JP4377691B2 - Screening method, screening system and screening program - Google Patents

Description

Technical field
    The present invention relates to a screening method, a screening system, and a screening program for performing virtual screening from a virtual library in drug discovery.
Background art
    Most pharmaceuticals currently in use have been developed by optimizing lead compounds. The lead compounds used here are mainly those in which plants, microorganisms, natural products derived from higher organisms, etc., including hormones and transmitters, have been discovered by screening techniques.
    In recent years, in the molecular design of these lead compound drug discovery, virtual screening technology predicted on a computer has become an important item in a series of drug discovery screening studies. In addition to the dramatic improvement in computer performance, the amount of structure-activity relationship information accumulated by high-throughput screening and combinatorial synthesis technology has increased dramatically. It is mentioned that the structural information of the number has increased dramatically.
    The molecular design technology that realizes the virtual screening is roughly classified into Ligand Based Drug Design (hereinafter referred to as LBDD) based on the known structure-activity relationship information and Structure-Base Drug Design (hereinafter referred to as LBDD) using the 3D structure information of the target protein. 2) (referred to as SBDD).
    In particular, in the technique called SBDD, when a plurality of drugs bind to the vicinity of the active site of the target protein, each drug has a complementary relationship with the protein, and the amount of change in free energy in each binding process is pharmacological. This is based on the knowledge that the level of activity is expressed. The target protein and ligand binding state and its pharmacological activity value are estimated on a computer. Predictive information on the structure-activity relationship is not required, but highly accurate activity value prediction can be expected. Has the advantage.
    On the other hand, LBDD, which is a drug discovery technology focusing on the fact that physicochemical parameters are homologous between drugs bound to a common site (such as overlay analysis and quantitative structure-activity relationship (QSAR)) Although it is an approximate solution, it has a high correct answer rate for various target systems, and has an advantage that a receptor structure (binding model) is not required unlike SBDD. (Reference: C.G.Wermuth, "From the latest drug discovery chemistry, from exploratory research to development," Technomic Co., Ltd., published on September 25, 1999, Chapter 11, Electric screening, Chapter 19, Quantitative structure-activity relationship methods)
    However, in SBDD, if it works well, an almost true solution can be obtained, but there are many target systems with extremely low correct answer rates. In addition, a receptor structure (binding model) is essential, and the prediction accuracy depends on the accuracy of the structure. Moreover, even if the structure-activity correlation information is accumulated, there is also an aspect that the prediction accuracy is not improved. For this reason, as described above, the general level of SBDD technology has not yet been applied to specific drug discovery research, and the mainstream of virtual screening is LBDD.
    Although LBDD has achieved results in actual drug discovery research, there is a limit to the improvement of prediction accuracy. One of the reasons is that it requires foresight information on pharmacological activity, and the prediction accuracy depends on the quality and quantity of the foresight information. The accuracy is low, and the accuracy is not satisfactory.
Disclosure of the invention
    The present invention has been made in view of the above points, and its object is to provide a screening method, a screening system, and a screening program having extremely high prediction accuracy and versatility showing a high correct answer rate for various target systems. Is to provide.
    That is, the screening method of the present invention is a programmed computer screening method including the following configuration:
  Based on structural information obtained in advance for a compound that is expected to have pharmacological activity against the target protein, a plurality of different conformations of the constituents of the compound are obtained, and a superposition analysis of these conformations is performed to obtain a plurality of conformations. Calculate the overlay candidate model,
  Calculate a plurality of binding models between the plurality of overlay candidate models and the target protein
  Calculate the binding free energy and pharmacological activity value for the multiple binding models obtained to obtain the correlation coefficient between them.And correcting a plurality of combined models based on the correlation coefficient
Next, a virtual screening is performed using the aggregate analysis result obtained by performing a statistical analysis with the physicochemical parameter as an element for the above binding model,
After this, a virtual screening using the pharmacophore map created by specifying the binding mode as an evaluation index, and the binding free energy function obtained by optimizing the binding free energy for the target protein based on the known structure-activity relationship information And virtual screening using as an evaluation index
    Moreover, the screening method of the present invention includes the following configuration in the screening method according to claim 1:
  Total analysis results obtained by performing statistical analysis with the physicochemical parameters as elements for the binding model,
  A pharmacophore map created by specifying the combination styleWhen,
  Optimize the binding free energy for the target protein based on the known structure-activity relationship informationObtainedBond free energy function and
As an evaluation index for virtual screening,theseThe evaluation index is updated based on the screening result.
  A screening system according to the present invention is a screening system having the following configuration:
  Stores the structural information obtained in advance for the target protein and the compound expected to have pharmacological activity against the target protein, and the conformation, superposition model, and binding model calculated based on the structural information of the compound. Based on the database part and the structure information of the compound stored in the database part, a plurality of different conformations of the constituents of the compound are obtained, and a superposition analysis of these conformations is performed to calculate a plurality of superposition candidate models. A combined analysis unit, a binding unit that calculates a plurality of binding models of the plurality of overlay candidate models and the target protein, and a binding free energy and a pharmacological activity value for the obtained binding models. The correlation coefficient between the twoA plurality of combined models are corrected based on the obtained correlation coefficientCalculation unit;
The calculation unit includes a total analysis result obtained by performing a statistical analysis using the physicochemical parameter as an element for the bond model, a pharmacophore map created by specifying the bond mode, and known structure-activity relationship information. Based on the above, the binding free energy function obtained by optimizing the binding free energy for the target protein is used as an evaluation index for virtual screening..
  At this time, the calculation unitThe above evaluation index is updated based on the screening results.If so, more preferable results can be obtained.
  The screening program according to the present invention is a screening program for causing a computer to execute the following steps.
Based on structural information obtained in advance for a compound that is expected to have pharmacological activity against the target protein, a plurality of different conformations of the constituents of the compound are obtained, and a superposition analysis of these conformations is performed to obtain a plurality of conformations. Calculating an overlay candidate model;
  Calculating a plurality of binding models of the plurality of overlay candidate models and the target protein,
  Calculate the binding free energy and pharmacological activity value for the multiple binding models obtained and calculate the correlation coefficient between them.Find and modify multiple combined models based on the correlation coefficientStep,
  the aboveA step of performing virtual screening using the aggregate analysis result obtained by performing a statistical analysis of the binding model as an element of a physicochemical parameter;
  A pharmacophore created with a specific binding stylePerforming virtual screening using the map as an evaluation index,
  Optimize the binding free energy for the target protein based on the known structure-activity relationship informationObtainedA step of performing virtual screening using the binding free energy function as an evaluation index.
    Further features of the present invention and the effects it provides will be understood in more detail based on the best mode for carrying out the invention described below with reference to the accompanying drawings.
BEST MODE FOR CARRYING OUT THE INVENTION
Example
    With reference to FIGS. 1-7, the screening system based on the Example of this invention is demonstrated in detail below.
    FIG. 2 shows a block diagram of a screening system realized by a computer and software (program) executed on the computer. It comprises a database unit 1, a conformation analysis unit 2, a superposition analysis unit 3, a molecular simulation unit 5, a calculation unit 6 for calculation of binding free energy and parameter optimization, an analysis search unit 7, and the like.
  The database unit 1 stores a structural formula, a conformation, a physicochemical parameter, and the like of a target protein, which will be described later, and a compound expected to have pharmacological activity for the target protein in a storage device in a computer. Conformational analysis unit 2 performs molecular simulation using force field parameters such as MM2, MM3, MMFF, AMBER, CHARMm, and OPLS to search for the most stable and metastable conformations (in this case, GB-SA or S-GB The superposition analysis unit 3 performs an exhaustive search using the homology between compounds in any combination of physicochemical parameters as an index. This is a list of superposition model candidates. The statistical analysis unit 4 includes statistics such as linear regression analysis, principal component analysis, CART method, multiple regression analysis, principal component analysis, K-Nearest Neighbors method, neural network, discriminant analysis, decision tree, support vector machine, and Bayesian network. The molecular simulation unit 5 performs molecular simulation by systematic search, molecular mechanics method, Monte Carlo method, molecular dynamics method, molecular orbital method, etc., and the calculation unit 6 calculates the binding free energy. And parameter optimization, more specifically, calculation of binding free energy from the difference in free energy between bound and unbound states, and assignment of contributions such as van der Waals potential and electrostatic potential from the target protein to each grid point Bind when given coordinate data for any ligand The free energy is calculated at a high speed, and the analysis search unit 7 performs a search based on the degree of matching with the statistical analysis result and the homology index.
    Note that these are functionally separated, and in fact, in addition to the database unit, a conformation analysis unit that generates conformations, a physicochemical parameter calculation unit that generates physicochemical parameters, superposition and Physicochemical parameter homology analysis unit, statistical analysis unit for homology search, etc.Such as creating a pharmacophore mapA program is constructed by modules such as a molecular simulation unit. In addition, these individual modules already exist as commercially available application software, but in the present invention, by using a program described later, an advanced prediction function that has not been obtained so far is achieved by organically linking them. Is.
  Moreover, the function of each part does not need to be independent. For example, it may be a combination of parts having some functions to complete the function of each part, or a combination of several parts having all the functions of each part. Furthermore, in the present invention, it is also possible to use modules having a part of functions in an organic manner in cooperation.
    In the present invention, the binding state between a specific target protein and a drug is basically searched by combining SBDD and LBDD. For each compound to be evaluated, “complementarity with target protein” and “known” "Homology with active compounds" is considered at the same time, and multiple drugs are considered at the same time, and "Homology of physicochemical parameters" and "Correlation between binding free energy and pharmacological activity" are simultaneously satisfied Find a solution. In obtaining this solution, here, as shown in FIG. 1, screening is performed in the steps of initial connection model construction, connection model correction, evaluation index optimization, virtual screening, and screening experiment.
    First, the initial binding model is created. Including those with relatively weak pharmacological activity, many active compounds are often obtained at the initial stage of research. Build a combined model. At this time, the estimation of the binding model usually evaluates whether or not the compound of interest is complementary to the target protein. Here, a superposition analysis of a plurality of molecules is performed in advance, and a candidate for the superposition model is determined. After enumeration, multiple molecule binding models are estimated simultaneously.
    That is, as shown in FIG. 4, conformations A1, A2, A3, B1, B2, B3, C1, C2, C3... Based on known active compounds A, B, C. Are first subjected to overlay analysis to create a plurality of overlay candidate models A1B3C3, A3B1C1,..., And then, from among the compounds used for overlay, the most rigid and large compound is selected, and the overlay candidate model is selected. Create a model that can bind to the target protein without any contradiction using the conformation. If the binding position and orientation of molecules are determined, the position and orientation of other molecules are uniquely determined based on the superposition model. Then, another compound is arranged on the obtained bond model based on the superposition model, and the free energy of the complex is minimized on an atomic coordinate basis by, for example, the simplex method.
    The binding model of a plurality of compounds obtained in this way is corrected. When a binding model of multiple active compounds is constructed using the correlation between binding free energy and pharmacological activity value as an index, it is estimated that the binding free energy calculated from the binding model and the pharmacological activity value have a linear relationship. For this reason, a model far from the true coupled state deviates greatly from this linear relationship, so such a model is removed from the coupled model (see FIG. 6). Identification of an unfavorable bond model is easy, and noise for subsequent prediction can be reduced.
    If we can estimate the binding state of multiple molecules with high reliability by searching the above initial binding model,
  Statistical analysis using physicochemical parameters that do not depend on the coordinate system as explanatory variables using the presence or absence of binding ability (for example,Pharmacophore(For example, the CART (Classification and Regression Tree) method with emphasis on statistical reliability can be suitably used as a statistical method using components and various structural descriptors)
  (B) Statistical analysis using the presence or absence of binding ability as an explanatory variable with physicochemical parameters that depend on the coordinate system.PharmacophoreUsing descriptors that assign components to grids, statistical methods such as simple aggregation and setting weighted homology scores based on the aggregation, discriminant analysis, decision trees, support vector machines, Bayesian networks, etc. Can be suitably used)
  C Optimization of binding free energy function specialized for target protein R based on known structure-activity relationship information
Perform 3 points.
    Physicochemical parameters include pharmacophore descriptor, WHIM descriptor, substituent length, substituent width, molecular refraction MR, Hammet substituent constant, Swain-Lupton electron effect parameters, dissociation constant, partial Electronic charge, hydrophobic constant of Hansch, hydrophobic constant of substituent, partition coefficient logP, hydrophobicity index measured by HPLC, calculated logP CLOGP, number of hydrogen bond acceptors, number of hydrogen bond donor groups, possible The total number of hydrogen bonds, dummy variables indicating whether hydrogen bonds are possible, and the like can be used.
    The above three evaluation indices are used for virtual screening. As shown in FIG. 5, the statistical model (1) obtained from (a) is first used for the purpose of quickly identifying the presence or absence of activity against the target protein. Screening based on the QSAR model (a) in FIG. 3 is performed. In this step, which can be positioned as a pre-screening for the subsequent detailed search, the LBDD technique (CART analysis using various structure descriptors) that enables screening independent of the coordinate system can be used as it is. In the statistical model, the evaluation is made only with the elements of the physicochemical parameters possessed by each candidate compound, so that high-speed processing is possible. In addition, when the CART method is applied as the LBDD technique, the analysis is generally performed while the active conformation and the overlap between the active compounds are unknown, but here, between the active conformation and the active compound with high reliability. Therefore, it is estimated that a more accurate prediction model can be obtained.
    By the way, for a group of compounds with known inhibitory activity against M-Calpain, 207 compounds (compound groups having an IC50 value of less than 10 μM) found from high-throughput screening of our own compound library and 10500 arbitrarily extracted from a commercially available compound library Using the compound as a target, using the coordinate data (Protein Data Bank registration: 1DF0) obtained from the X-ray crystal structure analysis of M-Calpain, a docking study of each compound was performed, and the top 10 items from the highest score It is possible to obtain binding models and perform analysis by the CART method (see Japanese Patent Application Laid-Open No. 2002-073626) using structural descriptors that are generally used using the coordinates of these binding models to identify the presence or absence of inhibitory activity. A structure-activity relationship model was obtained. The model was created using a partly extracted data set (150 active compounds and 10,000 inactive compounds). And as a result of determining how accurately the resulting model can predict the remaining data sets (57 active compounds and 500 inactive compounds), 49 active compounds (86%) and inactive compounds 446 (89%) of the group were correctly identified.
  On the other hand, instead of performing a docking study, using the coordinates of about 1 to several hundreds of stable and metastable conformational coordinates obtained from the conformational analysis in the solution, the analysis by the CART method is also performed to identify the presence or absence of inhibitory activity As a result of obtaining a structure-activity relationship model that can be obtained, only 44 cases (77%) of the active compound group and 406 cases (81%) of the inactive compound group were correctly identified.
    In the above screening process, a candidate compound that is estimated to have binding ability is a statistical analysis using the presence or absence of binding ability as an index and physicochemical parameters depending on the coordinate system as explanatory variables described in the above (b). A) or b).
  PharmacophoreIn the case of specifying the binding mode based on the map, at this stage, the preferred binding position and orientation are roughly specified using the degree of fitness of the known active compound group to the binding model as an index. However, since it is necessary to cover the binding position and orientation of the compound to be evaluated in the vicinity of the active site of the target protein, a certain amount of calculation is required. It is preferable to do this. The evaluation index is the total number of compounds for each lattice point for the presence of binding factors and atoms.PharmacophoreUse “Map”. The compound to be evaluated is translated and rotated in this grid,PharmacophoreThe position where the similarity with the map is optimal is specified. In the relative positional relationship between the compound to be evaluated and the target protein thus determined, if possible, water molecules are arranged between the two. this is,PharmacophoreIt will be arranged according to the hydrogen bond factor on the map.
    Since the bond model is obtained by this operation, the free energy of the complex is minimized on the basis of atomic coordinates by using, for example, the simplex method with the initial model as the initial coordinate, and the optimized bond free energy ΔG of (c) is calculated ( Perform c) in FIG. In this case, the bond free energy is calculated after optimizing the bond mode on the basis of atomic coordinates. For the optimization of coordinates, the polymer side is fixed and the energy is minimized. Examples of the bond free energy evaluation function include:
  dGbind = A * dGinter + B * dGsolv + C * dGconf + D
(DGinter: potential function, dGsolv: GB-SA or S-GB, dGconf: empirical function)
A linear combination form of a plurality of energy terms is adopted, and parameters optimized for the target protein of interest and the group of compounds to be searched are used.
    Some of the compounds judged to be promising in the virtual screening are actually subjected to pharmacological evaluation, and based on the obtained pharmacological activity information, the binding model is corrected and the three indicators used for the virtual screening are updated. That is, compounds that were predicted to be active but were actually inactive are removed from the binding model. The increase in reliable binding models and pharmacological activity data can improve the accuracy of each indicator.
    When the complex structure of a specific compound is revealed by X-ray or NMR, the binding model is corrected by replacing the corresponding compound and target protein in the model with actual measurement values, and the complex free energy of each compound is determined in the same manner as before. Minimize. At this time, if the predicted value of the pharmacological activity value and the actually measured value are far from each other, the overlay candidate model is extracted again, and the optimum solution is searched using the actually measured binding coordinates. Also at this time, the above three indexes are updated.
    By the way, in drug discovery research, there are a case where a peripheral substituent of a specific skeleton is searched and a case where a skeleton itself is searched. According to the above technique, the interaction between a compound and a protein is revealed. Because it is evaluated, it can be applied with versatility whatever the focus of the search is. In addition, various experimental data can be captured flexibly, and the accuracy of prediction can be improved sequentially by this capture. In addition, index optimization and virtual screening specialized for a specific group of compounds can be performed. Can respond.
    Furthermore, it is also possible to construct a binding model with a related protein for which a three-dimensional structure is not obtained from a binding model of a plurality of compounds constructed for a specific protein. The three-dimensional structure of the related protein is constructed by homology modeling and replaced with the protein of the previous binding model. Since a compound for which a binding model has been obtained already exists in its own library, the pharmacological activity against this related protein can be easily evaluated. Based on the obtained structure-activity relationship information, using the correlation between free binding energy and pharmacological activity as an evaluation function, we can optimize the vicinity of the active site of a related protein, and provide sufficient coordinates for subsequent virtual screening. Data can be obtained.
    In addition, it is also effective in considering compounds that bind through water molecules. In other words, compounds that bind to a specific protein are known to interact directly with the protein and to bind through several water molecules, and the latter compound and its binding mode can be simplified. A method for fast estimation is not known at present. However, according to the present invention, a compound (α in FIG. 3) that deviates greatly from the correlation between the free binding energy (ΔG) and the pharmacological activity value (Activity) in a plurality of compounds is bound via a water molecule. These compounds can be inferred thatPharmacophoreBased on the map, water molecules are arranged, and the free energy of the complex is calculated again by optimizing the complex free energy on the coordinate basis. In the virtual screening, the molecule of interest, target protein andPharmacophoreWhether or not the water molecule can be arranged can be instantaneously determined from the map correspondence.
    Since the units 2 to 6 shown in FIG. 2 can be parallelized, a significant speed improvement can be achieved by systematization on the premise of parallelization. In addition, the above-described filtering hierarchization can reduce the accuracy to a realistic calculation time. FIG. 3 schematically shows the flow of screening by the system according to the present invention. The arrows indicate the flow of structure data or activity data, and the open arrows indicate the flow of compound data.
    For screening, various statistical analyzes such as linear regression analysis and CART method are performed, and a database search based on the obtained statistical model, conformational analysis and calculation of binding free energy including water molecules are possible. Possible programs, programs that perform overlay analysis of multiple molecules and database search based on the results, programs that perform database search based on statistical models obtained by physicochemical parameter calculation and specific statistical analysis As described above, there are programs for estimating the bonding mode and strength of polymers in the polymer and programs for calculating the binding free energy and optimizing the functions. In addition, the method of the present invention is executed.
    That is, when coordinate data is given for an arbitrary molecule or a plurality of molecules, a new grid that includes them or a program that automatically corrects an existing grid,
A program for assigning information such as a binding factor of a molecule to each lattice point in the lid, a program for counting the above-mentioned assigned information for a plurality of molecules, and the above-mentioned assigned information with the above-mentioned total information for a plurality of molecules or a specific molecule In addition to these programs, a program that identifies the position, orientation, and conformation at which the above homology index is good for any molecule or molecules, and coordinate data for any protein Is assigned to each lattice point in the grid, the program for evaluating the interaction strength with any molecule, and the above interaction strength is good for any molecule or molecules A program that identifies the position, orientation, and conformation, and the homology index for any molecule or molecules Use intensity is programmed to identify the conformation and position and orientation to be the same time good.
    The above-mentioned conformational analysis, physicochemical parameter calculation, overlay analysis, grid-based high-speed docking study, bond free energy calculation, statistical analysis and function optimization By establishing programs that can be exchanged at any time, these programs
  Based on structural information obtained in advance for a compound that is expected to have pharmacological activity against the target protein, a plurality of different conformations of the constituents of the compound are obtained, and a superposition analysis of these conformations is performed to obtain a plurality of conformations. Calculating an overlay candidate model;
  Calculating a plurality of binding models of the plurality of overlay candidate models and the target protein;
  Calculate the binding free energy and pharmacological activity value for the multiple binding models obtained to obtain the correlation coefficient between them.And correcting a plurality of combined models based on the correlation coefficient;
  Performing virtual screening using the aggregate analysis result obtained by performing statistical analysis on the binding model as an element of physicochemical parameters;
  Created by specifying the binding stylePharmacophorePerforming virtual screening using the map as an evaluation index;
  Optimize the binding free energy for the target protein based on the known structure-activity relationship informationObtainedPerforming virtual screening using the binding free energy function as an evaluation index;
Can be executed on a computer.
Industrial applicability
    According to the present invention, the complementarity between the compound and the target protein and the homology of the binding ability characteristics among a plurality of compounds are considered at the same time, and the correlation between the measured pharmacological activity and the predicted value in the plurality of compounds is also considered. For this reason, it is possible to achieve a significant improvement in prediction accuracy, and it is possible to use optimally according to the purpose, such as a skeleton search and a peripheral substituent search. Moreover, even if there is no three-dimensional structure information of the target protein, virtual screening at a practical level is possible from known information on drugs and structures of related proteins.
    Furthermore, it is possible to estimate a compound that binds via a water molecule and to incorporate it into the subsequent prediction.
    Moreover, the evaluation method of each hierarchy can aim at the improvement of precision sequentially by taking in the data (pharmacological activity, structure, etc.) accumulated with the progress of research.
[Brief description of the drawings]
  FIG. 1 is a flowchart of an example of an embodiment of the present invention.
  FIG. 2 is a block diagram of the above.
  FIG. 3 is an explanatory diagram of the above.
  FIG. 4 is an explanatory diagram of the construction of the initial coupling model.
  FIG. 5 is an explanatory diagram of the application of the above index.
  FIG. 6 is a flowchart showing the detailed operation of the above.
  FIG. 7 is a flowchart showing the detailed operation of the above.

Claims (5)

A programmed computer screening method comprising:
Based on the structural information obtained in advance for the compound that is expected to have pharmacological activity against the target protein, a plurality of different conformations of the constituents of the compound are obtained, and a superposition analysis of these conformations is performed to obtain a plurality of conformations. Calculate the overlay candidate model,
Calculate a plurality of binding models between the plurality of overlay candidate models and the target protein, and calculate a binding free energy and a pharmacological activity value for the obtained plurality of binding models to obtain a correlation coefficient between them , A plurality of coupled models are corrected based on the correlation coefficient
Next, a virtual screening is performed using the aggregate analysis result obtained by performing a statistical analysis with the physicochemical parameter as an element for the above binding model,
After this, virtual screening using the pharmacophore map created by specifying the binding mode as an evaluation index, and the binding free energy function obtained by optimizing the binding free energy for the target protein based on the known structure-activity relationship information And virtual screening using as an evaluation index . The screening method according to claim 1 , wherein
Total analysis results obtained by performing statistical analysis with the physicochemical parameters as elements for the binding model,
A pharmacophore map created by specifying a combination style,
The binding free energy function obtained by optimizing the binding free energy for the target protein based on the known structure-activity relationship information
Is used as an evaluation index for virtual screening, and these evaluation indexes are updated based on the screening results . A screening system with the following configuration:
Stores the structure information obtained in advance for the target protein and the compound expected to have pharmacological activity against the target protein, and the conformation, superposition model, and binding model calculated based on the structure information of the compound. Database section;
A superposition analysis unit that obtains a plurality of different conformations of the constituents of the compound based on the structural information of the compound stored in the database unit and calculates a plurality of superposition candidate models by performing superposition analysis of these conformations;
A binding unit that calculates a plurality of binding models of the plurality of overlay candidate models and the target protein;
A calculation unit that calculates the binding free energy and the pharmacological activity value for the obtained plurality of binding models, obtains a correlation coefficient between them, and corrects the plurality of binding models based on the correlation coefficient;
The arithmetic unit is a total analysis result obtained by performing a statistical analysis with the physicochemical parameter as an element for the binding model,
A pharmacophore map created by specifying a combination style,
Virtual screening is performed using the binding free energy function obtained by optimizing the binding free energy for the target protein based on the known structure-activity relationship information as an evaluation index . The screening system according to claim 3 ,
The arithmetic unit updates the evaluation index based on the screening result . A screening program that causes a computer to perform the following steps:
Based on the structural information obtained in advance for the compound that is expected to have pharmacological activity against the target protein, a plurality of different conformations of the constituents of the compound are obtained, and a superposition analysis of these conformations is performed to obtain a plurality of conformations. Calculating an overlay candidate model;
Calculating a plurality of binding models of the plurality of overlay candidate models and the target protein,
Calculating a binding free energy and a pharmacological activity value for the obtained plurality of binding models to obtain a correlation coefficient between the two, and correcting the plurality of binding models based on the correlation coefficient;
A step of performing virtual screening using the aggregate analysis result obtained by performing a statistical analysis with the physicochemical parameter as an element for the binding model,
A step of performing virtual screening using a pharmacophore map created by specifying a binding mode as an evaluation index;
Performing virtual screening using the binding free energy function obtained by optimizing the binding free energy for the target protein based on the known structure-activity relationship information as an evaluation index .

Images