WO2022082740A1 - 基于虚拟现实的药物设计云计算流程控制系统及其方法 - Google Patents

基于虚拟现实的药物设计云计算流程控制系统及其方法 Download PDF

Info

Publication number
WO2022082740A1
WO2022082740A1 PCT/CN2020/123285 CN2020123285W WO2022082740A1 WO 2022082740 A1 WO2022082740 A1 WO 2022082740A1 CN 2020123285 W CN2020123285 W CN 2020123285W WO 2022082740 A1 WO2022082740 A1 WO 2022082740A1
Authority
WO
WIPO (PCT)
Prior art keywords
module
drug design
task
virtual reality
calculation
Prior art date
Application number
PCT/CN2020/123285
Other languages
English (en)
French (fr)
Inventor
方磊
马健
张佩宇
王辉
温书豪
赖力鹏
Original Assignee
深圳晶泰科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳晶泰科技有限公司 filed Critical 深圳晶泰科技有限公司
Priority to PCT/CN2020/123285 priority Critical patent/WO2022082740A1/zh
Publication of WO2022082740A1 publication Critical patent/WO2022082740A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
    • G16B15/00—ICT specially adapted for analysing two-dimensional or three-dimensional molecular structures, e.g. structural or functional relations or structure alignment
    • G16B15/30—Drug targeting using structural data; Docking or binding prediction

Definitions

  • the invention belongs to the field of drug design based on artificial intelligence, and in particular relates to a cloud computing process control method for drug design based on virtual reality technology.
  • Typical drug design methods or software include: Maestro software, MOE software, Amber molecular dynamics software, Autudock molecular docking software, Pymol display software, VMD display software, etc.
  • Limited process integration Open source software can generally complete specific computing tasks independently, but if you need to integrate these open source software to complete more complex functions, you need to write scripts to integrate multiple software, and different software has different inputs. The output format, etc., will be difficult to integrate.
  • Commercial software is more functional and has relatively complete processes, but software costs are generally more expensive. Because it is a commercial software, if you want to carry out secondary development, there is also a certain difficulty in accessing.
  • Visualization is mainly based on the 2D or pseudo-3D display of the display. It is difficult to display some complex three-dimensional structures, such as complexes of small molecules and targets.
  • the present invention provides a drug design cloud computing process control method based on virtual reality technology.
  • the present invention adopts the following scheme:
  • the present invention is mainly a drug design cloud computing process control system based on artificial intelligence and virtual reality technology, the system includes a drug design module group, an interaction and scheduling module and a virtual reality drug design module;
  • the drug design module group includes a verification module, a database module, a filter module, a preprocessing module, a method module, and an analysis module;
  • the drug design module group includes typical drug design methods and software, specifically including Maestro software, MOE software , Amber molecular dynamics software, Autudock molecular docking software, Pymol display software, VMD display software are displayed in the form of modules, and users can freely select and combine the corresponding modules to perform calculations according to project requirements;
  • the method modules include a molecular docking module (Molecular docking), a molecular dynamics module (Molecular Dynamics), a binding free energy calculation module (MMGBSA/MMPBSA), a free energy perturbation module (FEP), a pharmacophore module (Pharmacophore) , artificial intelligence molecule generation module, ADMET prediction module.
  • Molecular docking molecular docking
  • Molecular Dynamics molecular dynamics module
  • MMGBSA/MMPBSA binding free energy calculation module
  • FEP free energy perturbation module
  • pharmacophore module Pharmacophore
  • ADMET prediction module artificial intelligence molecule generation module
  • Described interaction and scheduling module includes Jupyter unit and Majorana task scheduling platform;
  • the Jupyter unit is an interface for interacting with the user, and the user can complete the corresponding drug design operation through the Jupyter unit issuing instructions to call various modules according to the project requirements; or call the virtual reality drug design module to make the entire drug design.
  • the operation is completed in the visual operation mode of virtual reality, and the result of the operation is displayed;
  • the Majorana task scheduling platform is an internally developed scheduling system, and its function is to call the corresponding task module in the drug design module group for calculation by receiving user instructions sent by the Jupyter unit, and submit the calculation task to the designated cloud computing. platform, and check the status of computing tasks; the status of checking computing tasks mainly includes whether the task is submitted successfully, the number of CPUs and GPUs occupied by the task, the status of the task (starting, executing, completing, etc.), task logs, output files, etc.
  • the cloud computing platforms include Amazon Cloud, Tencent Cloud, Facebook Cloud, and the Majorana task scheduling platform has built-in task execution scripts and task monitoring scripts that connect to various cloud computing platforms, and users can easily achieve cross-platform cloud computing requirements according to computing requirements;
  • the virtual reality drug design module includes virtual reality hardware equipment: a display, virtual reality glasses, a virtual reality gesture recognizer and a computer mouse.
  • virtual reality hardware equipment By operating the virtual reality hardware equipment, the drug design cloud computing process control method can be operated through the visualization of virtual reality. The mode operation is completed, and the results can be displayed in time after the calculation is completed.
  • the visualization operation described in the virtual reality drug design module can be operated in a 2D or 3D manner
  • the display results described in the virtual reality drug design module can be displayed in 2D or 3D;
  • the virtual reality drug design module can independently choose whether to operate in a virtual reality manner according to the needs of the user;
  • the cloud computing process control method for drug design based on virtual reality technology includes the following processes:
  • Small-scale verification first calls the verification module, including four sub-modules of preprocessing, database, method and analysis.
  • the required sub-modules can be selected according to the actual needs of the project, and the corresponding parameters can be obtained by selecting appropriate sub-modules for analysis and verification. , and record it.
  • the small-scale verification is a simple version of the process, the main purpose is to verify the performance of the small-scale calculation method in the corresponding research system, and the parameters after verification will be applied to the formal screening process; specific verification sub-modules It is mainly determined according to the actual needs of the project.
  • the filtering module to perform preliminary filtering and screening on the small molecule database, the main content is to filter out some molecules with poor drugability through the five rules of drugability;
  • the preprocessing module is called to process the target and small molecule data read by the database module, and the preprocessing work before formal calculation is performed.
  • General pretreatment is mainly divided into target pretreatment and small molecule pretreatment.
  • the preprocessing of the target mainly includes: repairing the missing residues of the target in the crystal structure, hydrogenation, charging, protonation of amino acids, etc., and save it in the standard PDB format;
  • the preprocessing method of the small molecule is as follows: each small molecule filtered in the database should be preprocessed, mainly including the conversion of planar two-dimensional structure to three-dimensional structure, lowest energy conformation search, PKa calculation, hydrogenation, charge addition, etc. Save to a standard format, such as SDF format.
  • the method module is called, and the corresponding module is called for calculation and analysis according to the actual calculation requirements.
  • the method module includes a molecular docking module (Molecular docking), a molecular dynamics module (Molecular Dynamics), and a combined free energy calculation module (MMGBSA/MMPBSA), Free energy perturbation module (FEP), pharmacophore module (Pharmacophore), artificial intelligence molecule generation module, ADMET prediction module and other main computational analysis modules;
  • Majorana task scheduling platform submit computing tasks to the specified cloud platform (Amazon Cloud, Tencent Cloud, Facebook Cloud, etc.) according to the computing requirements of the method module, and view the computing task status.
  • Majorana is an internally developed scheduling system, which mainly includes whether the task is successfully submitted, the number of CPU and GPU occupied by the task, the status of the task (starting, executing, completing, etc.), task logs, and output files.
  • the display results are mainly related to which methods are called in the method module.
  • the main contents include the binding mode, molecular dynamics trajectory, compound structure, compound properties, etc., using charts and animations to show the stability of these compound structures and the relationship between small molecule targets. key interactions, energy values, etc.
  • molecular docking and free energy calculation as an example.
  • the system can display the binding mode calculated by the software, mark the interaction between the key small molecule and the target, such as hydrogen bonding, hydrophobic interaction, polar interaction, etc., and can output the image in PNG format , GIF animation and other standard formats to facilitate future review and display.
  • the free energy calculation can display the calculated free energy value, error, dynamic trajectory stability, etc. It can also be output as PNG format pictures, GIF animations, etc. for future review and display.
  • Steps (b)-(g) will be executed cyclically, and whether to terminate depends on the preset indicators, such as predicting 100 compounds with better activity, predicting molecules with energy values in a certain range, etc., if the criteria are met, the cycle can be stopped. round calculation, etc.
  • the complete process of the cloud computing process control method for drug design based on virtual reality technology includes the above steps (a)-(h). Users can also freely combine steps (a)-(h) according to the needs of the project.
  • FIG. 1 is a schematic flowchart of a cloud computing process control method for drug design based on virtual reality technology.
  • Fig. 2 is the specific content of the verification module of the drug design cloud computing process control method based on virtual reality technology
  • Fig. 3 is the specific content of the method module of the drug design cloud computing process control method based on virtual reality technology
  • Fig. 4 is the correlation diagram of biological experiment test-calculation prediction activity of Example 1
  • FIG. 5 is a comparison of the CPU running time of the traditional process and the Majorana task scheduling system in the second embodiment.
  • FIG. 6 is a diagram showing the results of the drug design method in Example 3.
  • FIG. 6 is a diagram showing the results of the drug design method in Example 3.
  • FIG. 7 is a graph showing the results of molecular description in Example 4.
  • FIG. 8 is a diagram showing the results of the combination mode in Example 4.
  • a drug design cloud computing process control system based on artificial intelligence and virtual reality technology, the system includes a drug design module group, an interaction and scheduling module and a virtual reality drug design module;
  • the drug design module group includes a verification module, a database module, a filter module, a preprocessing module, a method module, and an analysis module;
  • the drug design module group includes typical drug design methods and software, specifically including Maestro software, MOE software , Amber molecular dynamics software, Autudock molecular docking software, Pymol display software, VMD display software are displayed in the form of modules, and users can freely select and combine the corresponding modules to perform calculations according to project requirements;
  • the method modules include a molecular docking module (Molecular docking), a molecular dynamics module (Molecular Dynamics), a binding free energy calculation module (MMGBSA/MMPBSA), a free energy perturbation module (FEP), a pharmacophore module (Pharmacophore) , artificial intelligence molecule generation module, ADMET prediction module.
  • Molecular docking molecular docking
  • Molecular Dynamics molecular dynamics module
  • MMGBSA/MMPBSA binding free energy calculation module
  • FEP free energy perturbation module
  • pharmacophore module Pharmacophore
  • ADMET prediction module artificial intelligence molecule generation module
  • Described interaction and scheduling module includes Jupyter unit and Majorana task scheduling platform;
  • the Jupyter unit is an interface for interacting with the user, and the user can complete the corresponding drug design operation through the Jupyter unit issuing instructions to call various modules according to the project requirements; or call the virtual reality drug design module to make the entire drug design.
  • the operation is completed in the visual operation mode of virtual reality, and the result of the operation is displayed;
  • the Majorana task scheduling platform is an internally developed scheduling system, and its function is to call the corresponding task module in the drug design module group for calculation by receiving user instructions sent by the Jupyter unit, and submit the calculation task to the designated cloud computing. platform, and check the status of computing tasks; the status of checking computing tasks mainly includes whether the task is submitted successfully, the number of CPUs and GPUs occupied by the task, the status of the task (starting, executing, completing, etc.), task logs, output files, etc.
  • the cloud computing platforms include Amazon Cloud, Tencent Cloud, Facebook Cloud, and the Majorana task scheduling platform has built-in task execution scripts and task monitoring scripts that connect to various cloud computing platforms, and users can easily achieve cross-platform cloud computing requirements according to computing requirements;
  • the virtual reality drug design module includes virtual reality hardware equipment: a display, virtual reality glasses, a virtual reality gesture recognizer and a computer mouse
  • the drug design cloud computing can be made by operating the virtual reality hardware equipment.
  • the process control method is completed through the visual operation mode of virtual reality, and the results can be displayed in time after the calculation is completed. This can bring better visual effects to users than traditional drug design software, as well as understand the molecular mechanism, and requires less professional background knowledge of users.
  • the small molecules in the database are called from the database module, the number is about 10 million, and then the preprocessing module is called to preprocess these molecules. Considering that the goal of the project is to screen the hit compounds, the calling method is selected.
  • the module validated the molecular docking method and found suitable screening parameters.
  • FIG. 4 shows the correlation diagram between the small molecule activity predicted by the calculation module of this embodiment and the experimental activity after it is recommended to the biological testing team. It can be seen from Fig. 4 that there is a high correlation between the calculated predicted activity and the experimental activity, which shows the reliability of the calculation method of this embodiment.
  • FIG. 5 shows the comparison of the CPU running time of the traditional drug design process and the Majorana task scheduling system. It can be seen that the use of the Majorana task scheduling system only accounts for 40% of the running time of the traditional process, which means that the streamlined and modular operation can significantly improve the work. efficiency.
  • Example 1 Some candidate hit compounds with good experimental properties were screened in Example 1, and the cloud computing process control method for drug design based on virtual reality technology was used to optimize this batch of hit compounds, and finally lead compounds were obtained.
  • the designed molecule and target structure are imported into the system, and the virtual reality drug design module is called to carry out the drug design based on virtual reality according to the combination mode.
  • Analyze the key binding mode between the target and the hit compound directly according to the binding mode displayed by the virtual reality call the corresponding method module according to the actual calculation requirements, then call the Majorana task scheduling platform, and submit the calculation task according to the calculation requirements of the method module Go to the cloud platform, and use the computing platform to conduct cloud computing evaluation in the background, and give feedback to team members in a timely manner.
  • the team members Based on the computational evaluation feedback, and with the help of virtual reality technology, the team members again watched the binding patterns and redesigned the molecule before submitting the backend computational evaluation again. After multiple rounds of iterations, reasonable candidate lead compounds are found. All data and scheduling in the process will be recorded in the database, which is convenient for retrospective review in the future.
  • the visualization of the drug design flow of the present embodiment is shown. Through this process, fragment severing and growth processes of molecules, as well as transition paths between different backbones, can be observed. From this visualization process, the step-by-step optimization process of the target molecule can be understood. Compared with other traditional methods, this method is firstly streamlined and modularized, which facilitates the optimization of lead compounds and improves efficiency. In this case, the drug design module of virtual reality is also used, which facilitates the analysis and interpretation of the results; at the same time, because of the use of the Majorana task scheduling system, the tracking and analysis of the calculation are more convenient, and the efficiency is further improved.
  • Example 2 From the operations of Example 1 and Example 2, some candidate lead compounds with good experimental properties were obtained. Using the cloud computing process control method for drug design based on virtual reality technology, call analysis to display this batch of calculation results, and display the relevant results of the lead compounds including binding mode, interaction, energy value, stability, principal component analysis, etc.
  • Fig. 7 shows the description of the molecular structure by the calculation method of this embodiment, mainly by means of molecular fingerprints and key-value pairs, which can quickly obtain the structural features of a large number of molecules, and digitally representation and storage.
  • Figure 8 shows the binding mode of molecules and proteins. It can be seen from the figure that not only shows the basic interaction between protein and ligand, but also further clustering the interaction types to obtain a pharmacophore-like analysis , to gain more insights into protein-ligand binding interactions and driving forces. Compared with other traditional methods, this method is firstly streamlined and modularized, which facilitates the display of results.

Landscapes

  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Pharmacology & Pharmacy (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Medicinal Chemistry (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biotechnology (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种基于虚拟现实的药物设计云计算流程控制系统及其方法，包括药物设计模块组、交互和调度模块和虚拟现实药物设计模块；所述药物设计模块组包括验证模块、数据库模块、过滤模块、预处理模块、方法模块、分析模块；所述药物设计模块组以模块的形式展现，用户根据项目需求自由选择和组合相应的模块来进行计算；所述方法模块包括分子对接模块、分子动力学模块、结合自由能计算模块、自由能微扰模块、药效团模块、人工智能分子生成模块，ADMET预测模块；所述交互和调度模块包括Jupyter单元和Majorana任务调度平台。实现了高通量、高并行的基于虚拟现实技术的药物设计云计算任务的跨平台调度。

Description

基于虚拟现实的药物设计云计算流程控制系统及其方法 技术领域
本发明属于基于人工智能的药物设计领域，尤其涉及基于虚拟现实技术的药物设计云计算流程控制方法。
背景技术
传统的计算机辅助药物设计方法，主要是基于结构的药物设计方法已经广泛的应用于新药研发领域。
这些传统的药物设计都是基于开源或者商业软件来完成，分为基于Linux平台或者基于Windows平台。
典型的药物设计方法或者软件有：Maestro软件，MOE软件，Amber分子动力学软件，Autudock分子对接软件，Pymol显示软件，VMD显示软件等。
这些药物设计方法虽然广泛使用，但是有如下缺陷：
1.流程整合有限：开源软件一般都能够独立完成特定的计算任务，但是如果需要整合这些开源软件来完成更为复杂的功能，需要自行编写脚本整合多个软件，而且不同的软件有不同的输入输出格式等，整合会存在一定的难度。商业软件功能上更加完善，而且有相对完善的流程，但是软件费用一般也较为昂贵。因为是商业软件，如果想进行二次开发，也存在一定的接入难度。
2.不能跨平台调度：当前几乎所有的软件，如果需要使用云计算平台，都需要先手动登录、编写任务脚本，再提交到各自的平台上计算。
3.可视化主要是基于显示器的2D或者伪3D显示，对一些复杂的三维结构，如小分子和靶点的复合物，存在一定的展示难度。
发明内容
为了解决以上技术问题，本发明提供一种基于虚拟现实技术的药物设计云计算流程控制方法。
为实现上述目的，本发明所采用以下方案：
本发明主要是基于人工智能和虚拟现实技术的药物设计云计算流程控 制系统，该系统包括药物设计模块组、交互和调度模块和虚拟现实药物设计模块；
所述的药物设计模块组包括验证模块、数据库模块、过滤模块、预处理模块、方法模块、分析模块；所述的药物设计模块组包括典型的药物设计方法和软件，具体包括Maestro软件，MOE软件，Amber分子动力学软件，Autudock分子对接软件，Pymol显示软件，VMD显示软件，以模块的形式展现，用户可以根据项目需求自由选择和组合相应的模块来进行计算；
所述的方法模块包括分子对接模块(Molecular docking)，分子动力学模块(Molecular Dynamics)，结合自由能计算模块(MMGBSA/MMPBSA)，自由能微扰模块(FEP)，药效团模块(Pharmacophore)，人工智能分子生成模块，ADMET预测模块。
所述的交互和调度模块包括Jupyter单元和Majorana任务调度平台；
所述的Jupyter单元是与用户进行交互的界面，用户可以项目的需求通过Jupyter单元发出调用各种模块的指令完成进行相应的药物设计操作；也可以通过调用虚拟现实药物设计模块，使整个药物设计操作以虚拟现实的可视化操作方式完成，并展示操作的结果；
所述的Majorana任务调度平台是一个内部开发的调度系统，其功能是通过接收Jupyter单元发出的用户指令在药物设计模块组中调取相应的任务模块来计算，将计算任务提交到指定的云计算平台，并查看计算任务状态；所述的查看计算任务状态主要包括任务是否提交成功，任务占用CPU，GPU数，任务的状态(启动，执行，完成等)，任务日志，输出文件等。
所述的云计算平台包括亚马逊云，腾讯云，阿里云，Majorana任务调度平台内置了对接各个云计算平台的任务执行脚本和任务监控脚本，用户可以根据计算的需求轻松实现跨平台云计算需求；
所述的虚拟现实药物设计模块包括虚拟现实硬件设备：显示器、虚拟现实眼镜，虚拟现实姿势识别器和电脑鼠标，通过操作虚拟现实硬件设备可以使药物设计云计算流程控制方法通过虚拟现实的可视化操作方式操作完成，并可以等计算完成后及时显示结果。
所述的虚拟现实药物设计模块中所述的可视化操作可以以2D或者3D 的方式进行操作；
所述的虚拟现实药物设计模块中所述的显示结果可以以2D或者3D的方式进行展示；
所述的虚拟现实药物设计模块可以根据用户的需求自主选择是否采用虚拟现实的方式来操作；
基于虚拟现实技术的药物设计云计算流程控制方法，包括以下过程：
(a)小规模的验证
小规模的验证首先调用验证模块，包括预处理，数据库，方法和分析四个子模块，可以根据项目的实际需求选择所需要的子模块，通过选用合适的子模块进行分析，验证，得到相应的参数，并记录下来。
所述的小规模的验证是一个简单版本的流程，主要目的是小规模的验证计算方法在相应研究体系中的表现，验证完成后的参数会应用到正式的筛选流程当中；具体验证的子模块主要根据项目的实际需求来确定。
(b)数据读入
调用数据库模块读取原始的数据和结构，主要是从数据库中将靶点和小分子化合物的结构信息读入到工作流程中；
(c)过滤筛选
调用过滤模块对小分子数据库进行初步的过滤筛选，主要内容是通过成药性五规则过滤掉一些成药性不佳的分子；
(d)预处理
调用预处理模块处理数据库模块读取的靶点和小分子数据，进行正式计算前的预处理工作。一般预处理主要分为靶点的预处理和小分子的预处理。靶点的预处理主要包括：晶体结构中靶点缺失残基的修补，加氢，加电荷，氨基酸的质子化等，并保存成标准的PDB格式；
所述小分子的预处理方法为：数据库过滤后的每个小分子都要进行预处理，主要包括平面二维结构转三维结构，最低能量构象搜索，PKa计算，加氢，加电荷等，并保存成标准的格式，如SDF格式。
(e)计算分析
调用方法模块，根据实际的计算需求，调用相应的模块进行计算分析， 方法模块中包括分子对接模块(Molecular docking)，分子动力学模块(Molecular Dynamics)，结合自由能计算模块(MMGBSA/MMPBSA)，自由能微扰模块(FEP)，药效团模块(Pharmacophore)，人工智能分子生成模块，ADMET预测模块等主要计算分析模块；
(f)云计算
调用Majorana任务调度平台，根据方法模块的计算需求，提交计算任务到指定云平台(亚马逊云，腾讯云，阿里云等)并查看计算任务状态。Majorana是内部开发的一个调度系统，主要包括任务是否提交成功，任务占用CPU，GPU数，任务的状态(启动，执行，完成等)，任务日志，输出文件。
(g)结果展示
调用分析模块分析展示结果，任务数据会自动写入到数据库模块中。展示结果主要根据方法模块中调用了哪些方法有关，主要内容包括对结合模式，分子动力学轨迹，化合物结构，化合物性质等，利用图表，动画展示这些化合物结构的稳定性，小分子靶点之间的关键相互作用，能量值等。此处以分子对接和自由能计算为例。分子对接完成后，系统可以展示软件计算得到的结合模式，标记关键的小分子和靶点之间的相互作用，如氢键，疏水相互作用，极性相互作用等，并可以输出成PNG格式图片，GIF动画等标准格式方便日后回看及展示。自由能计算可以展示如计算得到的自由能值，误差，动力学轨迹稳定性等，亦可以输出成PNG格式图片，GIF动画等方便日后回看及展示。
(h)循环执行
步骤(b)-(g)会循环执行，是否终止取决于预设的指标，比如预测得到100个活性较好的化合物，预测得到能量值在某个范围区间的分子等，满足标准可以停止次轮计算等。
基于虚拟现实技术的药物设计云计算流程控制方法的完整过程包括以上步骤(a)-(h)，用户也可以根据项目的需求自由的组合步骤(a)-(h)。
本发明带来了如下有益效果：
a)实现了高通量、高并行的基于虚拟现实技术的药物设计云计算任务的跨 平台调度。
b)无缝衔接了新药筛选流程的分子处理，评估，和分析等功能。
c)将所有分子对接及虚拟筛选任务的创建、提交和管理，结果分析和绘图等步骤集成到Jupyter中，实现了分子对接云计算流程的可视化操作。
d)引入了虚拟显示技术，可以更加真实的进行基于结构的理性药物设计。。
附图说明
图1是基于虚拟现实技术的药物设计云计算流程控制方法的流程示意图。
图2是基于虚拟现实技术的药物设计云计算流程控制方法的验证模块具体内容；
图3是基于虚拟现实技术的药物设计云计算流程控制方法的方法模块具体内容；
图4是实施例1生物实验测试-计算预测活性相关性图
图5是实施例2中，传统流程与Majorana任务调度系统的CPU运行时间的对比。
图6是实施例3中药物设计方法结果展示图。
图7是实施例4中分子描述结果展示图。
图8是实施例4中结合模式结果展示图。
具体实施方式
下面结合附图，对本发明的较优的实施例作进一步的详细说明：
实施例1
如图1-3所示，基于人工智能和虚拟现实技术的药物设计云计算流程控制系统，该系统包括药物设计模块组、交互和调度模块和虚拟现实药物设计模块；
所述的药物设计模块组包括验证模块、数据库模块、过滤模块、预处理模块、方法模块、分析模块；所述的药物设计模块组包括典型的药物设计方法和软件，具体包括Maestro软件，MOE软件，Amber分子动力学软件，Autudock分子对接软件，Pymol显示软件，VMD显示软件，以模块的 形式展现，用户可以根据项目需求自由选择和组合相应的模块来进行计算；
所述的方法模块包括分子对接模块(Molecular docking)，分子动力学模块(Molecular Dynamics)，结合自由能计算模块(MMGBSA/MMPBSA)，自由能微扰模块(FEP)，药效团模块(Pharmacophore)，人工智能分子生成模块，ADMET预测模块。
所述的交互和调度模块包括Jupyter单元和Majorana任务调度平台；
所述的Jupyter单元是与用户进行交互的界面，用户可以项目的需求通过Jupyter单元发出调用各种模块的指令完成进行相应的药物设计操作；也可以通过调用虚拟现实药物设计模块，使整个药物设计操作以虚拟现实的可视化操作方式完成，并展示操作的结果；
所述的Majorana任务调度平台是一个内部开发的调度系统，其功能是通过接收Jupyter单元发出的用户指令在药物设计模块组中调取相应的任务模块来计算，将计算任务提交到指定的云计算平台，并查看计算任务状态；所述的查看计算任务状态主要包括任务是否提交成功，任务占用CPU，GPU数，任务的状态(启动，执行，完成等)，任务日志，输出文件等。
所述的云计算平台包括亚马逊云，腾讯云，阿里云，Majorana任务调度平台内置了对接各个云计算平台的任务执行脚本和任务监控脚本，用户可以根据计算的需求轻松实现跨平台云计算需求；
本发明与已有技术不同在于：所述的虚拟现实药物设计模块包括虚拟现实硬件设备：显示器、虚拟现实眼镜，虚拟现实姿势识别器和电脑鼠标，通过操作虚拟现实硬件设备可以使药物设计云计算流程控制方法通过虚拟现实的可视化操作方式操作完成，并可以等计算完成后及时显示结果。这可以比传统的药物设计软件给用户带来更好的视觉效果，以及理解其中的分子机制，对用户的专业背景知识要求更少。
实施例2
针对一个靶点进行苗头化合物的筛选
首先进行小规模的验证，从数据库模块调用数据库中的小分子，数量在1千万左右，然后调用预处理模块对这些分子进行预处理，考虑到项目的目标是筛选苗头化合物，所以选择调用方法模块对分子对接方法进行了 方法验证，找到了合适的筛选参数。
然后调用数据库模块读取原始的数据和结构,接着利用小规模的验证得到的参数，调用Majorana任务调度平台提交筛选任务，并实时跟踪计算结果。计算完成后，调用分析模块分析计算结果。根据计算结果，推荐筛选出的苗头化合物给生物测试团队，完成了计算任务。图4表示用本实施例的计算模块预测的小分子活性，推荐给生物测试团队后，与实验活性的相关性图。从图4中可以看出，计算预测活性和实验活性具有高度的相关性，表明了本实施例计算方法的可靠性。
本方法和其他传统方法对比，首先是流程化以及模块化的操作，方便了筛选工作的进行，提升了效率。同时因为Majorana任务调度系统的使用，使得计算的跟踪，分析都更加方便，进一步提升了效率。图5表示，使用传统药物设计流程和Majorana任务调度系统的CPU运行时间对比，可以看出使用Majorana任务调度系统仅占传统流程运行时间的40％，表示流程化以及模块化的操作能显著提升工作效率。
实施例3
针对一个靶点开发一款新药的先导化合物
通过实施例1中筛选出了一些实验性质不错的候选苗头化合物，准备再次利用基于虚拟现实技术的药物设计云计算流程控制方法来优化此批苗头化合物，并最终得到先导化合物。
首先将设计的分子和靶点结构导入系统，调用虚拟现实药物设计模块，根据结合模式开展基于虚拟现实的药物设计。直接根据虚拟现实展示的结合模式来分析靶点和苗头化合物之间关键的结合模式，根据实际的计算需求调用相应的方法模块，接着调用Majorana任务调度平台，根据方法模块的计算需求，提交计算任务到云平台，并利用计算平台在后台进行云计算评估，并及时反馈给团队成员。团队成员根据计算评估反馈，并在虚拟现实技术的帮助下，再次观看结合模式，并重新设计分子，然后再次提交后台计算评估。经过多轮迭代，找到合理的候选先导化合物。过程中所有的数据，调度，都会记录到数据库，便于日后回溯复盘。
如图6所示，表示本实施例的药物设计流程的可视化。通过这个过程， 可以观察到分子的片段切断与生长过程，以及不同骨架之间的跃迁路径。从这个可视化过程，可以了解到目标分子的逐步优化过程。本方法和其他传统方法对比，首先是流程化以及模块化的操作，方便了先导化合物优化工作的进行，提升了效率。本案例中还使用了虚拟现实的药物设计模块，方便了对结果的分析和解读；同时因为Majorana任务调度系统的使用，使得计算的跟踪，分析都更加方便，进一步提升了效率。
实施例4
针对一个靶点开发一款新药并对外展示研发成果
从实施例1和实施例2的操作中得到了一些实验性质不错的候选先导化合物。利用基于虚拟现实技术的药物设计云计算流程控制方法，调用分析展示这批计算结果，展示出先导化合物包括结合模式，相互作用，能量值，稳定性，主成分分析在内的相关结果等。
如图7和图8所示，图7表示本实施例的计算方法对分子结构的描述，主要是通过分子指纹和键值对的方式，可以快速获得对大量分子的结构特征，并以数字化方式表示和存储。图8表示分子与蛋白的结合模式，从图中可以看出，不仅表示出了蛋白-配体间的基本相互作用，还进一步对相互作用类型做了聚类，得到一个类药效团的分析，可以获得更多关于蛋白-配体结合相互作用和驱动力的洞见。本方法和其他传统方法对比，首先是流程化以及模块化的操作，方便了结果的展示。
以上内容是结合具体的优选实施方式对本发明所作的进一步详细说明，不能认定本发明的具体实施只局限于这些说明。对于本发明所属技术领域的普通技术人员来说，在不脱离本发明构思的前提下，还可以做出若干简单推演或替换，都应当视为属于本发明的保护范围。

Claims (9)

  1. 一种基于虚拟现实的药物设计云计算流程控制系统，其特征在于，包括药物设计模块组、交互和调度模块和虚拟现实药物设计模块；
    所述药物设计模块组包括验证模块、数据库模块、过滤模块、预处理模块、方法模块、分析模块；
    所述药物设计模块组包括Maestro软件，MOE软件，Amber分子动力学软件，Autudock分子对接软件，Pymol显示软件，VMD显示软件，以模块的形式展现，用户根据项目需求自由选择和组合相应的模块来进行计算；
    所述方法模块包括分子对接模块、分子动力学模块、结合自由能计算模块、自由能微扰模块、药效团模块、人工智能分子生成模块，ADMET预测模块；
    所述交互和调度模块包括Jupyter单元和Majorana任务调度平台。
  2. 如权利要求1所述的系统，其特征在于，所述Jupyter单元是与用户进行交互的界面，根据项目需求通过Jupyter单元发出调用各种模块的指令完成进行相应的药物设计操作；或者通过调用虚拟现实药物设计模块，使整个药物设计操作以虚拟现实的可视化操作方式完成，并展示操作的结果。
  3. 如权利要求1所述的系统，其特征在于，Majorana任务调度平台内置了对接各个云计算平台的任务执行脚本和任务监控脚本，用户根据计算的需求轻松实现跨平台云计算需求；，通过接收Jupyter单元发出的用户指令在药物设计模块组中调取相应的任务模块来计算，将计算任务提交到指定的云计算平台，并查看计算任务状态。
  4. 如权利要求3所述的系统，其特征在于，所述查看计算任务状态主要包括任务是否提交成功，任务占用CPU，GPU数，任务的状态，任务日志，输出文件。
  5. 如权利要求1所述的系统，其特征在于，所述虚拟现实药物设计模块中所述的可视化操作可以以2D或者3D的方式进行操作；所述虚拟现实药物设计模块中所述的显示结果可以以2D或者3D的方式进行展示。
  6. 一种基于虚拟现实技术的药物设计云计算流程控制方法，其特征在 于，包括以下步骤：
    步骤(a)小规模验证：
    调用验证模块，包括预处理，数据库，方法和分析四个子模块，根据实际需求选择所需要的子模块，通过选用合适的子模块进行分析，验证，得到相应的参数；
    步骤(b)数据读入：
    调用数据库模块读取原始的数据和结构，主要是从数据库中将靶点和小分子化合物的结构信息读入到工作流程中；
    步骤(c)过滤筛选：
    调用过滤模块对小分子数据库进行初步的过滤筛选，通过成药性五规则过滤掉一些成药性不佳的分子；
    步骤(d)预处理：
    调用预处理模块处理数据库模块读取的靶点和小分子数据，进行正式计算前的预处理工作；
    步骤(e)计算分析：
    调用方法模块，根据实际的计算需求，调用相应的模块进行计算分析，
    步骤(f)云计算：
    调用Majorana任务调度平台，根据方法模块的计算需求，提交计算任务到指定云平台并查看计算任务状态；
    步骤(g)结果展示：
    调用分析模块分析展示结果，任务数据会自动写入到数据库模块中。步骤(h)循环执行：
    步骤(b)-(g)会循环执行，，基于虚拟现实技术的药物设计云计算流程控制方法的完整过程包括以上步骤(a)-(h)，是否终止取决于预设的指标。
  7. 如权利要求6所述的方法，其特征在于，所述步骤(d)中，预处理分为靶点的预处理和小分子的预处理；靶点的预处理包括：晶体结构中靶点缺失残基的修补，加氢，加电荷，氨基酸的质子化，并保存成标准的PDB格式；数据库过滤后的每个小分子都要进行预处理，主要包括平面二维结 构转三维结构，最低能量构象搜索，PKa计算，加氢，加电荷，并保存成标准的格式。
  8. 如权利要求6所述的方法，其特征在于，步骤(g)结果展示中，主要内容包括如下至少一种：结合模式，分子动力学轨迹，化合物结构，化合物性质，利用图表，动画展示这些化合物结构的稳定性，小分子靶点之间的关键相互作用，能量值。
  9. 如权利要求6所述的方法，其特征在于，所述步骤(h)中，所述指标包括如下至少一种：预测得到100个活性较好的化合物，预测得到能量值在某个范围区间的分子等，满足标准可以停止次轮计算。
PCT/CN2020/123285 2020-10-23 2020-10-23 基于虚拟现实的药物设计云计算流程控制系统及其方法 WO2022082740A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/123285 WO2022082740A1 (zh) 2020-10-23 2020-10-23 基于虚拟现实的药物设计云计算流程控制系统及其方法

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/123285 WO2022082740A1 (zh) 2020-10-23 2020-10-23 基于虚拟现实的药物设计云计算流程控制系统及其方法

Publications (1)

Publication Number Publication Date
WO2022082740A1 true WO2022082740A1 (zh) 2022-04-28

Family

ID=81291180

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/123285 WO2022082740A1 (zh) 2020-10-23 2020-10-23 基于虚拟现实的药物设计云计算流程控制系统及其方法

Country Status (1)

Country Link
WO (1) WO2022082740A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103294933A (zh) * 2013-05-10 2013-09-11 司宏宗 一种药物筛选方法
CN109885390A (zh) * 2019-02-21 2019-06-14 深圳晶泰科技有限公司 分子对接云计算流程控制方法
CN110867256A (zh) * 2019-10-17 2020-03-06 湖南中医药大学 一种优化肝复乐抗肝癌组方药理学分析方法和系统
CN110910951A (zh) * 2019-11-19 2020-03-24 江苏理工学院 一种基于渐进式神经网络预测蛋白质与配体结合自由能的方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103294933A (zh) * 2013-05-10 2013-09-11 司宏宗 一种药物筛选方法
CN109885390A (zh) * 2019-02-21 2019-06-14 深圳晶泰科技有限公司 分子对接云计算流程控制方法
CN110867256A (zh) * 2019-10-17 2020-03-06 湖南中医药大学 一种优化肝复乐抗肝癌组方药理学分析方法和系统
CN110910951A (zh) * 2019-11-19 2020-03-24 江苏理工学院 一种基于渐进式神经网络预测蛋白质与配体结合自由能的方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
KITCHEN, DOUGLAS B. ET AL.: "DOCKING AND SCORING IN VIRTUAL SCREENING FOR DRUG DISCOVERY:METHODS AND APPLICATION S", NATURE REVIEWS DRUG DISCOVERY, vol. 3, 30 November 2004 (2004-11-30), XP007905248, DOI: 10.1038/nrd1549 *

Similar Documents

Publication Publication Date Title
Nithya et al. Predictive analytics in health care using machine learning tools and techniques
US10255409B2 (en) Systems and methods for in silico evaluation of polymers
CN112201301A (zh) 基于虚拟现实的药物设计云计算流程控制系统及其方法
CN100492316C (zh) 测试自动化的系统和方法
US20070162482A1 (en) Method and system of using artifacts to identify elements of a component business model
US7991730B2 (en) Methods for similarity searching of chemical reactions
US20210279629A1 (en) Machine learning and computer-based generation of standard work matrices for improving execution of a standard work
CN109637592A (zh) 分子力场参数生成的计算任务管理分析系统及其运行方法
Back et al. A study for production simulation model generation system based on data model at a shipyard
CN109885390B (zh) 分子对接云计算流程控制方法
WO2022082740A1 (zh) 基于虚拟现实的药物设计云计算流程控制系统及其方法
CN111312342B (zh) 一种电子结构计算机辅助药物设计系统
CN111861757A (zh) 一种融资匹配方法、系统、设备和存储介质
CN108132802B (zh) 一种配置模型在系统开发中的应用方法
CN106940836A (zh) 一种数据分析方法及装置
US8429600B2 (en) Distributed collaborative environment definition system and method
US9830626B1 (en) Pre-calculated performance simulation
US11810197B2 (en) Systems and methods for benefit plan quality assurance and certification
CN114360637A (zh) 一种基于图注意力网络的蛋白质-配体亲和力评价方法
Hu et al. LS-HTC: an HTC system for large-scale jobs
US11609807B2 (en) Computing task management and analysis system for molecular force field parameter building and operation method thereof
Mushnoori et al. Pipelines for automating compliance-based elimination and extension (PACE2): a systematic framework for high-throughput biomolecular materials simulation workflows
WO2020168507A1 (zh) 分子对接云计算流程控制方法
US11080309B2 (en) Validating cluster results
Elmer et al. Strategic Plan for a Scientific Software Innovation Institute (S2I2) for High Energy Physics DRAFT

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20958322

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 31/08/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20958322

Country of ref document: EP

Kind code of ref document: A1