Method and system based on combination of adaptive round-trip enhanced sampling prediction and free energy
Through the adaptive round-trip enhancement sampling prediction combined with free energy method, the problem of high computational cost of FEP method is solved, the efficiency and accuracy of drug-target affinity prediction is improved, and more efficient tools are provided for drug design.
Patent Information
- Application Number
- CN202411100166.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-19
AI Technical Summary
In the prior art, the FEP method based on free energy perturbation has a low drug-target affinity prediction efficiency due to high computational cost, and the adaptive control method cannot reduce the computational cost while improving sampling efficiency.
Using the method based on adaptive round trip enhanced sampling prediction combined with free energy, forward and reverse molecular dynamics simulations are performed by splitting the free energy perturbation process, and convergence is dynamically evaluated, simulation time is adjusted, sampling efficiency is improved and Reduce calculation waste.
It significantly improves FEP sampling efficiency, reduces calculation costs, greatly accelerates drug-target affinity prediction, and provides more efficient drug design guidance.
Smart Images

Figure CN118983002A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of drug design, and more specifically, to a method and system for predicting binding free energy based on adaptive round-trip enhanced sampling. Background Art
[0002] High affinity between drugs and targets is crucial for drugs to exert therapeutic effects after binding to targets. Rapid and accurate prediction of drug-target affinity can greatly accelerate the lead compound discovery and optimization stages in the drug development process. At present, various computational methods for drug-target affinity prediction have been developed, including docking scoring based on empirical formulas, semi-empirical molecular mechanics / Poisson-Boltzmann (generalized Bern) surface area (MM / PB(GB)SA) methods based on implicit solvent models, and free energy calculation methods based on free energy perturbation (FEP) and alchemical transformation. Among these methods, the alchemical free energy calculation method based on FEP has a rigorous theoretical basis, is independent of empirical parameters, and explicitly considers the solvation effect and the entropy contribution of ligand-receptor binding. It is a very promising affinity prediction method. However, the widespread application of the FEP method is still limited due to the large amount of computational cost required to ensure the convergence of the calculation results.
[0003] The high computational cost of the FEP method is mainly due to the need for equilibrium samples to achieve accurate and convergent predictions. In order to obtain equilibrium samples, the FEP method needs to perform pre-equilibrium simulations to reach equilibrium. When the relaxation time of the system is long, more simulation time is required, resulting in low computational efficiency. At the same time, in the absence of an adaptive control method, fixed molecular dynamics simulation schemes often set longer pre-equilibrium times and product simulation times for systems of different perturbation sizes to ensure that convergent results are obtained as much as possible, resulting in computational waste.
[0004] In the prior art, enhanced sampling methods can improve sampling efficiency and speed up sampling to achieve a balance, thereby accelerating FEP calculations; adaptive sampling methods can reduce computational waste through customized simulation schemes, thereby reducing computational costs. However, existing enhanced sampling methods often require multiple copies to sample in parallel and frequently exchange sampling data, which reduces the sampling speed; some existing adaptive sampling methods are combined with the aforementioned enhanced sampling methods or do not use enhanced sampling algorithms, and cannot achieve more efficient sampling while adaptively controlling. Some adaptive sampling methods also require a large number of pre-simulations to determine parameters, which adds additional computational costs.
[0005] In summary, how to invent a method that can reduce the calculation cost of binding free energy is a problem that urgently needs to be solved in this technical field. Summary of the invention
[0006] In order to solve the problem of high cost of calculating binding free energy in the prior art, the present invention provides a method and system for predicting binding free energy based on adaptive round-trip enhanced sampling, which has the characteristics of improving sampling efficiency and reducing calculation waste, and accelerating the prediction of drug-target affinity in the field of drug design technology.
[0007] In order to achieve the above-mentioned purpose of the present invention, the technical scheme adopted is as follows: The method for predicting binding free energy based on adaptive round-trip enhanced sampling includes the following specific steps: Construct basic free energy perturbation parameters and adaptive round-trip enhanced sampling parameter files; According to the set segment window step , a window scheme for segmenting free energy perturbation processes; Perform round-trip enhanced sampling by performing one forward molecular dynamics simulation and at least one round-trip molecular dynamics simulation on the window of the first part; Dynamically evaluate the convergence of the last forward molecular dynamics simulation. If it is judged not to be converged, re-run the round-trip molecular dynamics simulation. If it is judged to be converged, perform round-trip enhanced sampling of the next part of the window. Continue to perform round-trip enhanced sampling and dynamic convergence judgment until all parts of the window simulation are completed; Binding free energies were calculated based on window simulations.
[0008] Preferably, the basic free energy perturbation parameters and the adaptive round-trip enhanced sampling parameter files are constructed, specifically: Construct the input file of the parameters required by the molecular dynamics engine and initialize the window scheme and segmented window step size , single simulation time , the total fluctuation allowed during the entire disturbance process , minimum number of round trips, maximum number of round trips.
[0009] Furthermore, according to the set segment window step size , the window scheme for dividing the free energy perturbation process, the specific steps are: Calculate the first window The following window Interval , according to the length , add Window to the first part ; The last window of the first part As the first window of the second part , calculation window To the next window Interval , add Window to the second part ; Repeat this step until all window solutions of the free energy perturbation process are divided and R parts of the window are obtained.
[0010] Furthermore, by performing one forward molecular dynamics simulation and at least one round-trip molecular dynamics simulation on the window of the first part, the specific steps are as follows: Perform a forward molecular dynamics simulation of the first part of the window. Start the simulation, and after the simulation is completed, pass the output conformation to the next window Perform molecular dynamics simulation and simulate until the last window of the first part , completed 1 forward molecular dynamics simulation; Reenter and perform a reverse molecular dynamics simulation of the first part of the window, starting from the last window of the first part Start the simulation, and after the simulation is completed, pass the output configuration to the previous window , simulate to the first window in sequence , and then return to perform forward molecular dynamics simulation, thus completing one round-trip molecular dynamics simulation.
[0011] Furthermore, after completing one round-trip molecular dynamics simulation each time, it is also determined whether the set minimum number of round trips has been reached. If so, the convergence judgment tool is further used to dynamically evaluate the convergence of the last forward molecular dynamics simulation. If not, it returns to complete another round-trip molecular dynamics simulation. Use the convergence judgment tool to dynamically evaluate the convergence of the last forward molecular dynamics simulation. If it is judged to be non-convergent, it is necessary to determine whether the set maximum number of round trips has been reached. If not, the round-trip molecular dynamics simulation is repeated. If so, the next part of the window is subjected to round-trip enhanced sampling.
[0012] Furthermore, the convergence judgment tool is used to dynamically evaluate the convergence of the last forward molecular dynamics simulation. The specific steps are as follows: Calculate the step size of the partial window participating in round-trip enhanced sampling :
[0013] Among them, r is the number of the part of the window when the window is split in step S2 According to the total fluctuation , calculate the allowed fluctuation of the partial window participating in round-trip enhanced sampling :
[0014] Use the convergence judgment tool to calculate the convergence index value of the forward simulation; The convergence index is greater than , judge that the simulation has not converged, repeat the round trip enhancement sampling, and use the convergence judgment tool to evaluate the convergence at the end of each round of forward simulation; if the convergence index is less than , judge that the simulation has reached convergence and proceed to simulate the next part of the window.
[0015] Furthermore, the convergence judgment tool is used to calculate the convergence index value of the forward simulation. The specific steps are as follows: Calculated using data accumulated throughout the simulation ; Use the last part of the simulated data; divide the last part of the data into n-1 non-overlapping moving time series, recorded as arrive ; Calculate n Standard deviation of the values:
[0016] Take out 0% to K% of the entire simulation cumulative data to obtain the forward cumulative time series and calculate value; Take out 100% to (100-K)% of the entire simulation cumulative data to obtain the reverse cumulative time series and calculate value; Calculates the absolute value of the difference between forward and backward time series estimates ; Calculate K% of M different proportions of equally spaced data Value, calculate these M The average of the values: .
[0017] Furthermore, we can calculate K% equally spaced M different proportions of data to obtain When the value ~ 80% of the values with equal spacing of endpoints.
[0018] Furthermore, the binding free energy is calculated based on the window simulation, specifically: The analysis program was run to calculate the binding free energy using the Bennet Acceptance Ratio algorithm and to construct a time series convergence plot and a heat map of the phase space overlap between adjacent windows generated using an optimized phase space overlap algorithm.
[0019] A system for predicting binding free energy based on adaptive round-trip enhanced sampling, including a cascaded task control module, a kinetic simulation module, a dynamic convergence monitoring module, and a statistical analysis module; The task control module is used to execute the control of all steps of the method and system for predicting binding free energy based on adaptive round-trip enhanced sampling; including constructing basic free energy perturbation parameters and adaptive round-trip enhanced sampling parameter files; dividing the window scheme of the free energy perturbation process; executing round-trip enhanced sampling simulation; and controlling the execution of other modules as a workflow; The dynamics simulation module is used to perform molecular dynamics simulation according to free energy perturbation parameters; The dynamic convergence monitoring module is used to dynamically evaluate the convergence of the last forward molecular dynamics simulation; The statistical analysis module is used to calculate binding free energy based on window simulation.
[0020] The beneficial effects of the present invention are as follows: The present invention discloses a method and system for predicting binding free energy based on adaptive round-trip enhanced sampling, which comprises the following steps: dividing a window scheme of a free energy perturbation process into several parts; performing forward molecular dynamics simulation and round-trip molecular dynamics simulation on the first part of the window; using a convergence judgment tool to determine the convergence of the last forward molecular dynamics simulation, and if it does not converge, repeating the round-trip enhanced sampling, and performing simulation on the next part of the window if convergence is achieved; repeating the simulation operation on the first part until the window simulation of all parts is completed, and calculating the binding free energy; thereby, the present invention improves the FEP sampling efficiency and adaptively adjusts the simulation time according to the size of different system perturbations, so that the system simulation reaches convergence with less computational cost, greatly accelerating the prediction of binding free energy, and providing guidance for the optimization and modification of drug lead compounds. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 It is a schematic flow chart of the method for predicting binding free energy based on adaptive round-trip enhanced sampling of the present invention.
[0022] Figure 2 It is a specific flow chart of the method for predicting binding free energy based on adaptive round-trip enhanced sampling in the embodiment of the present invention.
[0023] Figure 3 The method and system for predicting binding free energy based on adaptive round-trip enhanced sampling of the present invention is a thrombin system perturbation 3a-1b in the embodiment Convergence plot over time series.
[0024] Figure 4 It is a heat map of phase space overlap between adjacent windows generated by the method and system for predicting binding free energy based on adaptive round-trip enhanced sampling of the present invention for thrombin system perturbation 3a-1b using an optimized phase space overlap algorithm. DETAILED DESCRIPTION
[0025] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.
[0026] Example 1 like Figure 1 As shown, the method for predicting binding free energy based on adaptive round-trip enhanced sampling includes the following specific steps: Based on the input system topology and coordinate files, select the molecular dynamics engine, and construct the basic free energy perturbation parameters and adaptive round-trip enhanced sampling parameter files; According to the set segment window step , a window scheme for segmenting free energy perturbation processes; Through the window of the first part Perform one forward molecular dynamics simulation and at least one round-trip molecular dynamics simulation with round-trip enhanced sampling; Use the convergence judgment tool to dynamically evaluate the convergence of the last forward molecular dynamics simulation. If it is judged not to converge, re-run the round-trip molecular dynamics simulation. If it is judged to converge, perform round-trip enhanced sampling of the next part of the window. Continue to perform round-trip enhanced sampling and dynamic convergence judgment until all parts of the window simulation are completed; Binding free energies were calculated based on window simulations.
[0027] In this embodiment, Figure 2 As shown, selectable molecular dynamics engines include but are not limited to Amber, Open-MM, Gromacs, etc.
[0028] Example 2 In this example, relative binding free energy calculations were performed using a test set of 4 protein targets and 64 ligands.
[0029] In a specific embodiment, the basic free energy perturbation parameters and the adaptive round-trip enhanced sampling parameter files are constructed as follows: Construct the input file of the parameters required by the molecular dynamics engine and initialize the window scheme and segmented window step size , single simulation time , the total fluctuation allowed during the entire disturbance process , minimum number of round trips, maximum number of round trips.
[0030] In this embodiment, the input files are the topology and coordinate files of the protein receptor-small molecule ligand complex system and the small molecule ligand system in the solvent. The molecular dynamics simulation engine selects AMBER, the force field is AMBER ff14SB / GAFF, and the thermodynamic cycle for relative binding free energy calculation is constructed. In this embodiment, the fused state method (CS-FEP) and the one-step method (SS-FEP) are used in parallel, corresponding to two thermodynamic cycle processes respectively. The calculation steps for each thermodynamic process of the two FEP methods are the same. The initial molecular dynamics simulation parameters constructed include arrive Windowing scheme linearly set at intervals of 0.02, segmented window step size , single simulation time , the total fluctuation allowed during the entire perturbation process , the minimum number of round trips is 1 round trip, and the maximum number of round trips is 50 round trips.
[0031] In a specific embodiment, the segmented window step size is set according to the adaptive round-trip enhanced sampling parameter. Window scheme for segmenting free energy perturbation processes ( arrive ), the first part of the window includes arrive The second part of the window includes arrive , and so on, dividing the window scheme into 5 parts.
[0032] Furthermore, by opening the window in the first part Perform one forward molecular dynamics simulation and at least one round-trip molecular dynamics simulation. The specific steps are as follows: Perform a forward molecular dynamics simulation of the first part of the window. Start the simulation, and after the simulation is completed, pass the output conformation to the next window Perform molecular dynamics simulation and simulate until the last window of the first part , completed 1 forward molecular dynamics simulation; Reenter and perform a reverse molecular dynamics simulation of the first part of the window, starting from the last window of the first part Start the simulation, and after the simulation is completed, pass the output configuration to the previous window , simulate to the first window in sequence , and then return to perform forward molecular dynamics simulation, thus completing one round-trip molecular dynamics simulation.
[0033] In this embodiment, the first part of the window Forward molecular dynamics simulation, from the first window Start simulation, simulation time is After the simulation is completed, the final conformation is passed to the next window , simulate in sequence until the last window of the first part is completed . To the window When performing molecular dynamics simulation, the molecular dynamics trajectory of the current window is sampled. Based on the molecular dynamics trajectory of the current window, the force field files of multiple adjacent windows are used for calculation and the potential energy between adjacent windows is recorded. U , which is used to subsequently draw the phase space overlap heat map of adjacent windows.
[0034] In this embodiment, the reverse molecular dynamics simulation of the first part of the window is performed, and the last window of the first part is Start simulating the first window in the same way as in the previous step. , and return to perform the forward molecular dynamics simulation described in the previous step to complete a round-trip enhanced sampling.
[0035] In a specific embodiment, Figure 2 As shown, after completing one round-trip molecular dynamics simulation each time, it is also determined whether the set minimum number of round trips has been reached. If so, the convergence judgment tool is further used to dynamically evaluate the convergence of the last forward molecular dynamics simulation. If not, it returns to continue to complete one round-trip molecular dynamics simulation. Use the convergence judgment tool to dynamically evaluate the convergence of the last forward molecular dynamics simulation. If it is judged to be non-convergent, it is necessary to determine whether the set maximum number of round trips has been reached. If not, the round-trip molecular dynamics simulation is repeated. If so, the next part of the window is subjected to round-trip enhanced sampling.
[0036] In a specific embodiment, the convergence of the last forward molecular dynamics simulation is dynamically evaluated using a convergence judgment tool, and the specific steps are as follows: Calculate the step size of the partial window participating in round-trip enhanced sampling :
[0037] Among them, r is the number of the part of the window when the window is split in step S2 According to the total fluctuation , calculate the allowed fluctuation of the partial window participating in round-trip enhanced sampling :
[0038] Use the convergence judgment tool to calculate the convergence index value of the forward simulation; Convergence index is greater than , judge that the simulation has not converged, repeat the round trip enhancement sampling, and use the convergence judgment tool to evaluate the convergence at the end of each round of forward simulation; if the convergence index is less than , judge that the simulation has reached convergence and proceed to simulate the next part of the window.
[0039] In this embodiment, the step length of the first part of the window participating in the round-trip enhanced sampling is calculated ; The total fluctuation allowed for the entire perturbation process according to the adaptive round-trip enhanced sampling parameter setting , calculate the allowed fluctuation of the partial window participating in round-trip enhanced sampling : .
[0040] In a specific embodiment, the convergence index value of the forward simulation is calculated using a convergence judgment tool, and the specific steps are: Computes the moving series estimate stability (MES) using data accumulated throughout the simulation ; Use the last 50% of the simulation data, divided into 4 parts of non-overlapping moving time series accounting for 10% of the total cumulative data, and calculate arrive ; Calculate 5 Standard deviation of the values:
[0041] Take out 0% to K% of the entire simulation cumulative data to obtain the forward cumulative time series and calculate value; Take out 100% to (100-K)% of the entire simulation cumulative data to obtain the reverse cumulative time series and calculate value; Calculates the absolute value of the difference between forward and backward time series estimates ; Calculate K% with endpoints 10 different proportions of data with equal distances between ~ 80% Value, calculate these 10 The average of the values: .
[0042] In this example, the calculated MES value and FREC value are compared with If the convergence index is greater than , it means that the simulation has not reached convergence, and a second round-trip enhanced sampling consisting of reverse and forward molecular dynamics simulations is performed. When the forward simulation is completed, repeat this step and use the convergence judgment tool to evaluate the convergence; If the convergence index is less than , it means that the simulation of the first part of the window has reached convergence, and the second part of the window is carried out arrive simulation.
[0043] In one embodiment, the second part window starts The specific steps of the simulation are the same as those of the simulation of the first part of the window, including round-trip enhanced sampling and dynamic convergence monitoring steps. When convergence is achieved, the next part of the simulation is carried out; until the simulation of all parts of the window is completed; when the simulation of the last part of the window reaches convergence, the molecular dynamics simulation program is stopped; the analysis program is run, the binding free energy is calculated by the Bennet Acceptance Ratio algorithm, and a time series convergence graph is constructed, and the time series convergence graph is used to judge the energy convergence of each simulation window; a heat map of the phase space overlap between adjacent windows is generated using an optimized phase space overlap algorithm, and the heat map of the phase space overlap between adjacent windows is used to judge whether the phase space overlap of adjacent windows under enhanced sampling is sufficient.
[0044] In this embodiment, RBFE calculations are performed three times in parallel for each pair of ligands. The simulation time is adaptively adjusted based on convergence under the control of the method (CAR) described in this patent. The simulation time consumed for each calculation is recorded (unit: ns). The average simulation time for executing all calculations is shown in Table 1: Table 1
[0045] Table 1 shows the average time taken to calculate the free energy of four protein systems by combining the method (CAR) described in this patent with the CS-FEP method and the SS-FEP method, including the time taken for the complex system composed of protein receptor-small molecule ligand and the small molecule ligand system in the solvent, and the total time composed of the sum of the two. It can be seen from Table 1 that the method (CAR) described in this patent can customize the simulation time for different systems, and the calculation time for the complex side with large perturbations and difficult convergence is significantly higher than the calculation time for the ligand side in the solvent. It shows that the method of the present invention can effectively and adaptively adjust the simulation time for systems with different perturbations. In addition, the overall average simulation time of the two FEP calculation methods with different thermodynamic cycles is 16.29ns and 15.23ns, respectively. Compared with the traditional FEP calculation method that is widely used and takes 120~200ns, the speed of the method of the present invention is increased by 7 to 13 times. It shows that the method of the present invention can efficiently enhance sampling and customize simulation time, greatly improve calculation efficiency, and reduce calculation cost.
[0046] In this embodiment, RBFE calculations were performed three times in parallel for each pair of ligands, and the average of the three calculation results was taken to obtain the ΔΔG value, and then the loop closure convergence tool was used to obtain the final ΔG value. The comparison of the obtained test results with AMBER20-SSC (2) is shown in Table 2: Table 2
[0047] Table 2 shows the mean absolute error (MAE), root mean square error (RMSE) and correlation coefficient of the method described in this patent (CAR) combined with the CS-FEP method and the SS-FEP method and the AMBER20-SSC (2) method for predicting ΔG of the four protein systems. R 2 From Table 2, it can be seen that after adopting the method described in this patent (CAR), compared with the results of AMBER20-SSC (2), the correlation of the prediction results of most systems is improved when using a shorter simulation time, and the overall prediction ability remains the same. This shows that the method of the present invention has a high accuracy.
[0048] The method described in this patent (CAR) is selected to calculate a perturbation in the Thrombin system (3a-1b), and the time series convergence diagram is constructed using the analysis program. As the time series converges, Figure 3 As shown in the figure, the red line represents the forward cumulative time series estimate, the blue line represents the reverse cumulative time series estimate, the green line represents the moving time series estimate, and the bright yellow line is the final prediction result calculated using the Bennet Acceptance Ratio (BAR) method. The energy curves of the forward time series estimate and the reverse time series estimate converge rapidly to the BAR calculation result from the upper and lower ends, respectively. The moving time series estimate shows Fluctuations over time remain stable.
[0049] The method (CAR) described in this patent is selected to calculate a perturbation (3a-1b) in the Thrombin system, and the optimized phase space overlap algorithm is used to generate a heat map of the phase space overlap between adjacent windows. The phase space overlap heat map is as follows: Figure 4 As shown, the color of each small grid is lighter, and the phase space overlap between adjacent windows is greater than 0.8, indicating that the phase space overlap between adjacent windows is high, the simulated disturbance is small and reaches equilibrium, and the energy calculation results are relatively convergent and reliable.
[0050] It can be seen that the test set of 4 targets and 64 ligands is used as an example, and the method proposed in the present invention is used to simulate and calculate the example system. The mean absolute error (MAE), root mean square error (RMSE) and correlation coefficient R2 of the ΔG of the 4 protein systems obtained are close to the AMBER20-SSC (2) method, and the simulation time is shortened by several to more than ten times. This shows that the method and system for predicting binding free energy based on adaptive round-trip enhanced sampling described in the present invention can customize the simulation time according to the disturbance size of different systems and thermodynamic processes, and effectively accelerate the simulation through the enhanced sampling method. At the same time, the method described in the present invention guarantees convergence during the calculation process, and quickly obtains accurate and convergent binding free energy prediction results, which greatly reduces the amount of calculation, and is expected to promote the large-scale application of the FEP method in the field of drug design to guide the design of drug lead compounds and structural modification.
[0051] Example 3 A system for predicting binding free energy based on adaptive round-trip enhanced sampling, including a cascaded task control module, a kinetic simulation module, a dynamic convergence monitoring module, and a statistical analysis module; The task control module is used to execute the control of all steps of the method and system for predicting binding free energy based on adaptive round-trip enhanced sampling; including constructing basic free energy perturbation parameters and adaptive round-trip enhanced sampling parameter files; dividing the window scheme of the free energy perturbation process; executing round-trip enhanced sampling simulation; and controlling the execution of other modules as a workflow; The dynamics simulation module is used to perform molecular dynamics simulation according to free energy perturbation parameters; The dynamic convergence monitoring module is used to dynamically evaluate the convergence of the last forward molecular dynamics simulation using a convergence judgment tool; The statistical analysis module is used to calculate binding free energy based on window simulation.
[0052] Obviously, the above embodiments of the present invention are only examples for clearly explaining the present invention, and are not intended to limit the implementation methods of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the claims of the present invention.
Claims
1. A method for predicting binding free energy based on adaptive round-trip enhanced sampling, characterized in that: The specific steps include: Construct basic free energy perturbation parameters and adaptive round-trip enhanced sampling parameter files; According to the set segment window step , a window scheme for segmenting free energy perturbation processes; Perform round-trip enhanced sampling by performing one forward molecular dynamics simulation and at least one round-trip molecular dynamics simulation on the window of the first part; Dynamically evaluate the convergence of the last forward molecular dynamics simulation. If it is judged not to be converged, re-run the round-trip molecular dynamics simulation. If it is judged to be converged, perform round-trip enhanced sampling of the next part of the window. Continue to perform round-trip enhanced sampling and dynamic convergence judgment until all parts of the window simulation are completed; Binding free energies were calculated based on window simulations.
2. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 1, characterized in that: Construct the basic free energy perturbation parameters and adaptive round-trip enhanced sampling parameter files, specifically: Construct the input file of the parameters required by the molecular dynamics engine and initialize the window scheme and segmented window step size , single simulation time , the total fluctuation allowed during the entire disturbance process , minimum number of round trips, maximum number of round trips.
3. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 2, characterized in that: According to the set segment window step , the window scheme for dividing the free energy perturbation process, the specific steps are: Calculate the first window The following window Interval , according to the length , add Window to the first part ; The last window of the first part As the first window of the second part , calculation window To the next window Interval , add Window to the second part ; Repeat this step until all window solutions of the free energy perturbation process are divided and R parts of the window are obtained.
4. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 3, characterized in that: Perform one forward molecular dynamics simulation and at least one round-trip molecular dynamics simulation on the window in the first part. The specific steps are as follows: Perform a forward molecular dynamics simulation of the first part of the window. Start the simulation, and after the simulation is completed, pass the output conformation to the next window Perform molecular dynamics simulation and simulate until the last window of the first part , completed 1 forward molecular dynamics simulation; Reenter and perform a reverse molecular dynamics simulation of the first part of the window, starting from the last window of the first part Start the simulation, and after the simulation is completed, pass the output configuration to the previous window , simulate to the first window in sequence , and then return to perform forward molecular dynamics simulation, thus completing one round-trip molecular dynamics simulation.
5. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 4, characterized in that: After completing a round-trip molecular dynamics simulation each time, it is also determined whether the set minimum number of round trips has been reached. If so, the convergence judgment tool is further used to dynamically evaluate the convergence of the last forward molecular dynamics simulation. If not, it returns to complete another round-trip molecular dynamics simulation. Use the convergence judgment tool to dynamically evaluate the convergence of the last forward molecular dynamics simulation. If it is judged to be non-convergent, it is necessary to determine whether the set maximum number of round trips has been reached. If not, the round-trip molecular dynamics simulation is repeated. If so, the next part of the window is subjected to round-trip enhanced sampling.
6. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 5, characterized in that: Use the convergence judgment tool to dynamically evaluate the convergence of the last forward molecular dynamics simulation. The specific steps are as follows: Calculate the step size of the partial window participating in round-trip enhanced sampling : Among them, r is the number of the part of the window when the window is split in step S2 According to the total fluctuation , calculate the allowed fluctuation of the partial window participating in round-trip enhanced sampling : Use the convergence judgment tool to calculate the convergence index value of the forward simulation; Convergence index is greater than , judge that the simulation has not converged, repeat the round trip enhancement sampling, and use the convergence judgment tool to evaluate the convergence at the end of each round of forward simulation; if the convergence index is less than , judge that the simulation has reached convergence and proceed to simulate the next part of the window.
7. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 6, characterized in that: Use the convergence judgment tool to calculate the convergence index value of the forward simulation. The specific steps are: Calculated using data accumulated throughout the simulation ; Use the last part of the simulated data; divide the last part of the data into n-1 non-overlapping moving time series, recorded as arrive ; Calculate n Standard deviation of the values: Take out 0% to K% of the entire simulation cumulative data to obtain the forward cumulative time series and calculate value; Take out 100% to (100-K)% of the entire simulation cumulative data to obtain the reverse cumulative time series and calculate value; Calculates the absolute value of the difference between forward and backward time series estimates ; Calculate K% of M different proportions of equally spaced data Value, calculate these M The average of the values: 。 8. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 7, characterized in that: Calculate K% of M different proportions of equally spaced data When the value ~ 80% of the values with equal spacing of endpoints.
9. The method for predicting binding free energy based on adaptive round-trip enhanced sampling according to claim 1, characterized in that: The binding free energy is calculated based on the window simulation, specifically: The analysis program was run to calculate the binding free energy using the Bennet Acceptance Ratio algorithm and to construct a time series convergence plot and a heat map of the phase space overlap between adjacent windows generated using an optimized phase space overlap algorithm.
10. A system for predicting binding free energy based on adaptive round-trip enhanced sampling, characterized in that: It includes cascaded task control module, dynamics simulation module, dynamic convergence monitoring module and statistical analysis module; The task control module is used to execute the control of all steps of the method and system for predicting binding free energy based on adaptive round-trip enhanced sampling; including constructing basic free energy perturbation parameters and adaptive round-trip enhanced sampling parameter files; dividing the window scheme of the free energy perturbation process; executing round-trip enhanced sampling simulation; and controlling the execution of other modules as a workflow; The dynamics simulation module is used to perform molecular dynamics simulation according to free energy perturbation parameters; The dynamic convergence monitoring module is used to dynamically evaluate the convergence of the last forward molecular dynamics simulation; The statistical analysis module is used to calculate binding free energy based on window simulation.
