A method and system for predicting ligand-target dissociation kinetics parameters

Abstract

This application relates to the field of computer-aided drug design technology, and in particular to a method and system for predicting ligand-target dissociation kinetic parameters. The method includes the following steps: constructing and parameterizing a protein-ligand system; applying constraints to the system to maintain structural stability; setting multiple sets of simulation condition pairs; globally scaling the non-bonded interaction potential energy of the system or the force derived from that potential energy under each set of simulation condition pairs; performing molecular dynamics simulations and detecting ligand dissociation events under each set of simulation condition pairs; estimating the dissociation rate under each simulation condition pair; and extrapolating to the target condition based on the relationship between the dissociation rate and a unified extrapolation independent variable to obtain the dissociation rate and residence time under the target condition. This invention solves the problem in existing technologies where it is difficult to reliably obtain the dissociation rate of ligands and corresponding target receptors at an affordable computational cost, and it also features controllable parameters and ease of parallelization.

Description

Technical Field

[0001] This application relates to the field of computer-aided drug design technology, and in particular to a method and system for predicting ligand-target dissociation kinetic parameters. Background Technology

[0002] Protein-ligand binding kinetics parameters are crucial in drug development. The dissociation rate constant directly characterizes the dynamics of ligand-target protein binding at the binding site, providing quantitative assessments of drug binding behavior, selectivity, in vivo exposure window, and dosing frequency. Therefore, predicting the dissociation rate constant rapidly and efficiently in computer-aided drug design workflows is a key research focus at the intersection of molecular dynamics simulation and drug design.

[0003] However, in conventional molecular dynamics simulations, the dissociation of ligands from their binding pockets is typically a rare event. At physiological temperatures, the system needs to overcome a high energy barrier, and the dissociation time scale is usually in the range of microseconds to seconds. Therefore, observing natural dissociation events through a single long-duration molecular dynamics simulation trajectory at physiological temperatures is often impractical within acceptable simulation costs.

[0004] To address the challenge of sampling rare events, existing technologies have proposed several classes of augmented sampling algorithms. The first class of methods accelerates sampling by constructing set variables and adding bias potentials or forces to the corresponding set variable dimensions; examples include meta-dynamics and its variants, umbrella sampling, and adaptive bias forces. The second class of methods does not rely on set variables, such as accelerated molecular dynamics and copy-swapping molecular dynamics simulations. These methods can accelerate the exploration of dissociation paths to some extent, but they still face the following problems in practical engineering applications.

[0005] First, some of the methods mentioned above require the pre-selection of appropriate set variables. A good set variable selection often depends on the researcher's experience. Generally, the selected set variables are required to describe the process under study and capture the relevant slow degrees of freedom processes; otherwise, it will lead to sampling bias or insufficient enhancement of sampling efficiency.

[0006] Secondly, the aforementioned enhanced sampling algorithm involves a large number of empirical parameter settings, and its implementation is also quite complex. Different systems may require different empirical parameters, and it is difficult to maintain a uniform standard in high-throughput virtual screening scenarios.

[0007] Furthermore, the interpretability and extrapolability of the dynamics are insufficient. Although the partial augmentation sampling algorithm can promote the faster occurrence of dissociation events, it requires additional reweighting operations and rigorous theoretical derivation to transform the augmented sampling observations into real observations under normal conditions.

[0008] In addition, existing technologies also employ methods that increase the simulation temperature to accelerate the dissociation process. However, these methods still face the following challenges: On the one hand, simply increasing the temperature may lead to significant rearrangement or even local instability of the protein binding pocket structure, thereby introducing dissociation pathways or non-physical dissociation mechanisms that are not intended for the target state; on the other hand, whether the kinetic processes at different temperatures satisfy a simple linear extrapolation relationship is closely related to factors such as the energy barrier morphology, solvent effects, and conformational entropy contributions of the system. Without a unified extrapolation independent variable and statistical modeling framework, the extrapolation results may still be unrobust and the confidence intervals may be difficult to quantify.

[0009] Therefore, there is still a need to develop a more engineered, parameter-controllable, and easily parallelizable computational process: under the premise of ensuring the stability of the architecture and the interpretability of the dissociation mechanism, it should be able to generate sufficient dissociation event samples within a limited simulation budget, and output the results of dissociation rate and residence time and their uncertainties at the target temperature through multi-condition fitting and extrapolation. Summary of the Invention

[0010] To address the problem that existing technologies struggle to reliably calculate the dissociation rate between ligands and their corresponding target receptors at an affordable computational cost, this invention proposes a method and system for predicting ligand-target dissociation kinetic parameters, which features controllable parameters and ease of parallelization.

[0011] To achieve the above-mentioned objectives of this invention, the technical solution adopted is as follows: A method for predicting ligand-target dissociation kinetic parameters includes the following steps: The protein-ligand system was constructed and parameterized to obtain the initial configuration and force field parameters for molecular dynamics simulations; Constraints are applied to the system to maintain structural stability, and the heat bath temperature is set to a simulated temperature. Multiple sets of simulation condition pairs are set up, each set of simulation condition pairs includes simulation temperature and global situation energy scaling factor, and a uniform extrapolation independent variable is defined as the ratio of the global situation energy scaling factor to the simulation temperature. Under each set of simulation conditions, the nonbonded interaction potential energy of the system or the force derived from the potential energy is globally scaled to obtain the scaled molecular dynamics sampling trajectory. Molecular dynamics simulations were performed under each set of simulation conditions, and ligand dissociation events were detected to obtain dissociation time samples containing censored information of non-dissociation. Based on the dissociation time samples, the dissociation rate under each simulation condition pair is estimated; Based on the relationship between the dissociation rate and the unified extrapolation independent variable, the dissociation rate and residence time under the target conditions are extrapolated to obtain the target conditions.

[0012] Preferably, performing molecular dynamics simulations and detecting ligand dissociation events under each set of simulation conditions specifically includes: performing multiple copies of molecular dynamics simulations for each set of simulation conditions, detecting whether the ligand has dissociated based on a preset dissociation criterion, and recording dissociation time samples; when the simulation runs for a preset maximum time and dissociation has not occurred, the corresponding sample is marked as right-censored data.

[0013] Furthermore, the dissociation criterion includes both a distance threshold criterion and a root mean square deviation criterion; wherein, the distance threshold criterion is that the distance between the ligand centroid and the binding site reference point is greater than a preset distance threshold; the root mean square deviation criterion is that the root mean square deviation of the ligand relative to the binding conformation reference conformation is greater than a preset deviation threshold; an event that simultaneously satisfies the above criteria and lasts for a duration not less than a preset duration is defined as a dissociation event.

[0014] Furthermore, the dissociation rate under each simulation condition pair is estimated based on the dissociation time sample as follows: assuming that the dissociation process follows first-order dynamics, the maximum likelihood estimation method is used to construct a likelihood function containing dissociated samples and undissociated right-censored samples, and the dissociation rate under each condition pair is solved by maximizing the likelihood function.

[0015] Furthermore, the likelihood function is:

[0016] Where E is the set of samples that have dissociated, C is the set of samples that have not dissociated, k is the dissociation rate, and t is the observed dissociation time. max The simulation duration is for the corresponding replica; the dissociation rate is obtained by maximizing the log-likelihood function.

[0017] Furthermore, extrapolating the relationship between the dissociation rate and the unified extrapolation independent variable to the target condition specifically involves: performing linear regression fitting with the unified extrapolation independent variable as the independent variable and the natural logarithm of the dissociation rate as the dependent variable; substituting the overall situation energy scaling factor of 1 and the simulated temperature as the target temperature into the fitted linear relationship to obtain the dissociation rate and residence time under the target condition.

[0018] Furthermore, the linear regression fitting is based on the generalized Arrhenius equation, and the fitting form is:

[0019] in, The intercept corresponds to the logarithm of the pre-exponential factor; The slope corresponds to the negative value of the ratio of the energy barrier to the Boltzmann constant; This is the global state energy scaling factor; To simulate temperature.

[0020] Furthermore, the global scaling satisfies any of the following:

[0021] Alternatively, the force derived from the original potential energy can be scaled:

[0022] in, The global situation energy scaling factor is... This is the potential energy function of the original non-bonded interactions. The force is derived from the original potential energy.

[0023] Furthermore, it also includes an uncertainty assessment step: performing bootstrap resampling or cross-validation on the dissociation time samples, and outputting the confidence interval or standard deviation of the dissociation rate and residence time under the target conditions.

[0024] A ligand-target dissociation kinetic parameter prediction system includes: a data processing module, a condition initialization module, a scaling and simulation module, a rate estimation and extrapolation module, and a result output module; The data processing module is used to construct and parameterize the protein-ligand system to obtain the initial configuration and force field parameters for molecular dynamics simulation. The condition initialization module is used to apply constraints to the system to maintain structural stability, set multiple sets of simulation condition pairs, each set of simulation condition pairs includes simulation temperature and global state energy scaling factor, and defines a uniform extrapolation independent variable as the ratio of the global state energy scaling factor to the simulation temperature; The scaling and simulation module is used to globally scale the nonbonded interaction potential energy of the system or the force derived from the potential energy under each set of simulation conditions and perform molecular dynamics simulations, detect ligand dissociation events, and obtain dissociation time samples containing censored information of non-dissociation. The rate estimation and extrapolation module is used to estimate the dissociation rate under each simulation condition based on the dissociation time sample, and extrapolate to the target condition based on the relationship between the dissociation rate and the unified extrapolation independent variable to obtain the dissociation rate and residence time under the target condition. The result output module is used to output the dissociation rate and residence time under the target conditions.

[0025] The beneficial effects of this invention are as follows: This invention achieves dual energy barrier compression by globally scaling the potential energy or force of non-bonded interactions and combining it with multiple temperature conditions. This makes rare dissociation events more likely to occur within a limited simulation time, thus obtaining more statistically usable dissociation samples under similar computational budgets and significantly improving the sampling efficiency of dissociation events. This invention applies constraints to proteins to maintain structural stability, ensuring the stability of binding site structures under heating and scaling conditions. This reduces non-physical dissociation mechanisms caused by protein instability and guarantees the physical rationality of dissociation paths. This invention defines a unified extrapolation independent variable as the ratio of the global potential energy scaling factor to the simulation temperature. This allows dissociation rates under different simulation conditions to be fitted linearly in the logarithmic rate space and extrapolated to the target conditions, facilitating automated implementation and result interpretation. This invention employs a statistical estimation method including right-censored data, fully utilizing the temporal information of undissociated trajectories and avoiding the information waste caused by simply discarding undissociated data, thereby improving the accuracy of dissociation rate estimation. The computational process proposed in this invention is engineering-friendly, highly parallel, easy to generate tasks in batches and perform multi-replica parallelism, and is adapted to high-throughput virtual screening and lead compound optimization processes. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a method for predicting ligand-target dissociation kinetic parameters according to one embodiment of the present invention.

[0027] Figure 2 This is a system block diagram of a ligand-target dissociation kinetic parameter prediction system according to one embodiment of the present invention. Detailed Implementation

[0028] 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.

[0029] Example 1 like Figure 1 As shown, a method for predicting ligand-target dissociation kinetic parameters includes the following steps: The protein-ligand system was constructed and parameterized to obtain the initial configuration and force field parameters for molecular dynamics simulations; Constraints are applied to the system to maintain structural stability, and the heat bath temperature is set to a simulated temperature. Multiple sets of simulation condition pairs are set up, each set of simulation condition pairs includes simulation temperature and global situation energy scaling factor, and a uniform extrapolation independent variable is defined as the ratio of the global situation energy scaling factor to the simulation temperature. Under each set of simulation conditions, the nonbonded interaction potential energy of the system or the force derived from the potential energy is globally scaled to obtain the scaled molecular dynamics sampling trajectory. Molecular dynamics simulations were performed under each set of simulation conditions, and ligand dissociation events were detected to obtain dissociation time samples containing censored information of non-dissociation. Based on the dissociation time samples, the dissociation rate under each simulation condition pair is estimated; Based on the relationship between the dissociation rate and the unified extrapolation independent variable, the dissociation rate and residence time under the target conditions are extrapolated to obtain the target conditions.

[0030] The initial configuration and force field parameters include any one or a combination of the following: specifying the protonation state of the protein and ligand; generating topological parameters and initial coordinate files for the protein-ligand complex; and performing energy minimization, stepwise heating, and isobaric equilibrium simulations to obtain the initial conformation for the production stage simulation.

[0031] The target temperature of the target condition It is 298 K or 300 K.

[0032] The constraints include any one or a combination of the following: imposing positional restrictions on key residues in the protein backbone or binding pocket; imposing constraints on specific bond lengths or bond angles of the protein.

[0033] Example 2 In this embodiment, the specific method for predicting ligand-target dissociation kinetic parameters includes the following steps: System Construction and Parameterization: Crystal structures of protein-ligand complexes were downloaded from the PDB database. The protonation states of the protein and ligands were specified, and the pKa values ​​were determined using PropKa. A solvent model (TIP3P) and a water model (and counterions) were added to neutralize the system charge. After system construction, force field parameters were specified: the AMBERff14SB force field was used for the protein, and the GAFF2 force field was used for the ligands. Energy minimization, stepwise temperature increase, and isobaric equilibrium simulations were performed to obtain the initial configuration and force field parameters for molecular dynamics simulations.

[0034] To prevent non-physical denaturation or unfolding of the protein during subsequent high-temperature or potential-scale simulations, constraints are imposed on the system to maintain structural stability. Specifically, positional constraints are imposed on the main chain atoms of the protein or domains far from the binding pocket, or distance constraints are imposed on key hydrogen bonds. Simultaneously, the heat bath temperature is set to the simulation temperature.

[0035] Multiple simulation condition pairs are set up, each pair including simulation temperature and global state energy scaling factor λ. The value of λ can range from 0.5 to 1.0, and the value of simulation temperature can range from 400K to 700K. The uniform extrapolation independent variable is defined as the ratio of λ to simulation temperature.

[0036] Under each set of simulation conditions, the nonbonded interaction potential energy (van der Waals forces and electrostatic forces) of the system is globally scaled. The scaling method can be a potential energy scaling approach:

[0037] Alternatively, force scaling can be used:

[0038] By reducing the dissociation energy barrier through the above scaling, the accelerated molecular dynamics sampling trajectory under this scaling condition is obtained.

[0039] For each set of simulation conditions, a multi-copy molecular dynamics simulation is performed. During the simulation, the dissociation of the ligand is detected in real time or post-processing based on a preset dissociation criterion. The moment of dissociation is recorded as a dissociation time sample; if dissociation does not occur after the simulation has run for a preset maximum time, the sample is marked as right-censored data.

[0040] Based on the dissociation time samples under each set of simulation conditions, including the precise time of dissociation that has occurred and the right censoring time of non-dissociation, the dissociation rate under that condition is estimated.

[0041] A linear regression was performed with the unified extrapolation independent variable as the independent variable and the natural logarithm of the dissociation rate as the dependent variable. The obtained linear relationship was extrapolated to the target conditions, with a target temperature and a global potential energy scaling factor of 1, to obtain the predicted dissociation rate under the target conditions.

[0042] Output the dissociation rate and residence time under the target conditions.

[0043] Example 3 This embodiment, based on embodiment 2, provides further detailed explanations of the parameter settings and calculation models for key steps.

[0044] Regarding the dissociation criterion, this embodiment employs a combined criterion to ensure the authenticity of the dissociation event. The dissociation criterion includes both a distance threshold criterion and a root-mean-square (RMS) deviation criterion. The distance threshold criterion calculates the distance between the ligand's centroid and the binding site reference point, requiring this distance to be greater than a preset distance threshold of 10 Å. The RMS deviation criterion calculates the RMS deviation of the ligand relative to the binding conformation reference conformation, requiring it to be greater than a preset deviation threshold of 2.5 Å. Furthermore, the duration of the aforementioned state is required to be no less than a preset duration of 100 ps.

[0045] Regarding rate estimation, this embodiment employs the maximum likelihood estimation method. It is assumed that the dissociation process follows a first-order dynamical Poisson process, with its probability density function being... The survival function is For a sample set containing right-censored data, construct the likelihood function:

[0046] Where E is the set of samples that dissociated, t is the observed dissociation time, and C is the set of samples that did not dissociate. Let k be the simulation duration for the corresponding replica. The optimal dissociation rate k is obtained by maximizing the log-likelihood function.

[0047] in, This represents the total number of events in which dissociation occurred. This formula makes full use of information from the undissociated trajectory, improving the accuracy of the estimation.

[0048] Regarding the linear regression and extrapolation model, this embodiment derives it based on the generalized Arrhenius equation. Under the approximation of a global scaling λ of the potential energy surface, the energy barrier height... Become According to the Arrhenius formula It can be deduced that:

[0049] Among them, intercept The slope corresponds to the logarithm of the pre-exponential factor. The sign is negative when the slope β is defined as positive, and positive when the slope β is defined as negative. By plotting the natural logarithm of the dissociation rate against the unified extrapolation independent variable and performing linear regression, we can verify whether the dynamic behavior conforms to the physical model and perform extrapolation accordingly. By substituting the overall state energy scaling factor of 1 and the simulated temperature as the target temperature into the fitting relationship, we can obtain the dissociation rate and residence time under the target conditions.

[0050] Regarding uncertainty assessment, this embodiment uses a bootstrap method to resample the duplicate trajectory or dissociation time sample to quantify the robustness of the extrapolation results.

[0051] Example 4 This embodiment uses the T4 lysozyme L99A mutant and benzene ligand system as an example to illustrate the application effect of the present invention.

[0052] Data preparation: The complex structure of the T4 lysozyme L99A mutant and benzene ligand was selected. The protein was parameterized using an AMBERff14SB force field, and the ligand was parameterized using a GAFF2 force field. A TIP3P water box was added, and energy minimization and equilibrium simulations were performed to obtain the initial configuration for the production stage simulation.

[0053] Structural constraint setting: Positional constraints are applied to the Cα atoms of the protein backbone to maintain the structural stability of the binding pocket under high temperature and scaling conditions.

[0054] Condition settings: Set 5 sets of simulation temperatures, 400K, 450K, 500K, 550K, and 600K, and 5 sets of global potential scaling factors, 1.0, 0.9, 0.8, 0.7, and 0.6, to generate 25 simulation condition pairs.

[0055] Simulation execution: Ten copies of 50 ns molecular dynamics simulations were run under each simulation condition pair. Under high temperature and low λ conditions (600 K, λ = 0.6), rapid dissociation of the ligands was observed within 1–5 ns; under low temperature and high λ conditions (e.g., 400 K, λ = 1.0), most trajectories did not dissociate within 50 ns, which were recorded as right-censored data.

[0056] Dissociation detection: The combined criteria described in Example 2 are used, with the distance threshold set to 10 Å, the root mean square deviation threshold set to 2.5 Å, and the duration threshold set to 100 ps.

[0057] Rate estimation and fitting: The dissociation rate for each simulation condition pair was calculated using the maximum likelihood estimation method from Example 2. The independent variable λ / T was extrapolated uniformly. sim A scatter plot was drawn with the x-axis representing the natural logarithm of the dissociation rate and the y-axis representing the distance from the x-axis. The results show that the data points exhibit a good linear relationship, with a linear correlation coefficient greater than 0.95, indicating that the unified extrapolation independent variable proposed in this invention effectively captures the combined effects of temperature and potential energy scaling on the kinetics.

[0058] Extrapolation results: Extrapolating the straight line to the target temperature of 298K and λ=1, we obtain the predicted dissociation rate under the target conditions.

[0059] Uncertainty analysis: By resampling 1000 times using the bootstrap method, the 95% confidence interval of the prediction results was obtained, demonstrating the robustness of the method.

[0060] Example 5 like Figure 2 As shown, a ligand-target dissociation kinetic parameter prediction system includes: a data processing module, a condition initialization module, a scaling and simulation module, a rate estimation and extrapolation module, and a result output module; The data processing module is used to construct and parameterize the protein-ligand system to obtain the initial configuration and force field parameters for molecular dynamics simulation. The condition initialization module is used to apply constraints to the system to maintain structural stability, setting multiple sets of simulation condition pairs. Each set of simulation condition pairs includes a simulation temperature and a global state energy scaling factor, and defines a uniform extrapolation independent variable as the ratio of the global state energy scaling factor to the simulation temperature. ,in ,and satisfy and set the target temperature It is 298 K or 300 K; The scaling and simulation module is used to globally scale the nonbonded interaction potential energy of the system or the force derived from the potential energy under each set of simulation conditions and perform molecular dynamics simulations, detect ligand dissociation events, and obtain dissociation time samples containing censored information of non-dissociation. The rate estimation and extrapolation module is used to estimate the dissociation rate under each simulation condition based on the dissociation time sample, and extrapolate to the target condition based on the relationship between the dissociation rate and the unified extrapolation independent variable to obtain the dissociation rate and residence time under the target condition. The result output module is used to output the dissociation rate and residence time under the target conditions.

Claims

1. A method for predicting ligand-target dissociation kinetic parameters, characterized in that, Includes the following steps: The protein-ligand system was constructed and parameterized to obtain the initial configuration and force field parameters for molecular dynamics simulations; Constraints are applied to the system to maintain structural stability, and the heat bath temperature is set to a simulated temperature. Multiple sets of simulation condition pairs are set up, each set of simulation condition pairs includes simulation temperature and global situation energy scaling factor, and a uniform extrapolation independent variable is defined as the ratio of the global situation energy scaling factor to the simulation temperature. Under each set of simulation conditions, the nonbonded interaction potential energy of the system or the force derived from the potential energy is globally scaled to obtain the scaled molecular dynamics sampling trajectory. Molecular dynamics simulations were performed under each set of simulation conditions, and ligand dissociation events were detected to obtain dissociation time samples containing censored information of non-dissociation. Based on the dissociation time samples, the dissociation rate under each simulation condition pair is estimated; Based on the relationship between the dissociation rate and the unified extrapolation independent variable, the dissociation rate and residence time under the target conditions are extrapolated to obtain the target conditions.

2. The method according to claim 1, characterized in that, Performing molecular dynamics simulations and detecting ligand dissociation events under each set of simulation conditions specifically includes: performing multiple copies of molecular dynamics simulations for each set of simulation conditions, detecting whether the ligands have dissociated based on a preset dissociation criterion, and recording dissociation time samples; when the simulation runs for a preset maximum time and dissociation has not occurred, the corresponding sample is marked as right-censored data.

3. The method according to claim 2, characterized in that, The dissociation criteria include both a distance threshold criterion and a root mean square deviation criterion; wherein, the distance threshold criterion is that the distance between the ligand centroid and the binding site reference point is greater than a preset distance threshold; the root mean square deviation criterion is that the root mean square deviation of the ligand relative to the binding conformation reference conformation is greater than a preset deviation threshold; an event that simultaneously satisfies the above criteria and lasts for a duration not less than a preset duration is defined as a dissociation event.

4. The method according to claim 1, characterized in that, The specific method for estimating the dissociation rate under each simulation condition pair based on dissociation time samples is as follows: assuming that the dissociation process follows first-order dynamics, the maximum likelihood estimation method is used to construct a likelihood function containing dissociated samples and undissociated right-censored samples, and the dissociation rate under each condition pair is solved by maximizing the likelihood function.

5. The method according to claim 4, characterized in that, The likelihood function is: in, Let C be the set of samples that have dissociated, and k be the set of samples that have not dissociated. Let k be the dissociation rate, and t be the observed dissociation time. The simulation duration is for the corresponding replica; the dissociation rate is obtained by maximizing the log-likelihood function.

6. The method according to claim 1, characterized in that, Extrapolating the relationship between dissociation rate and uniform extrapolation independent variable to the target condition specifically involves: performing linear regression fitting with the uniform extrapolation independent variable as the independent variable and the natural logarithm of the dissociation rate as the dependent variable; substituting the overall situation energy scaling factor of 1 and the simulated temperature as the target temperature into the fitted linear relationship to obtain the dissociation rate and residence time under the target condition.

7. The method according to claim 6, characterized in that, The linear regression fitting is based on the generalized Arrhenius equation, and the fitting form is: in, The intercept corresponds to the logarithm of the pre-exponential factor; The slope corresponds to the negative value of the ratio of the energy barrier to the Boltzmann constant; This is the global state energy scaling factor; To simulate temperature.

8. The method according to claim 1, characterized in that, The global scaling condition satisfies any of the following: Alternatively, the force derived from the original potential energy can be scaled: in, The global situation energy scaling factor is... This is the potential energy function of the original non-bonded interactions. The force is derived from the original potential energy.

9. The method according to claim 1, characterized in that, It also includes an uncertainty assessment step: performing bootstrap resampling or cross-validation on the dissociation time samples, and outputting the confidence interval or standard deviation of the dissociation rate and residence time under the target conditions.

10. A ligand-target dissociation kinetic parameter prediction system, characterized in that, include: Data processing module, condition initialization module, scaling and simulation module, rate estimation and extrapolation module, and result output module; The data processing module is used to construct and parameterize the protein-ligand system to obtain the initial configuration and force field parameters for molecular dynamics simulation. The condition initialization module is used to apply constraints to the system to maintain structural stability, set multiple sets of simulation condition pairs, each set of simulation condition pairs includes simulation temperature and global state energy scaling factor, and defines a unified extrapolation independent variable as the ratio of the global state energy scaling factor to the simulation temperature; The scaling and simulation module is used to globally scale the nonbonded interaction potential energy of the system or the force derived from the potential energy under each set of simulation conditions and perform molecular dynamics simulations, detect ligand dissociation events, and obtain dissociation time samples containing censored information of non-dissociation. The rate estimation and extrapolation module is used to estimate the dissociation rate under each simulation condition based on the dissociation time sample, and extrapolate to the target condition based on the relationship between the dissociation rate and the unified extrapolation independent variable to obtain the dissociation rate and residence time under the target condition. The result output module is used to output the dissociation rate and residence time under the target conditions.

Images