Method and system for constructing free energy surface of molecular binding / dissociation path
By employing mutual information analysis and reinforcement learning-based adaptive path sampling, the subjectivity of fixed reaction coordinates and the omission of intermediate states in traditional molecular dynamics simulations are resolved, improving the accuracy and efficiency of free energy surface calculations and providing a more reliable tool for studying molecular interaction mechanisms.
Patent Information
- Application Number
- CN202511589818.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-03
- Publication Date
- 2026-03-03
AI Technical Summary
In traditional molecular dynamics simulations, the selection of fixed reaction coordinates relies on subjective experience, leading to the omission of intermediate states and deviations in free energy calculations, thus failing to fully explore the conformation space.
Candidate reaction coordinates are screened using mutual information analysis, and adaptive path sampling is performed through reinforcement learning to construct the free energy surface of molecular binding/dissociation paths.
It improves the accuracy and efficiency of free energy surface calculations and provides a more reliable tool for studying molecular interaction mechanisms.
Smart Images
Figure CN121601053A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of molecular free energy surface construction technology, and in particular to a method and system for constructing molecular binding / dissociation path free energy surfaces. Background Technology
[0002] In the field of molecular dynamics simulations, accurately calculating the free energy surface of molecular binding / dissociation processes is a key challenge. Traditional methods typically use fixed reaction coordinates (RC) to guide the simulation process. This approach has two main drawbacks: first, the choice of a fixed RC often depends on subjective experience, which may lead to the omission of important intermediate states; second, molecular dynamics simulations under fixed RC constraints cannot fully explore the conformational space, resulting in biased free energy calculations. Summary of the Invention
[0003] Therefore, it is necessary to provide a method and system for constructing the free energy surface of molecular binding / dissociation pathways to address the problems in related technologies.
[0004] To achieve the above objectives, in a first aspect, this application provides a method for constructing a molecular binding / dissociation path free energy surface, the method comprising:
[0005] Mutual information analysis was used to quantify the correlation between the distance / angle of atomic pairs and the bonding state in order to screen out candidate reaction coordinates; short-term pre-simulation was performed on the candidate reaction coordinates to determine the dominant reaction coordinates;
[0006] Using the dominant response coordinates as navigation, adaptive path sampling is performed through reinforcement learning to obtain an optimized path;
[0007] Calculate the free energy of the optimized path, and construct the free energy surface of the molecular binding / dissociation path based on the obtained free energy.
[0008] The above-mentioned method for constructing the free energy surface of molecular binding / dissociation paths can solve the problems of subjectivity and intermediate state omission in traditional fixed reaction coordinate methods by automatically identifying the dominant reaction coordinates and using adaptive path sampling technology. This can improve the calculation accuracy and efficiency of the free energy surface and provide a more reliable theoretical tool for the study of molecular interaction mechanisms.
[0009] In some embodiments, mutual information analysis is used to quantify the correlation between atom pair distance / angle and binding state to screen candidate reaction coordinates; short-time pre-simulation verification is performed on the candidate reaction coordinates to determine the dominant reaction coordinates, including:
[0010] Extract geometric parameters to obtain a parameter set;
[0011] Define a combined state indicator function and calculate the mutual information value between each geometric parameter and the combined state indicator function;
[0012] Candidate reaction coordinates are selected based on the mutual information values;
[0013] The candidate reaction coordinates are verified by short-time pre-simulation to determine the dominant reaction coordinates.
[0014] In some embodiments, geometric parameters are extracted to obtain a parameter set, including calculating the distances and angles between all atom pairs in the molecular system to obtain the parameter set.
[0015] In some embodiments, the formula for the mutual information value between each geometric parameter and the binding state indicator function is:
[0016]
[0017] Among them, I(p) i B) represents the mutual information value; p i Let p be the geometric parameter; B be the state indication function; and p(x,y) be the joint probability, i.e., the geometric parameter p. i The joint probability of taking the value x and having the molecular state y; p(x) is the geometric parameter p i The marginal probability of taking the value x; y(y) is the marginal probability of the molecule being state y.
[0018] In some embodiments, using the dominant response coordinates as navigation, adaptive path sampling is performed through reinforcement learning to obtain an optimized path, including:
[0019] Build a reinforcement learning environment;
[0020] Construct a policy network based on the reinforcement learning environment;
[0021] Adaptive sampling is performed based on the policy network to obtain the motion trajectory; several transition samples are obtained in the process of obtaining the motion trajectory.
[0022] The network parameters of the policy network are updated based on the transferred samples, and the adaptive sampling is updated based on the updated policy network to update the motion trajectory.
[0023] The process is repeated until the termination condition is met, in order to obtain the optimized path.
[0024] In some embodiments, the process of constructing a policy network based on the reinforcement learning environment includes: a deep deterministic policy gradient algorithm with a dual-network structure, wherein the dual-network structure includes an Actor network and a Critic network.
[0025] In some embodiments, the termination conditions include: path integrity, energy barrier convergence, and sufficient sampling.
[0026] In some embodiments, the free energy of the optimized path is calculated, and a molecular binding / dissociation path free energy surface is constructed based on the obtained free energy, including:
[0027] Calculate the solvation energy and potential energy of each sampled conformation in the optimized path;
[0028] Based on the aforementioned potential energy smoothing energy curve;
[0029] Calculate the entropy term for each sampling conformation in the optimized path;
[0030] The free energy of the optimized path is obtained based on the energy curve and the entropy term;
[0031] Based on the obtained free energy, a continuous free energy curve is obtained, and the free energy surface of molecular binding / dissociation path is constructed.
[0032] In some embodiments, the energy curve is smoothed using cubic spline interpolation based on the potential energy.
[0033] Secondly, this application also provides a system for constructing molecular binding / dissociation path free energy surfaces, the system comprising:
[0034] The dominant reaction coordinate acquisition module is used to quantify the correlation between the distance / angle of atomic pairs and the bonding state using mutual information analysis to screen out candidate reaction coordinates; and to perform short-time pre-simulation verification on the candidate reaction coordinates to determine the dominant reaction coordinates.
[0035] The optimized path acquisition module is used to obtain an optimized path by using the dominant response coordinates as navigation and driving unbiased sampling through reinforcement learning.
[0036] The free energy surface acquisition module is used to calculate the free energy of the optimized path and construct the free energy surface of the molecular binding / dissociation path based on the obtained free energy.
[0037] The aforementioned system for constructing the free energy surface of molecular binding / dissociation paths automatically identifies the dominant reaction coordinates through the dominant reaction coordinate acquisition module and adopts adaptive path sampling technology through the optimized path acquisition module. This solves the subjectivity and intermediate state omission problems of traditional fixed reaction coordinate methods, improves the calculation accuracy and efficiency of free energy surfaces, and provides a more reliable theoretical tool for the study of molecular interaction mechanisms. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0039] Figure 1 This is a flowchart of a method for constructing the free energy surface of a molecular binding / dissociation pathway provided in one embodiment of this application;
[0040] Figure 2 This is a structural block diagram of a system for constructing the molecular binding / dissociation path free energy surface provided in another embodiment of this application.
[0041] Figure labeling: 10, Dominant reaction coordinate acquisition module; 20, Optimized path acquisition module; 30, Free energy surface acquisition module. Detailed Implementation
[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0043] In one embodiment, see Figure 1 This application also provides a method for constructing a molecular binding / dissociation path free energy surface, the method comprising the following steps: S10 to S30.
[0044] S10: The mutual information analysis method is used to quantify the correlation between the distance / angle of the atom pair and the bonding state to screen out candidate reaction coordinates; the candidate reaction coordinates are verified by short-time pre-simulation to determine the dominant reaction coordinates.
[0045] S20: Using the dominant reaction coordinates as navigation, adaptive path sampling is performed through reinforcement learning to obtain an optimized path.
[0046] S30: Calculate the free energy of the optimized path and construct the free energy surface of the molecular binding / dissociation path based on the obtained free energy.
[0047] The method for constructing the free energy surface of molecular binding / dissociation paths in this embodiment can solve the problems of subjectivity and intermediate state omission in traditional fixed reaction coordinate methods by automatically identifying the dominant reaction coordinates and using adaptive path sampling technology. This can improve the calculation accuracy and efficiency of the free energy surface and provide a more reliable theoretical tool for the study of molecular interaction mechanisms.
[0048] In step S10, please refer to Figure 1 In step S10, mutual information analysis is used to quantify the correlation between the distance / angle of atomic pairs and the binding state in order to screen out candidate reaction coordinates; the candidate reaction coordinates are then verified by short-time pre-simulation to determine the dominant reaction coordinates.
[0049] As an example, in step S10, the mutual information analysis method is used to quantify the correlation between the distance / angle of the atom pair and the binding state in order to screen out candidate reaction coordinates; the candidate reaction coordinates are verified by short-time pre-simulation to determine the dominant reaction coordinates, which may include the following steps: S101 to S104.
[0050] S101: Extract geometric parameters to obtain a parameter set.
[0051] S102: Define the combined state indicator function and calculate the mutual information value between each geometric parameter and the combined state indicator function.
[0052] S103: Filter candidate reaction coordinates based on the mutual information value.
[0053] S104: Perform short-time pre-simulation verification on the candidate reaction coordinates to determine the dominant reaction coordinates.
[0054] As an example, in step S101, geometric parameters are extracted to obtain a parameter set, including: calculating the distance d between all atom pairs (including ligand-acceptor, ligand-solvent, acceptor-solvent, etc.) in the molecular system. ij and angle θ ijk This is to obtain the parameter set. For N-atom systems, co-extraction can be performed. distance parameters and There are several angle parameters; the parameter set P = {p1, p2, ..., p...} is formed by these parameters. M}; where p1, p2, ..., p M There are M parameters.
[0055] As an example, in step S102, a binding state indicator function B(t) is defined (binding state is 1, dissociation state is 0), and each geometric parameter p is calculated. i The formula for the mutual information value of the combined state indication function B(t) can be:
[0056]
[0057] Among them, I(p) i B) represents the mutual information value; p i is the geometric parameter; b is the state indication function; p(x,y) is the joint probability, i.e., the geometric parameter p i The joint probability of taking the value x and having the molecular state y; p(x) is the geometric parameter pi The marginal probability of taking the value x; p(y) is the marginal probability of the molecule being in state y. Mutual information value I(p i The larger B is, the greater the geometric parameter p is. i The stronger the statistical correlation with the associated state (state indicator function B).
[0058] As an example, in step S103, during the process of filtering candidate reaction coordinates based on the mutual information values, all geometric parameters can be sorted in descending order of mutual information values, and the first three geometric parameters can be selected as candidate reaction coordinates RC. cand ={rc1,rc2,rc3}, where are three geometric parameters selected in descending order of mutual information values; to ensure the physical rationality of the candidate response coordinates, the following conditions must be met:
[0059] Distance parameter range:
[0060] Angular parameter range: 30° < θ ijk <150°.
[0061] As an example, step S104, performing a short-time pre-simulation verification on the candidate reaction coordinates to determine the dominant reaction coordinates, may include the following: performing a 10 ns short-time molecular dynamics simulation on the candidate reaction coordinates to monitor their dynamic behavior:
[0062] In the bound state: the values of the candidate reaction coordinates should be stable within a specific range (e.g., fluctuation < 10%);
[0063] During the dissociation process, the values of the candidate reaction coordinates should change monotonically without any jumps;
[0064] Energy barrier height: Calculate the free energy profile along the candidate reaction coordinates, and select the candidate reaction coordinate with the largest energy barrier height as the dominant reaction coordinate.
[0065] In step S20, please refer to Figure 1 In step S20, the dominant reaction coordinates are used as navigation, and adaptive path sampling is performed through reinforcement learning to obtain an optimized path.
[0066] As an example, in step S20, using the dominant response coordinates as navigation, adaptive path sampling is performed through reinforcement learning to obtain an optimized path, which may include the following steps: S201 to S205.
[0067] S201: Constructing a reinforcement learning environment.
[0068] S202: Construct a policy network based on the reinforcement learning environment.
[0069] S203: Adaptive sampling is performed based on the policy network to obtain the motion trajectory; several transfer samples are obtained in the process of obtaining the motion trajectory.
[0070] S204: Update the network parameters of the policy network based on the transferred samples, and update the adaptive sampling based on the updated policy network to update the motion trajectory.
[0071] S205: Update and iterate until the termination condition is met to obtain the optimized path.
[0072] As an example, in step S201, constructing the reinforcement learning environment may include the following:
[0073] The state space S is defined as the molecular conformation space, consisting of the dominant reaction coordinates s and the auxiliary geometric parameters p. i composition;
[0074] Action space A contains 6 basic motion directions: {Δx} + ,Δx - ,Δy + ,Δy - ,Δz + ,Δz -};
[0075] Reward function R(s) t ,a t ,s t+1 The design is as follows:
[0076]
[0077] Among them, s t The dominant reaction coordinate value at the current time step; s t+1 The coordinates of the dominant reaction at the next time step after the action is performed; Δt is the fixed molecular dynamics simulation step size. To the current dominant reaction coordinate value s t The free energy gradient at point α; α and β are weighting parameters.
[0078] As an example, in step S202, the process of constructing the policy network based on the reinforcement learning environment may include: using a deep deterministic policy gradient (DDPG) algorithm with a dual-network structure, wherein the dual-network structure includes an Actor network and a Critic network.
[0079] Specifically, for the Actor network: input state s t The output action probability distribution π(a|s) can be represented by the following formula:
[0080] π(a|s)=softmax(W2·ReLU(W1·s+b1)+b2)
[0081] Where W1 is the weight matrix, connecting the parameters of the input layer and the hidden layer; W2 is the weight matrix, connecting the parameters of the hidden layer and the output layer; s is the state vector (i.e., the input state); b1 is the hidden layer bias term; b2 is the output layer bias term; ReLU is the non-linear activation function; and softmax is the normalized exponential function.
[0082] Specifically, for the Critic network: the formula for evaluating the state-action value Q(s,a) can be:
[0083] Q(s,a)=V(s)+A(s,a)
[0084] Where V(s) is the state value function and A(s,a) is the advantage function.
[0085] As an example, in step S203, adaptive sampling is performed based on the policy network to obtain the motion trajectory; in the process of obtaining the motion trajectory, several transfer samples are obtained, which may include the following:
[0086] Initialization: Load the pre-trained policy network from the associative state (s0);
[0087] Motion trajectory generation: At each step, select action a based on the current strategy. t ~π(a|s t ), π(a|s t = softmax(W2·ReLU(W1·s) t +b1)+b2);
[0088] The molecular dynamics step size Δt = 0.5 ps was used.
[0089] Record new state s t+1 And instant rewards.
[0090] As an example, in step S204, updating the network parameters of the policy network based on the transferred samples and updating the adaptive sampling based on the updated policy network to update the motion trajectory may include the following:
[0091] Priority playback: Store and transfer samples (s) t ,a t ,r t ,s t+1 Samples are taken first from the playback buffer based on the timing difference error δ:
[0092] δ=r t +γQ(s t+1 ,a t+1 )-Q(s t ,a t )
[0093] Where, r t For instant rewards; γ is the discount factor; Q(s) t+1 ,a t+1 The assessment of future value, that is, the assessment of the value of action a. t After that, it reaches a new state s t+1 Below, the Critic network determines the optimal future action a. t+1 Value assessment; Q(s) t ,a t The old value assessment refers to the assessment performed on action a. t Previously, the Critic network focused on state s t Next, execute action a t The value assessment of this behavior;
[0094] Network update: Update the parameters θ of the policy network every 100 steps.
[0095] As an example, in step S205, the update iteration continues until the termination conditions are met to obtain the optimized path. The termination conditions include: path integrity, energy barrier convergence, and sufficient sampling.
[0096] Specifically, path integrity: the dominant reaction's reaction coordinate s value is in [s min ,s max The interval was traversed 3 times; among them, s min As the minimum value of the dominant reaction coordinate, s max The maximum value of the dominant reaction coordinate.
[0097] Specifically, energy barrier convergence: the difference in energy barrier height between two consecutive traversals is <0.5 kcal / mol.
[0098] Specifically, sampling adequacy: the statistical error σ(s) of all s intervals is less than 0.3 kcal / mol.
[0099] In step S30, please refer to Figure 1 In step S30, the free energy of the optimized path is calculated, and the free energy surface of the molecular binding / dissociation path is constructed based on the obtained free energy.
[0100] As an example, in step S30, the free energy of the optimized path is calculated, and the free energy surface of the molecular binding / dissociation path is constructed based on the obtained free energy, which may include the following steps: S301 to S305.
[0101] S301: Calculate the solvation energy and potential energy of each sampled conformation in the optimized path.
[0102] S302: Smooth the energy curve based on the potential energy.
[0103] S303: Calculate the entropy term for each sampling conformation in the optimized path.
[0104] S304: Obtain the free energy of the optimized path based on the energy curve and the entropy term.
[0105] S305: Based on the obtained free energy, a continuous free energy curve is obtained, and the free energy surface of molecular binding / dissociation path is constructed.
[0106] As an example, in step S301, each sampling conformation s in the optimized path can be calculated based on the following formula. i Potential energy E(s) i ):
[0107] E(s i ) = E MM (s i )+E solv (s i )
[0108] Among them, E MM (s i ) is a conformation s i The molecular mechanical energy; E solv (s i ) is a conformation s i The solvation energy.
[0109] As an example, in step S302, the energy curve can be smoothed using cubic spline interpolation based on the potential energy, and the corresponding formula is:
[0110]
[0111] in, E(s) represents the smoothed, continuous energy. i ) is the sampling conformation s i Potential energy; B i (s) represents the B-spline basis function; N represents the number of sampled configurations.
[0112] As an example, in step S303, the entropy term S(s) for each sampling conformation in the optimized path can be calculated based on the following formula:
[0113] S(s)=k B ·D(s)
[0114] Where, k B Where is the Boltzmann constant; D(s) is the conformational diversity index, and the corresponding formula is:
[0115]
[0116] pk Let be the probability of conformation k appearing within the window s±Δs; M is the number of unique conformations within the window.
[0117] As an example, in step S304, the free energy F(s) of the optimized path can be obtained based on the energy curve and the entropy term using the following formula:
[0118]
[0119] in, S is the smoothed continuous energy; S(s) is the entropy term; T is the absolute temperature in Kelvin.
[0120] As an example, in step S305, a continuous free energy curve can be obtained based on the obtained free energy, and a molecular binding / dissociation path free energy surface can be constructed based on the free energy region.
[0121] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.
[0122] In another embodiment, please refer to Figure 1 See Figure 2 This application also provides a system for constructing the free energy surface of molecular binding / dissociation paths. The system includes: a dominant reaction coordinate acquisition module 10, an optimized path acquisition module 20, and a free energy surface acquisition module 30. The dominant reaction coordinate acquisition module 10 is used to quantify the correlation between the distance / angle of atomic pairs and the binding state using mutual information analysis to screen candidate reaction coordinates; the candidate reaction coordinates are then subjected to short-time pre-simulation verification to determine the dominant reaction coordinates; the optimized path acquisition module 20 is used to obtain an optimized path by using the dominant reaction coordinates as navigation and driving unbiased sampling through reinforcement learning; the free energy surface acquisition module 30 is used to calculate the free energy of the optimized path and construct the free energy surface of molecular binding / dissociation paths based on the obtained free energy.
[0123] In the molecular binding / dissociation path free energy surface construction system of this embodiment, the dominant reaction coordinates are automatically identified by the dominant reaction coordinate acquisition module 10, and the adaptive path sampling technology is adopted by the optimized path acquisition module 20. This can solve the subjectivity and intermediate state omission problems of the traditional fixed reaction coordinate method, improve the calculation accuracy and efficiency of the free energy surface, and provide a more reliable theoretical tool for the study of molecular interaction mechanisms.
[0124] As an example, the molecular binding / dissociation path free energy surface construction system of this embodiment can be used to perform, for example... Figure 1 The method for constructing the molecular binding / dissociation path free energy surface in the corresponding embodiments.
[0125] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features of the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0126] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.
Claims
1. A method for constructing the free energy surface of a molecular binding / dissociation path, characterized in that, include: Mutual information analysis was used to quantify the correlation between the distance / angle of atomic pairs and the bonding state in order to screen out candidate reaction coordinates; short-term pre-simulation was performed on the candidate reaction coordinates to determine the dominant reaction coordinates; Using the dominant response coordinates as navigation, adaptive path sampling is performed through reinforcement learning to obtain an optimized path; Calculate the free energy of the optimized path, and construct the free energy surface of the molecular binding / dissociation path based on the obtained free energy.
2. The method for constructing the molecular binding / dissociation path free energy surface according to claim 1, characterized in that, Mutual information analysis is used to quantify the correlation between atomic pair distance / angle and bonding state to screen candidate reaction coordinates; short-term pre-simulation is performed on the candidate reaction coordinates to determine the dominant reaction coordinates, including: Extract geometric parameters to obtain a parameter set; Define a combined state indicator function and calculate the mutual information value between each geometric parameter and the combined state indicator function; Candidate reaction coordinates are selected based on the mutual information values; The candidate reaction coordinates are verified by short-time pre-simulation to determine the dominant reaction coordinates.
3. The method for constructing the molecular binding / dissociation path free energy surface according to claim 2, characterized in that, Extracting geometric parameters to obtain a parameter set includes calculating the distances and angles between all atom pairs in the molecular system to obtain the parameter set.
4. The method for constructing the molecular binding / dissociation path free energy surface according to claim 2, characterized in that, The formula for the mutual information value between each geometric parameter and the binding state indicator function is: Among them, I(p) i B) represents the mutual information value; p i Let p be the geometric parameter; B be the state indication function; and p(x,y) be the joint probability, i.e., the geometric parameter p. i The joint probability of taking the value x and having the molecular state y; p(x) is the geometric parameter p i The marginal probability of taking the value x; p(y) is the marginal probability of the molecule being in state y.
5. The method for constructing the molecular binding / dissociation path free energy surface according to claim 1, characterized in that, Using the dominant response coordinates as navigation, adaptive path sampling is performed through reinforcement learning to obtain an optimized path, including: Build a reinforcement learning environment; Construct a policy network based on the reinforcement learning environment; Adaptive sampling is performed based on the policy network to obtain the motion trajectory; several transition samples are obtained in the process of obtaining the motion trajectory. The network parameters of the policy network are updated based on the transferred samples, and the adaptive sampling is updated based on the updated policy network to update the motion trajectory. The process is repeated until the termination condition is met, in order to obtain the optimized path.
6. The method for constructing the molecular binding / dissociation path free energy surface according to claim 5, characterized in that, The process of constructing a policy network based on the reinforcement learning environment includes: a deep deterministic policy gradient algorithm with a dual-network structure, wherein the dual-network structure includes an Actor network and a Critic network.
7. The method for constructing the molecular binding / dissociation path free energy surface according to claim 5, characterized in that, The termination conditions include: path integrity, energy barrier convergence, and sufficient sampling.
8. The method for constructing the molecular binding / dissociation path free energy surface according to any one of claims 1 to 7, characterized in that, Calculate the free energy of the optimized path, and construct the molecular binding / dissociation path free energy surface based on the obtained free energy, including: Calculate the solvation energy and potential energy of each sampled conformation in the optimized path; Based on the aforementioned potential energy smoothing energy curve; Calculate the entropy term for each sampling conformation in the optimized path; The free energy of the optimized path is obtained based on the energy curve and the entropy term; Based on the obtained free energy, a continuous free energy curve is obtained, and the free energy surface of molecular binding / dissociation path is constructed.
9. The method for constructing the molecular binding / dissociation path free energy surface according to claim 8, characterized in that, The energy curve is smoothed using cubic spline interpolation based on the potential energy.
10. A system for constructing the free energy surface of a molecular binding / dissociation path, characterized in that, include: The dominant reaction coordinate acquisition module is used to quantify the correlation between the distance / angle of atomic pairs and the bonding state using mutual information analysis to screen out candidate reaction coordinates; and to perform short-time pre-simulation verification on the candidate reaction coordinates to determine the dominant reaction coordinates. The optimized path acquisition module is used to obtain an optimized path by using the dominant response coordinates as navigation and driving unbiased sampling through reinforcement learning. The free energy surface acquisition module is used to calculate the free energy of the optimized path and construct the free energy surface of the molecular binding / dissociation path based on the obtained free energy.