible combinations of interaction form types on any interaction point, If a heavy atom or a heavy atom to which a hydrogen atom is added is arranged and these heavy atoms are within the bond distance range, they can be connected by a bond, and if they are outside the bond range, they can be connected by a linker through an arbitrary linker. it can. Also, hydrogen atoms can be added later to each heavy atom based on the generated bond information.
In addition, by partially superimposing single or multiple interaction points and functional groups or atomic groups, the compounds are similarly designed using single or multiple functional groups or atomic groups arranged. You can also.
Examples of the present invention will be described below, but the present invention is not limited thereto.

In order to show the effectiveness of the present invention, as an example of plasma human retinol-binding protein, potential ligand binding in the previous three-dimensional structure (hereinafter referred to as apo form) that forms a complex with the retinol molecule that is a natural ligand. It shows that a site can be predicted.
Pretreatment of the biopolymer system is performed using Homology Modeling for HyperChem (Homology Modeling for HyperChem, Revision A3; Institute of Molecular Function; Saitama, Japan, http://www.molfunction.com) developed by the present inventor. Carried out. In addition, the three-dimensional structure of the biopolymer system used was provided from Protein Data Bank (hereinafter referred to as PDB, http://www.pdbj.org/index.html). As the interaction point information database, a database created by the inventor of the present application (version created on January 15, 2006) was used.
The three-dimensional structure of plasma human retinol binding protein was PDB ID: 1BRQ provided by PDB. The N- and C-termini of this protein are treated as Zitter ions, and the initial coordinate of the hydrogen atom of the water molecule contained in the three-dimensional structure is the initial coordinate evaluation program for hydrogen atoms of Homology Modeling for HyperChem (http: // www .molfunction.com / algorithm.htm). All hydrogen atoms contained in the protein system were optimized using the Amber94 force field and Polak-Ribiere minimization algorithm with an RMS gradient of 1.0 kcal / Åmol as the termination condition.

The position of the interaction point on the protein was identified based on the interaction point information database. At this time, interaction points for hydrogen atoms of hydroxyl groups and thiol groups are generated in increments of 60 degrees, and interaction points for heavy atoms of hydroxyl groups, thiol groups, and thioethers (ie, oxygen atoms and sulfur atoms) are also generated in increments of 60 degrees. I let you. The interaction requirement with the constituent atoms of the biopolymer was taken into consideration the distance to the hydrogen atom, and the grid size was set to 2.5 mm and the interaction distance condition was also set to 2.5 mm (FIG. 6). At this time, water molecules contained in the protein system were not considered. In the step of removing the interaction point that interacts with the solvent molecule, for comparison with Example 2, the boundary cell condition (FIG. 7) with a grid of 0.8 cm was used and the solvent contact surface was used. The interaction point was predicted under the above conditions.
FIG. 8 shows the prediction result of the interaction point in the apo-type retinol binding protein. The fine points in the figure are oxygen atoms of water molecules. The number of interaction points obtained was 21 in total, 16 points inside the protein and 5 points in total at 4 points on the protein surface. The 16 interaction points calculated inside the protein are predicted within a limited range, and it can be easily estimated that this part is an interaction site. In addition, there are one or two interaction points calculated on the protein surface, each of which is a candidate interaction point that could not be completely removed by solvent contact surface treatment, solvent molecules, etc. And can be clearly distinguished from the ligand binding site. As the physicochemical information of the interaction points, four weakly positively charged electrostatic term-dominated interaction points were calculated inside the protein and one point on the protein surface. Also, there is one strongly positively charged electrostatic term-dominated interaction point on the protein surface (not shown because it is hidden behind the leftmost interaction point in the overall view), and the rest is Both were interaction points governed by electron correlation terms.
The number of interaction points in each step was 3367 points in the second step, 644 points in the third step, and 21 points in the fourth step.

This protein has a plurality of cavities and irregularities, but when the present invention is used, a plurality of interaction points are predicted intensively at only one site as described above. FIG. 8 also shows, for comparison, retinol molecules obtained by superimposing the same type retinol-binding proteins forming a complex (hereinafter referred to as holo). Some of the interaction sites predicted by the present invention were completely consistent with the fat-soluble portion of the retinol molecule that actually formed the complex.
On the other hand, although the interaction point corresponding to the alcohol part of the retinol molecule is not predicted, as shown in FIG. 8, the interaction point where the alcohol functional group of the retinol molecule should interact with the apo-type retinol-binding protein is shown in FIG. This is because the phenylalanine side chain of the protein itself is occupied. In other words, the interaction point or interaction site involved in ligand recognition brings about an important chemical consideration that the apo type is used for stabilization of the internal structure.
Thus, when the present invention is used, even if there are a plurality of cavities and irregularities in the inside or the surface of the protein, even if it is an apo type that has stabilized inside, It turns out that the complex formation site | part of an external factor can be estimated correctly.

Next, it is shown that the holo type can accurately predict not only the complex formation site of the ligand but also the shape of the ligand molecule and the chemical requirement of the ligand.
The three-dimensional structure of the protein was PDB ID: 1BRP provided by PDB. In the pretreatment of the protein system, the retinol molecule was assigned binding information, hydrogen atom, Amber94 force field atom type, and MNDO / d atomic charge immediately before the optimization of the hydrogen atom structure. Other pretreatments, interaction conditions with constituent atoms of the biopolymer (in addition to water molecules, retinol molecules were not considered) and interaction conditions with solvent molecules were all carried out in the same manner as in Example 1.
FIG. 9 shows the result of predicting the interaction point for the holo-type plasma human retinol binding protein. The fine points in the figure are oxygen atoms of water molecules. The figure also shows retinol molecules forming a complex at the same time for comparison. The number of interaction points obtained was 16 in total, 13 points inside the protein and 3 points in total on the protein surface. Any interaction point calculated on the protein surface is determined to be an interaction point candidate that could not be completely removed by the solvent contact surface treatment, or an interaction point with a solvent molecule or the like. In contrast, the predicted region containing 13 points inside the protein was completely consistent with the ligand binding site. The physicochemical information of the interaction point is that there are three weakly positively charged electrostatic term-dominated interaction points inside the protein and two points on the protein surface, weakly negatively charged electrostatic term-dominated interactions There was one point inside the protein. The rest were all interaction points governed by electron correlation terms.
The number of interaction points in each step was 3367 points in the second step, 631 points in the third step, and 16 points in the fourth step.

It can be seen that the shape presumed by three-dimensionally connecting the predicted interaction point to the ligand binding site agrees fairly well with the retinol molecule forming the complex. Also, in the holo type, the occupation by phenylalanine is eliminated, and the positively charged electrostatic term-dominated interaction point on the alcohol oxygen atom of the retinol molecule, therefore, due to the electrostatic interaction with the alcohol oxygen having a large negative charge of the retinol molecule. A stabilizing interaction point is predicted.
Furthermore, the holo-type phenylalanine has moved to the upper interaction point predicted by the apo-type of FIG. 8, which indicates that the present invention has been applied to the dynamic behavior of the internal structure of the biopolymer, such as the induction fitting effect. It also shows that it can be predicted.
The predicted physicochemical information of the interaction point is dominated by the electron correlation term at the interaction point corresponding to the fat-soluble part of retinol, and positively charged at the interaction point corresponding to the alcohol part as described above. Is accurately predicted to be dominated by the electrostatic term. Also, one weakly positively charged electrostatic term dominated interaction point is correctly predicted on the conjugated olefin surface.

Next, the interaction point prediction of the glucosamine-6-phosphate synthase region of Escherichia coli glutamine amide transferase will be described. This enzyme is characterized by a substrate-binding site in a fairly shallow dent on the protein surface and multiple cavities and irregularities. The present invention shows that the interaction point can be accurately predicted even in such a case.
As the three-dimensional structure of this enzyme, PDB ID: 1MOR provided by PDB was used. The same conditions as in Example 2 were adopted except that a solvent contact surface under a boundary cell condition of 1.0 mm was used as an interaction condition with solvent molecules.
The number of interaction points in each process was 6676 points in the second process, 986 points in the third process, and 21 points in the fourth process.
FIG. 10 shows the result of predicting the interaction point for the glucosamine-6-phosphate synthase region of E. coli glutamine amide transferase. The fine points in the figure are oxygen atoms of water molecules. The figure also shows glucose-6-phosphate molecules forming a complex at the same time for comparison.

There were a total of 21 interaction points, a total of 19 points at 6 locations on the protein surface and 2 points at 2 locations inside the protein. Of the six predicted on the protein surface, four were in full agreement with the coordinates of the crystal water. In addition, the negatively charged electrostatic term-dominated interaction point calculated at the top in the overall diagram of FIG. 10 is an interaction point candidate or solvent molecule that could not be completely removed by the solvent contact surface treatment. The remaining one point where 14 interaction points were predicted was perfectly consistent with the substrate binding site of the glucosamine-6-phosphate synthase region. In addition, although interaction points dominated by the electron correlation term are predicted one by one at two different positions in the protein (one point is almost in the center of the whole figure, and the other point is glucose-6-phosphate. This enzyme has a tunnel mechanism for ammonia molecules to move from a position far away from the site where glucose-6-phosphate is bound. The interaction points inside these proteins may be predicted within this tunnel.
The physicochemical information of each of the 14 interaction points at the glucosamine-6-phosphate binding site is based on the interaction point dominated by the electron correlation term and the weakly positively charged electrostatic term on the pyranose ring of the glucose-6-phosphate molecule. A plurality of dominant interaction points are predicted, and only a plurality of positively charged electrostatic term-dominated interaction points are predicted in the phosphate portion. It accurately predicted the chemical characteristics of acid molecules or known substrates.

Next, the effectiveness of the present invention is demonstrated using another protein, namely, a human retinoic acid receptor ligand binding region (gamma subtype) employing an agonist conformation and a natural ligand, retinoic acid. This receptor is characterized by having a ligand binding site in the cavity inside the protein.
As the three-dimensional structure of the receptor, PDB ID: 2LBD provided by PDB was used. The employed conditions were the same as those in Example 2 except that the interaction distance condition 2.4 Å was used as the interaction point condition with the constituent atoms of the biopolymer.
The number of interaction points in each step was 4420 points in the second step, 765 points in the third step, and 26 points in the fourth step.
FIG. 11 and FIG. 12 simultaneously show the predicted interaction points and retinoic acid molecules forming the complex for comparison, and are a front view (FIG. 11) and a side view (FIG. 12), respectively. The fine points in the figure are oxygen atoms of water molecules. As in Examples 1 to 3, this protein has a plurality of cavities and irregularities, but the predicted interaction site completely matches the actual complex formation site. The total number of interaction points obtained was 26 on a total of 8 points on 6 protein surfaces and 18 on the inside of the protein. Some of the interaction points calculated on the protein surface are considered as interaction point candidates that could not be completely removed by solvent contact surface treatment, or interaction points with solvent molecules. The three points were near the C-terminal and were in the same location as the part where the structure following the C-terminal called the F region was bound in other nuclear receptors. In addition, one other point was predicted at a portion where an activator or a suppressor is bound. As the physicochemical information of the interaction point, there are weakly positively charged electrostatic term-dominated interaction points, one on each protein surface and two on the ligand binding site, weakly negatively charged electrostatic One term-dominated interaction point was predicted for each of the two protein surfaces. The rest were all interaction points governed by electron correlation terms.
The shape estimated by three-dimensionally connecting the predicted interaction points to the ligand binding site and the physicochemical information of each interaction point are extremely useful information for narrowing down natural ligand candidates from a vast number of organic compounds. It turns out that it is. That is, the physicochemical information at each interaction point is predicted to be dominated by the electron correlation term in the fat-soluble part of retinoic acid, and dominated by the positively charged electrostatic term in the carboxylic acid part. By combining with the shape information inferred from the interaction point, even if the natural ligand is unknown, it can be easily inferred from this information that the natural ligand of the receptor is a fatty acid.

Next, an example will be shown in which interaction point information predicted by the present invention can provide effective guidelines for structural modification of a ligand aiming at high affinity or design and search of a novel low molecular weight compound.
FIG. 13 shows a ligand binding site of mouse retinoid X receptor ligand binding region F318A mutant (alpha subtype) adopting antagonist conformation and an oleic acid molecule forming a complex at that site. The fine points in the figure are oxygen atoms of water molecules.
The three-dimensional structure of this receptor was PDB ID: 1DFK provided by PDB. Since 1DFK is a heterodimer with the retinoic acid receptor, water molecules within 4Å from the receptor and molecules other than the oleic acid molecule incorporated into the receptor were separated and used. The receptor helix 1-helix 3 loop was modeled using the coordinates of the main chain of the human retinoid X receptor ligand binding region (alpha subtype) adopting agonist conformation as a template. The interaction distance condition is set to 2.1 mm in consideration of water molecules in determining the interaction with the constituent atoms of the biopolymer, and the van der Waals surface by the boundary cell condition of 1.5 mm grid is used as the interaction condition with the solvent molecules. The interaction point prediction was performed under the same pretreatment and conditions as in Example 2 except that they were used.
The number of interaction points in each process was 4234 in the second process, 767 in the third process, and 25 in the fourth process.

As in the previous examples, the protein has a plurality of cavities and irregularities, but the predicted interaction site completely coincides with the actual complex formation site. The total number of interaction points obtained was a total of 25 points, 4 points at 4 locations on the protein surface and 21 points at 2 locations inside the protein. Any interaction point calculated on the protein surface is determined to be an interaction point candidate that could not be completely removed by van der Waals surface treatment, or an interaction point with a solvent molecule or the like. Of the two interaction sites inside the protein, 19 interaction points were predicted for the ligand binding site. On the other hand, the remaining one part in the protein coincided with a portion to which helix 12, an activator or an inhibitor that adopts the antagonist form is bound. As the physicochemical information of the interaction point, there are two weakly positively charged electrostatic term-dominated interaction points on the protein surface and one point on the ligand binding site, and the weakly negatively charged electrostatic term-dominated interaction. Similarly, there were 2 points on the protein surface and 1 point on the portion of the ligand binding site not occupied by the oleic acid molecule. The rest were all interaction points governed by electron correlation terms.
The shape and physicochemical information inferred from these interaction points is the same as in the previous examples. I predict well. In other words, the 17 interaction points located on the carbon atom of the oleic acid molecule are all governed by the electron correlation term, and only one interaction point located on the carboxylate oxygen atom is a positively charged electrostatic term. It was the interaction point of dominance.

In addition, one weakly negatively charged electrostatic term dominated interaction point that is not occupied by the structure of oleic acid is predicted at the complex formation site.
That is, from any position of oleic acid, if it is possible to introduce a substituent charged weakly positive results in the stabilization in the interaction point, suggesting that can further stabilize the complex.
Also, in the previous holo-type retinol-binding protein (FIG. 9), a weakly positively charged electrostatic term-dominated interaction point that is not occupied by the ligand is predicted at the lower part of the yonon ring of retinol. This suggests the possibility that a retinol derivative oxidized at the corresponding position can be a ligand of the protein.

Finally, an example of predicting the most stable complex structure between the biopolymer and any low molecular weight compound using the predicted interaction point information is shown below.
It is a representative retinoid that has a stable conformation separately for the human retinoic acid receptor ligand binding region (including water molecules) that adopts the agonist conformation used in Example 4 by molecular dynamics calculation. FIG. 14 shows a procedure of one aspect for docking Am80.
Unlike Example 4, water molecules are also considered in the interaction determination with the constituent atoms of the biopolymer, so that a single interaction point on the protein surface is not predicted, and only the actual ligand binding site interacts. The point of action is predicted.
The total number of interaction points predicted is 16, but this time in FIG. 11, the electron correlation term located on the methyl group of the nonone ring of retinoic acid, the olefinic carbon at the 10th position, and the alpha carbon at the 14th position The three dominant interaction points were used as the superposition points on the protein side.
The four twisted bonds of Am80 are rotated in 60 degree increments (81 conformations are generated), allowing the same information assigned to each interaction atom type and each heavy atom constituting Am80. Overlay was performed on combinations (9240 per conformation), and for dockable combinations, Am80 was further structurally optimized and the interaction energy was estimated.
The lower right figure of FIG. 14 shows the most stable complex structure of Am80 obtained by docking simulation, and the lower left figure of FIG. 14 shows the actual complex structure with retinoic acid.
The most stable complex structure of Am80 obtained by docking simulation is found to be very similar to the actual complex structure with retinoic acid, and the interaction point information predicted by the present invention is extremely effective in docking simulation. It was shown that there is.

The conceptual diagram about one aspect | mode of this invention using the human retinoic acid receptor ligand binding area | region which employ | adopts an agonist conformation is shown. The conceptual diagram of fragmentation is shown. The initial structure for a two-body interaction model calculation of alanine, leucine, isoleucine, valine and threonine side chain terminal methyl group is shown. It is one aspect | mode of the method of database-izing the interaction point information about alanine. A method for specifying the positions of all interaction points on an alanine residue based on interaction point information is shown. The interaction point judgment method related to internal structure stabilization is shown. The determination method of the solvent contact surface external interaction point using a grid is shown. The interaction point prediction result in apo-type retinol binding protein is shown. The interaction point prediction result in a holo-type retinol binding protein is shown. The interaction point prediction result in the glucosamine-6-phosphate synthase region of glutamine amide transferase is shown. The front view of the interaction point prediction result in the human retinoic acid receptor ligand binding region which employs agonist conformation is shown. The side view of the interaction point prediction result in the human retinoic acid receptor ligand binding region which employs agonist conformation is shown. The interaction point prediction result in the mouse | mouth retinoid X receptor ligand binding area | region F318A mutant which employ | adopted antagonist conformation is shown. The conceptual diagram about the one aspect | mode which implements the docking simulation between the human retinoic acid receptor ligand binding region and Am80 which employ | adopt an agonist conformation using the interaction point information estimated by this invention is shown.

Claims (10)

  Interaction sites in biopolymersBy computerA prediction method,
Atomic coordinate information of biopolymersReadFirst step,
In the second step of identifying the coordinates of the interaction point candidate of the biopolymer based on the interaction point information database, the interaction point information database isThe position of the interaction point with respect to each constituent unit of the biopolymer is defined by a relative coordinate system with respect to the constituent unit.Including location information,For the structural unit of the biopolymer, access to the database to obtain the position information of the corresponding interaction point,The informationAnd saidFor atomic coordinate information of structural units of biopolymersOn the basis ofThe step of identifying the location of the interaction point candidate;
  Of the interaction point candidates identified in the second step, the third step of removing the interaction point candidates involved in the interaction with the constituent atoms of the biopolymer,
Including the above method. The computer, in the third step, for each interaction point candidate, when the constituent atoms of the biopolymer within 4Å from the interaction point candidate exists, you remove the interaction point candidate, claim The method according to 1. The computer further includes, after the third step,
A fourth step of removing candidate interaction points involved in the interaction between the surface of the biopolymer and the solvent molecule;
The execution method according to claim 1 or 2. The computer, in the fourth step, remove interaction point candidate present outside the solvent accessible surface or van der Waals surface of the biopolymer The method of claim 3. The method according to claim 1 , wherein the computer further executes a fifth step of displaying the predicted position of the interaction point on a display. In the fifth step , the computer causes the interaction point to be displayed in different colors so as to distinguish whether the interaction energy at the interaction point is dominated by an electronic correlation term or an electrostatic term. Item 6. The method according to Item 5.   The method according to any one of claims 1 to 6, wherein the biopolymer is a protein. A computer program for predicting an interaction site in a biopolymer,
First step to read atomic coordinate information of biopolymer,
A second procedure for specifying coordinates of the interaction point candidate of the biopolymer with reference to an interaction point information database, wherein the interaction point information database is an interaction for each structural unit of the biopolymer. Including position information defining a position of a point by a relative coordinate system with respect to the structural unit, and for the structural unit of the biopolymer, accessing the database to obtain position information of a corresponding interaction point; second instructions for identifying the location of the interaction point candidates based on the information and the atomic coordinate information of the structural unit of the biopolymer,
For the interaction point candidates identified in the second step,
Determining whether or not a constituent atom of the biopolymer exists within a predetermined distance from the interaction point candidate, and when present within the predetermined distance, a third procedure for removing the interaction point candidate;
A computer program for causing a computer to execute a procedure for displaying a predicted position of an interaction point on a display. A computer program for predicting an interaction site in a biopolymer,
First step to read atomic coordinate information of biopolymer,
A second procedure for specifying coordinates of the interaction point candidate of the biopolymer with reference to an interaction point information database, wherein the interaction point information database is an interaction for each structural unit of the biopolymer. the position of the point includes position information defining the relative coordinate system with respect to the structural unit, the second for identifying the position of the interaction point candidates based on the information and the atomic coordinate information of the structural unit of the biopolymer procedure,
For the interaction point candidates identified in the second step,
Determining whether or not a constituent atom of the biopolymer exists within a predetermined distance from the interaction point candidate, and when present within the predetermined distance, a third procedure for removing the interaction point candidate;
A fourth procedure for removing candidate interaction points existing outside the solvent contact surface or van der Waals surface of the biopolymer,
A computer program for causing a computer to execute a procedure for displaying a predicted position of an interaction point on a display.   The computer program according to claim 8 or 9, wherein the biopolymer is a protein.