De novo drug design method and apparatus based on photonic quantum computer
Patent Information
- Application Number
- PCT/CN2025/075023
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2024-03-07
- Filing Date
- 2025-01-25
- Publication Date
- 2025-10-02
Smart Images
Figure CN2025075023_02102025_PF_FP_ABST
Abstract
Description
A de novo drug design method and device based on optical quantum computer Technical Field
[0001] The present application relates to the field of quantum computing technology, and in particular to a method and device for de novo drug design based on an optical quantum computer. Background Art
[0002] Traditional drug screening is an expensive and resource-intensive process, typically costing billions of dollars, with a success rate of only about 10%. In recent years, with the development of powerful molecular modeling tools and the increasing number of resolved structures of protein-small molecule complexes, computer-aided drug design has become an indispensable tool in new drug development.
[0003] For example, existing de novo drug design methods aim to generate novel molecules with specific properties through molecular generation to explore chemical space and replenish compound libraries. This method uses molecular generation to generate and optimize individual molecules, ultimately adapting them to the target pocket.
[0004] De novo drug design focuses on designing and synthesizing drugs from fragment libraries to meet target pockets. It aims to optimize the affinity between proteins and ligands and enhance the various properties of small molecule drugs. It is a crucial step in both the early stages of drug screening and the later stages of drug optimization.
[0005] In existing technologies, there are two main approaches to ab initio drug design based on traditional computers: structure-based design and ligand-based design. The three-dimensional structure of a receptor can usually be determined through X-ray crystallography, nuclear magnetic resonance, or electron microscopy. When the receptor structure is unknown, homology modeling can be used to obtain a suitable structure for new drug design. However, the quality of the homology model depends on the quality of the template structure and sequence similarity. When structural data for the biological target are unavailable, ligand-based approaches are often used.
[0006] In the existing technology, most traditional model screening methods in the pharmaceutical field use heuristic algorithms and systematic search algorithms. However, these algorithms generally have the following disadvantages:
[0007] (1) Use heuristic algorithms: It takes a lot of time to iterate, is time-consuming and computationally intensive, and may not yield the optimal solution. This results in a high rate of false positives during drug development.
[0008] (2) Use systematic search algorithms: This may lead to combinatorial explosion problems, and traditional computers are less capable of handling such problems.
[0009] Therefore, traditional models in the existing technology require a large amount of sampling to obtain a relatively good ligand molecule, which consumes a lot of time and computing power, and the ligand structure may not be the global optimal solution. Summary of the Invention
[0010] In view of this, the present invention provides a method and apparatus for de novo drug design based on an optical quantum computer, thereby enabling rapid and accurate de novo drug design.
[0011] The technical solution of the present invention is specifically achieved as follows:
[0012] A de novo drug design method based on an optical quantum computer, the method comprising:
[0013] Obtain the ligand-receptor complex structure that needs to be optimized;
[0014] Extracting ligand molecules and receptor structures from the ligand-receptor complex structure, marking the optimizable parts of the ligand molecules, and pre-setting a fragment library for replacement;
[0015] Determine the properties to be optimized based on the marked optimizable parts; calculate the corresponding linear coefficients and quadratic coefficients based on the properties to be optimized and the fragment library
[0016] Construct a corresponding mathematical model based on the calculated linear term coefficient and quadratic term coefficient;
[0017] The constructed mathematical model is converted into the corresponding Ising model and input into an optical quantum computer for solution, and the optimized structure and ligand information are calculated.
[0018] Preferably, the method further comprises:
[0019] Output the optimized structure and ligand information.
[0020] Preferably, a fragment-based chemical library is pre-set by mapping relevant drug groups to receptor pockets, serving as a fragment library for replacement.
[0021] Preferably, the properties to be optimized are binding force and properties of the ligand itself.
[0022] Preferably, the corresponding linear coefficients and quadratic coefficients calculated based on the properties to be optimized and the fragment library include:
[0023] After the fragments are replaced according to the fragment library, the properties to be optimized after the fragment replacement are calculated, and the calculated properties are used as the corresponding linear term coefficients or quadratic term coefficients.
[0024] Preferably, when the property to be optimized is binding strength, after replacing a single fragment according to the fragment library, the binding strength after the single fragment replacement is calculated, and the calculated binding strength is used as the linear coefficient;
[0025] When the property to be optimized is the property of the ligand itself, after fragment replacement according to the fragment library, the property of the ligand itself after fragment replacement is calculated, and the calculated property of the ligand itself is used as the quadratic term coefficient.
[0026] Preferably, the mathematical model is a quadratic unconstrained binary optimization model;
[0027] In the quadratic unconstrained binary optimization model, the calculated binding force is used as the linear coefficient, and the ligand's own properties are used as the quadratic coefficient.
[0028] Preferably, the quadratic unconstrained binary optimization model is expressed by the following formula:
[0029] Among them, QUBO is a quadratic unconstrained binary optimization model, n represents the overall fragment length of the optimized ligand, i and j represent the fragment positions that can be optimized, and x i and x j Represents a decision variable, with a value of {0,1}, o i is the linear coefficient, which represents the properties of the ligand itself; t i,j is the coefficient of the quadratic term, which represents the molecular properties after fragment replacement.
[0030] Preferably, the optical quantum computer is a coherent Ising quantum computer.
[0031] The present invention also provides a de novo drug design device based on an optical quantum computer, the device comprising: a marker, a fragment library, a model building converter and an optical quantum computer;
[0032] The marker is used to obtain the ligand-receptor complex structure that needs to be optimized; extract the ligand molecule and the receptor structure from the ligand-receptor complex structure, and mark the optimizable part of the ligand molecule;
[0033] The fragment library is a fragment-based chemical library used for fragment replacement;
[0034] The model construction converter is used to determine the attribute to be optimized based on the marked optimizable portion; calculate the corresponding linear term coefficient and quadratic term coefficient based on the attribute to be optimized and the fragment library; construct a corresponding mathematical model based on the calculated linear term coefficient and quadratic term coefficient, and convert the constructed mathematical model into a corresponding Ising model;
[0035] The optical quantum computer is used to solve the Ising model and calculate the optimized structure and ligand information.
[0036] As can be seen from the above, in the de novo drug design method and device based on optical quantum computers in the present invention, a large-scale fragment-based chemical library can be pre-set by mapping the relevant drug groups to the receptor pocket. As a fragment library for replacement, a large-scale fragment-based chemical library can be pre-set by mapping the relevant drug groups to the receptor pocket. As a fragment library for replacement, the process of synthesizing new molecules based on the fragment library is then converted into a quadratic unconstrained binary optimization problem, thereby converting the de novo drug design problem, which is very important in the drug screening process, into a corresponding mathematical model, and then using a quantum computer to solve the mathematical model, so that the improvement of the affinity of small molecule drugs and the optimization of their properties can be achieved through the mathematical model. Therefore, by using the technical solution of the present invention, the affinity of small molecule drugs and proteins can be quickly optimized, and their various properties can be improved, helping researchers to screen out potential lead compounds and assisting the drug development process. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] FIG1 is a flow chart of a de novo drug design method based on an optical quantum computer in a specific embodiment of the present invention.
[0038] FIG2 is a schematic diagram of the structure of the composite according to a specific embodiment of the present invention.
[0039] FIG3 is a schematic diagram showing the parts of the ligand molecule and the groups that need to be optimized in a specific embodiment of the present invention.
[0040] FIG4 is a schematic structural diagram of a de novo drug design device based on an optical quantum computer in a specific embodiment of the present invention. DETAILED DESCRIPTION
[0041] In order to make the technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0042] In the technical solution of the present invention, a method and apparatus for de novo drug design based on an optical quantum computer are provided, thereby enabling rapid and accurate de novo drug design.
[0043] FIG1 is a flow chart of a method for de novo drug design based on an optical quantum computer in a specific embodiment of the present invention. As shown in FIG1 , in a specific embodiment of the present application, the method for de novo drug design based on an optical quantum computer includes:
[0044] Step 101: Obtain the ligand-receptor complex structure that needs to be optimized.
[0045] In this step, the ligand-receptor complex structure that needs to be optimized can be obtained in advance to facilitate subsequent steps.
[0046] For example, in a specific embodiment of the present application, the user may input a corresponding ligand-receptor complex structure (eg, protein structure) that needs to be optimized.
[0047] For another example, as an example, in a specific embodiment of the present application, the ligand-receptor complex structure that needs to be optimized can also be downloaded from the protein data bank (PDB) (for example, the complex structure with ID number 5HMZ is downloaded, as shown in Figure 2).
[0048] Step 102: extracting the ligand molecule and the receptor structure from the complex structure, marking the optimizable portion of the ligand molecule, and pre-setting a fragment library for replacement.
[0049] After obtaining the above-mentioned ligand-receptor complex structure, the corresponding ligand molecule and receptor structure can be extracted from the ligand-receptor complex structure, and the optimizable parts can be marked in the ligand molecule (for example, the various parts of the ligand molecule and the groups that need to be optimized are marked, as shown in Figure 3), and the corresponding fragment library is pre-set to facilitate the replacement of the marked optimizable parts.
[0050] For example, in a specific embodiment of the present application, a large (greater than billions) fragment-based chemical library can be pre-set by mapping relevant drug groups to receptor pockets as a fragment library for replacement.
[0051] Step 103: Determine the attributes to be optimized based on the marked optimizable parts; and calculate the corresponding linear term coefficients and quadratic term coefficients based on the attributes to be optimized and the fragment library.
[0052] In this step, the property to be optimized can be determined based on the optimizable part of the above-mentioned labeled ligand molecule, and then calculations can be performed based on the property to be optimized and the pre-set fragment library for replacement to obtain the linear term coefficient and the quadratic term coefficient corresponding to the property to be optimized.
[0053] For example, in a specific embodiment of the present application, the properties to be optimized may be binding force and properties of the ligand itself.
[0054] For example, in a specific embodiment of the present application, the ligand's own properties may be: quantitative estimate of drug-likeness (QED), molecular weight (MW), lipid-water partition coefficient (ClogP), hydrogen-bond donors (HBD), hydrogen-bond acceptor (HBA), molecular polar surface area (PSA), etc. Of course, other ligand-specific properties may also be used, which are not listed here one by one.
[0055] For another example, in a specific embodiment of the present application, the calculation of the corresponding linear term coefficient and quadratic term coefficient based on the attribute to be optimized and the fragment library may include:
[0056] After the fragments are replaced according to the fragment library, the properties to be optimized after the fragment replacement are calculated, and the calculated properties are used as the corresponding linear term coefficients (single term coefficients) or quadratic term coefficients.
[0057] For example, as an example, in a specific embodiment of the present application, when the property to be optimized is binding force, after a single fragment is replaced according to the fragment library, the binding force after the single fragment replacement can be calculated, and the calculated binding force is used as the first-order term coefficient (single-order term coefficient).
[0058] In addition, as an example, in a specific embodiment of the present application, the binding force after a single fragment is replaced can be calculated using a scoring function.
[0059] For another example, as an example, in a specific embodiment of the present application, when the property to be optimized is the property of the ligand itself, after fragment replacement according to the fragment library, the property of the ligand itself after fragment replacement is calculated, and the calculated property of the ligand itself is used as the quadratic term coefficient.
[0060] For example, as an example, in a specific embodiment of the present application, when the attribute to be optimized is quantitative evaluation of drug-likeness (QED), molecular weight (MW), lipid-water partition coefficient (ClogP), hydrogen bond donor (HBD), hydrogen bond acceptor (HBA) or molecular polar surface area (PSA), after fragment replacement according to the fragment library, the QED, MW, ClogP, HBD, HBA or PSA after fragment replacement can be calculated respectively, and the calculated QED, MW, ClogP, HBD, HBA or PSA can be used as the corresponding quadratic term coefficients.
[0061] In addition, as an example, in a specific embodiment of the present application, the properties of the ligand itself after fragment replacement can be calculated by tool software (for example, the open source chemical information toolkit RDKit, etc.) to obtain the corresponding linear term coefficient.
[0062] Step 104: construct a corresponding mathematical model based on the calculated linear term coefficients and quadratic term coefficients.
[0063] After the linear term coefficient and the quadratic term coefficient are calculated, the corresponding mathematical model can be constructed based on the obtained linear term coefficient and the quadratic term coefficient.
[0064] For example, in a specific embodiment of the present application, the mathematical model may be a Quadratic Unconstrained Binary Optimization (QUBO) model.
[0065] For example, as an example, in a specific embodiment of the present application, in the QUBO model, the calculated binding force can be used as the linear term coefficient, and the ligand's own properties can be used as the quadratic term coefficient.
[0066] Through the above method, the properties of the ligand can be optimized to the best while finding the fragment composition that maximizes the binding force.
[0067] For example, in a specific embodiment of the present application, the QUBO model can be expressed by the following formula:
[0068] Where n represents the overall fragment length of the optimized ligand, i and j represent the positions of the fragments that can be optimized, and x i and x j Represents a decision variable, with a value of {0,1}, o i is the linear coefficient, which represents the properties of the ligand itself; t i,j is the coefficient of the quadratic term, which represents the molecular properties after fragment replacement.
[0069] The aforementioned QUBO model is a mathematical model for solving combinatorial optimization problems. This method transforms the process of synthesizing new molecules from fragment libraries into a quadratic unconstrained binary optimization problem. This translates the crucial de novo drug design problem in drug screening into a corresponding QUBO model, facilitating the subsequent use of quantum computers for calculations. This QUBO model can be used to improve the affinity of small molecule drugs and optimize their properties.
[0070] Of course, in the technical solution of the present invention, the above mathematical model may also be other suitable mathematical models that can be used for de novo drug design, which will not be listed here one by one.
[0071] Step 105 , converting the constructed mathematical model into a corresponding Ising model, inputting the model into an optical quantum computer for solution, and calculating and obtaining the optimized structure and ligand information.
[0072] After constructing the above-mentioned QUBO mathematical model, the mathematical model can be converted into the corresponding Ising model, and then solved using an optical quantum computer. The optimized structure and ligand information can be obtained through calculation.
[0073] For example, as an example, in a specific embodiment of the present application, when the above-mentioned mathematical model is a QUBO model, since each binary variable in the Ising model can be represented by the state of a quantum bit, and the Hamiltonian can correspond to the energy of the quantum system, the QUBO problem and the Ising model are one-to-one corresponding, so the QUBO model can be converted into the corresponding Ising model, thereby converting the QUBO problem into a special Ising model.
[0074] After converting the QUBO model into the corresponding Ising model, the Ising model can be input into the coherent Ising quantum computer CIM for solution. Since quantum computers have the characteristics of entangled states, superposition states, and full connectivity, the corresponding optimal solution can be calculated quickly and accurately through this quantum computer, thereby obtaining the optimized structure and ligand information. CIM can use nonlinear optical effects based on optical interference to solve the QUBO problem. Specifically, after converting the QUBO problem into a special Ising model, CIM can encode the Ising model into the amplitude and phase of a set of optical pulses, and then use the interference effect to transmit the coupling relationship between different spins to the optical interference; then, the nonlinear optical effect is used to convert the light pulses into the corresponding spin values. Finally, the ground state of the Ising model can be calculated, and the optimal solution can be read from the ground state of the quantum bit to obtain the optimal solution to the QUBO problem.
[0075] The coherent Ising machine can efficiently process large-scale QUBO problems with high parallelism and fast response speed. Therefore, the optimal solution can be read from the ground state of the quantum bit, thereby obtaining the optimized structure and ligand information.
[0076] Therefore, through the above steps 101 to 105, de novo drug design based on optical quantum computers can be achieved.
[0077] In addition, as an example, in a specific embodiment of the present application, the above method may further include:
[0078] Step 106: output the optimized structure and ligand information.
[0079] After the optimized structure and ligand information are calculated, the optimized structure and ligand information can be further output.
[0080] For example, as an example, in a specific embodiment of the present application, the obtained optimized 3D binding structure and ligand information can be displayed to the user through a display device (eg, a display, a user interface, etc.).
[0081] Therefore, in the technical solution of the present invention, the user only needs to input the ligand-receptor complex structure file that needs to be optimized on the user side, and the initial ligand structure and receptor structure of the complex structure can be displayed on the user side, which is convenient for the user to perform visual three-dimensional operations; then, after the user determines the groups that need to be replaced and the properties that need to be improved (that is, marking the optimizable parts of the ligand molecule), the server side can construct a corresponding mathematical model (for example, a QUBO model) according to the parameters determined by the user, and then calculate the mathematical model through a quantum computer to obtain the optimal solution, thereby obtaining the optimized structure and ligand information.
[0082] Since in the technical solution of the present invention, the entire process is calculated using an optical quantum computer, the time consumption will be greatly reduced, and the user can see the screened structural models as well as the scoring and illustration attribute information on the user-side interface; these molecules can then be used for the next step of experimental verification to obtain lead compounds.
[0083] In addition, in the technical solution of this application, a de novo drug design device based on an optical quantum computer is also proposed.
[0084] FIG4 is a schematic diagram of the structure of a de novo drug design device based on an optical quantum computer in a specific embodiment of the present invention. As shown in FIG1 , in a specific embodiment of the present application, the de novo drug design device based on an optical quantum computer includes: a marker 41, a fragment library 42, a model construction converter 43, and an optical quantum computer 44;
[0085] The marker is used to obtain the ligand-receptor complex structure that needs to be optimized; extract the ligand molecule and the receptor structure from the complex structure, and mark the optimizable part of the ligand molecule;
[0086] The fragment library is a fragment-based chemical library used for fragment replacement;
[0087] The model construction converter is used to determine the attribute to be optimized based on the marked optimizable portion; calculate the corresponding linear term coefficient and quadratic term coefficient based on the attribute to be optimized and the fragment library; construct a corresponding mathematical model based on the calculated linear term coefficient and quadratic term coefficient, and convert the constructed mathematical model into a corresponding Ising model;
[0088] The optical quantum computer is used to solve the Ising model and calculate the optimized structure and ligand information.
[0089] In addition, as an example, in a specific embodiment of the present application, after the above-mentioned optical quantum computer calculates the optimized structure and ligand information, it can further output the obtained optimized structure and ligand information.
[0090] In addition, as an example, in a specific embodiment of the present application, the above-mentioned de novo drug design device based on an optical quantum computer may further include: a display device (e.g., a display, a user interface, etc.);
[0091] The display device is used to display the obtained optimized 3D binding structure and ligand information.
[0092] In addition, as an example, in a specific embodiment of the present application, the optical quantum computer may be a coherent Ising quantum computer, or may be other suitable optical quantum computers.
[0093] In summary, by using the technical solution of the present invention, a huge drug library can be screened with high throughput. The user only needs to input the corresponding protein structure, and the server can calculate the suitable drug molecules and return them to the user, thereby accelerating the drug development process.
[0094] In addition, in the technical solution of the present invention, a large (greater than billions) fragment-based chemical library can be pre-set by mapping the relevant drug groups to the receptor pocket, as a fragment library for replacement. A large (greater than billions) fragment-based chemical library can be pre-set by mapping the relevant drug groups to the receptor pocket, as a fragment library for replacement, and then the process of synthesizing new molecules based on the fragment library is converted into a quadratic unconstrained binary optimization problem, so that the de novo drug design problem, which is very important in the drug screening process, can be converted into a corresponding QUBO mathematical model, and then the CIM quantum computer is used to solve the mathematical model, so that the mathematical model can be used to achieve the improvement of the affinity of small molecule drugs and the optimization of their properties. Therefore, by using the technical solution of the present invention, the affinity of small molecule drugs and proteins can be quickly optimized, and their various properties can be improved, helping researchers to screen out potential lead compounds and assist in the drug development process.
[0095] Furthermore, the technical solution of the present invention limits the search space for molecules with drug-like properties that match the binding pocket, and then optimizes them for synthetic feasibility and novelty, thereby providing significant commercial advantages over existing technologies. Users can upload different protein structures, which are then converted by the server and the optimal calculation results are given by the quantum computer, displaying the three-dimensional models of small molecules and proteins to the user. For de novo drug design problems, quantum computers are more efficient and faster than traditional computers, so they can provide results faster and more accurately.
[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method and device for de novo drug design based on optical quantum computers, characterized in that: The method includes: Obtain the ligand-receptor complex structure that needs to be optimized; Extracting ligand molecules and receptor structures from the ligand-receptor complex structure, marking the optimizable parts of the ligand molecules, and pre-setting a fragment library for replacement; Determine the properties to be optimized based on the marked optimizable parts; calculate the corresponding linear coefficients and quadratic coefficients based on the properties to be optimized and the fragment library Construct a corresponding mathematical model based on the calculated linear term coefficient and quadratic term coefficient; The constructed mathematical model is converted into the corresponding Ising model and input into an optical quantum computer for solution, and the optimized structure and ligand information are calculated.
2. The method according to claim 1, characterized in that The method further includes: Output the optimized structure and ligand information.
3. The method according to claim 1, wherein: A fragment-based chemical library is pre-configured by mapping relevant drug groups to receptor pockets, serving as a fragment library for substitution.
4. The method according to claim 1, wherein: The properties to be optimized are binding force and ligand properties.
5. The method according to claim 1, wherein The corresponding linear coefficients and quadratic coefficients calculated based on the properties to be optimized and the fragment library include: After the fragments are replaced according to the fragment library, the properties to be optimized after the fragment replacement are calculated, and the calculated properties are used as the corresponding linear term coefficients or quadratic term coefficients.
6. The method according to claim 5, characterized in that: When the property to be optimized is binding strength, after replacing a single fragment according to the fragment library, the binding strength after the single fragment replacement is calculated, and the calculated binding strength is used as the linear term coefficient; When the property to be optimized is the property of the ligand itself, after fragment replacement according to the fragment library, the property of the ligand itself after fragment replacement is calculated, and the calculated property of the ligand itself is used as the quadratic term coefficient.
7. The method according to claim 1, wherein: The mathematical model is a quadratic unconstrained binary optimization model; In the quadratic unconstrained binary optimization model, the calculated binding force is used as the linear coefficient, and the ligand's own properties are used as the quadratic coefficient.
8. The method according to claim 7, characterized in that The quadratic unconstrained binary optimization model is expressed by the following formula: Among them, QUBO is a quadratic unconstrained binary optimization model, n represents the overall fragment length of the optimized ligand, i and j represent the fragment positions that can be optimized, and x i and x j Represents a decision variable, with a value of {0,1}, o i is the linear coefficient, which represents the properties of the ligand itself; t i,j is the coefficient of the quadratic term, which represents the molecular properties after fragment replacement.
9. The method according to claim 1, wherein: The optical quantum computer is a coherent Ising quantum computer.
10. A de novo drug design device based on an optical quantum computer, characterized in that: The device includes: a tagger, a fragment library, a model building converter and an optical quantum computer; The marker is used to obtain the ligand-receptor complex structure that needs to be optimized; extract the ligand molecule and the receptor structure from the ligand-receptor complex structure, and mark the optimizable part of the ligand molecule; The fragment library is a fragment-based chemical library used for fragment replacement; The model construction converter is used to determine the attribute to be optimized based on the marked optimizable portion; calculate the corresponding linear term coefficient and quadratic term coefficient based on the attribute to be optimized and the fragment library; construct a corresponding mathematical model based on the calculated linear term coefficient and quadratic term coefficient, and convert the constructed mathematical model into a corresponding Ising model; The optical quantum computer is used to solve the Ising model and calculate the optimized structure and ligand information.