PharmaScreen-AI drug auxiliary research and development platform

Abstract

Functions of traditional PDB library, molecular drawing, molecular SMILES sequence generation websites, compound search, target prediction, molecular inverse synthesis and the like are dispersed. The platform integrates a plurality of protein, medicine and small molecule databases, platforms and toolkits, and aims to build a one-stop AI + medicine auxiliary research and development platform, so that time is saved for learning and research, and the efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of smart pharmaceutical manufacturing (AI+small molecule pharmaceutical manufacturing), and more specifically, to a software platform for "AI+drug-assisted research and development". Background Art

[0002] The traditional drug development process is complex and uncertain, and researchers often face problems such as data silos, information asymmetry, and low efficiency. These problems not only prolong the time it takes for drugs to go from the laboratory to the market, but also significantly increase the cost of research and development, and require knowledge and skills across multiple fields such as science, technology, regulations, and the market. Therefore, the drug development auxiliary platform came into being, aiming to design a platform that can integrate resources and provide strong support for drug development.

[0003] With the development of technologies such as artificial intelligence, big data analysis, and cloud computing, the field of drug research and development has ushered in unprecedented opportunities. The application of these technologies can help researchers process and analyze large amounts of data more quickly, predict the effects and side effects of drugs, and even simulate complex biological reaction processes. The design of the drug research and development auxiliary platform is based on the maturity and feasibility of these technologies, in order to maximize their potential in drug research and development.

[0004] Traditional PDB libraries, molecular drawing, websites for generating molecular SMILES sequences, and functions such as compound search, target prediction, and molecular reverse synthesis are scattered. This platform integrates several protein, drug, and small molecule databases, platforms, and toolkits, aiming to build a one-stop AI+drug-assisted R&D platform to save time and improve efficiency for learning and research. Summary of the invention

[0005] The main functions are as follows: Since the platform needs to load resources when accessing each function for the first time, the resource preloading interface will be entered when the resources are loaded. Please wait patiently. If the waiting time is too long, try to clear the browser cache and refresh and reload the page again.

[0006] 1. PDB query preview: Manually upload offline pdb files, or use the online public PDB library to preview the protein 3D structure. You can use the mouse to zoom in and rotate the structure, and you can also view the protein active pocket. This can facilitate experimental research on the role of various groups in enzyme molecules and their effects on the structure, properties and functions of enzyme molecules. It can also be used to study the essential groups in the active pocket of the enzyme.

[0007] 2. Molecule query: Use the elements, ring structures and chemical bonds in the molecule drawing status bar to draw molecules, and you can get the SMILES sequence of the molecule. After copying, you can query the molecule information in the molecule preview, or import MOL, MOL2 files to query related information.

[0008] 3. Compound search: Get related compounds by entering the compound name, ChEMBL ID, and SMILES sequence. You can view the chemical information after downloading.

[0009] 4. Molecular generation: Random drug small molecules can be generated from scratch, and drug small molecules can be generated based on existing SMILES sequences.

[0010] 5. Molecular retrosynthesis route planning, using traditional machine learning algorithms or deep learning networks to plan retrosynthesis routes. Using the known drug molecular structure, cut the appropriate chemical bonds, convert the target molecule into some smaller intermediates, and then use these intermediates as new target molecules, cut them into new smaller intermediates, and so on, until you get the starting materials that can be easily purchased.

[0011] The main technical solutions and software basic development frameworks adopted: pydantic, fastapi, paddlepaddle, paddlenlp, paddlets, sklearn, pinia, molstarvolseg, molstar-viewer-vue, ChemDoodleWeb.js, ThreeJS, and TresJS. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1: The opening screen loading interface.

[0013] Fig. 2 PCD previews the protein three-dimensional structure interface.

[0014] Figure 3 Drag to display the schematic diagram of protein 3D structure.

[0015] Figure 4 PDB browsing history (the browser system where the page is displayed is Ubutun22.04LTS).

[0016] Figure 5 Schematic diagram of clicking preview history to continue previewing.

[0017] Figure 6 Schematic diagram of clearing preview history.

[0018] Figure 7 Two preview methods.

[0019] Figure 8 The download pop-up window that pops up in Chrome after clicking Download.

[0020] Figure 9 shows the Chrome download record (the browser system of the displayed page is Win11).

[0021] Figure 10 Skeleton screen interface when the molecule preview is loaded.

[0022] Figure 11 is located atFig.10 The input box just above.

[0023] Fig. 12 Molecular structure (2D / 3D) display frame.

[0024] Figure 13 shows how to click on the lower right corner of the display box to download the molecule file.

[0025] Fig. 14 Schematic diagram of clearing history in molecule preview.

[0026] Fig. 15 Schematic diagram of the molecular drawing interface.

[0027] Figure 16 Compound search interface.

[0028] Figure 17 The search box above the compound search.

[0029] Figure 18 Schematic diagram of compound search results for "aspirin".

[0030] Figure 19 Schematic diagram of selecting the option to download compound search results.

[0031] Figure 20 All pages have a "Contact Us" button in the lower left corner, with the author information above it. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] PDB query preview: see the interface Figure 2 .

[0034] This functional interface can preview the 3D structure of proteins. Users can manually upload offline pdb files or use the online public PDB repository to search and display protein 3D structures.

[0035] Among them, it has been connected to online public databases such as RCSB (a data archive on the three-dimensional structure of biological macromolecules maintained by the Brookhaven National Laboratory in the United States, which includes the atomic coordinates, references, first- and second-level structure information of biological macromolecules, as well as crystal structure factors and NMR experimental data).

[0036] Click on the online open source search to search for PDB in the pop-up search box and select display. The default PDB source is RCSB.

[0037] The protein 3D structure can be rotated and manipulated to view the structure. The middle view uses the open source Molstar protein structure display tool (Molstar), which can easily rotate and display the 3D structure of the protein. Clicking and dragging with the mouse and using the arrow keys on the numeric keypad can adjust the display view and clear the history.

[0038] When you open the website, if the browser supports and allows LocalStorage, the browsing history will be saved after the user browses the protein structure so that they can be viewed again from time to time.

[0039] When you don’t need it, you can right-click to delete or click the Clear button at the top to clear the record: The folder icon indicates that this PDB file is a local file, such as Figure 4 The first PDB file is located in the ` / plus / tmp` folder of Ubutun22.04LTS (the operating system where the browser is located). If this file does not exist, the record will be automatically deleted after a pop-up warning.

[0040] The protein icon indicates the pdb file retrieved from the online PDB library. After more PDB libraries are connected, they will be replaced with the logos of the PDB online libraries. Figure 3 .

[0041] If you need to download the PDB file to your local computer, you can click the download icon on the right side of the browsing history generated each time you browse to download it. After downloading, you can view it in the download content of the corresponding browser and the default download folder. Figure 7 Figure 8 Fig. 9 .

[0042] Molecule preview: see the interface Fig.10 Fig.12 .

[0043] Upload a molecular file or enter a molecular SMILES sequence to query molecular information. PS: This version currently only supports: MOL, MOL2, SMI file formats, and will gradually support more mainstream formats in the future.

[0044] After inputting the SMILES sequence or uploading the supported file format, MOL2 and SMILES will be automatically generated and converted, and the molecular information will be automatically retrieved and displayed in the TCMSP-E database.

[0045] Molecular drawing: see the interface Fig.15 .

[0046] Draw molecules and dynamically generate SMILES sequences corresponding to the molecules.

[0047] Compound search: see the interface Fig.16 .

[0048] After entering the relevant information, the ChEMBL database will be searched and displayed.

[0049] Take the search for `aspirin` as an example: you can see that compounds related to aspirin are retrieved. There are two display modes: card display and list display. Other display modes will be launched after the platform is improved. You can select to download the selected search information, see Fig.18 .

[0050] The remaining features are currently under development.

Claims

1. A one-stop AI+ smart drug assisted R&D platform, featuring: It integrates commonly used tools for molecules and proteins, such as drawing molecules and copying their SMILES sequences, searching for molecular structures based on SMILES sequences, viewing protein three-dimensional structures, converting common molecular files to SMILES sequences, etc. It also carries many popular models for drug-assisted synthesis, such as small molecule generation algorithms, target prediction algorithms, etc. For specific algorithms and tools, see the manual.

2. The function of "drawing a molecule and copying its SMILES sequence" described in claim 1 is located in the "molecule drawing" function bar. Clicking this function will pop up the "molecule drawing" subpage. You can draw the specified 2D molecule by dragging or clicking the relevant tools in the toolbar. The generated molecular formula will generate a SMILES sequence at the bottom of the interface, which can be copied. You can also directly enter the SMILES sequence and modify it on the original 2D molecular structure.

3. The function of "Searching for molecular structure based on SMILES sequence" described in claim 1 is located in the "Molecular Preview" function bar. Clicking this function will pop up the "Molecular Preview" subpage, where you can preview the 2D / 3D structure of the molecule, as well as Molecule ID information, etc. The right sidebar has a history record function, which is convenient for users to view molecules that have been viewed in the past. It supports direct drag and drop to upload molecular files, such as: "1055.MOL2". In addition, the remaining sub-functions have supporting "history record" sub-functions, see the instructions for details.

4. The "View Protein 3D Structure" in claim 1 is located in the "PDB Preview" sub-function. After clicking it, you can search for the corresponding protein online and preview its 3D structure. You can also upload a local protein 3D structure to preview the 3D structure.

5. The sub-functions of "common molecule file and SMILES sequence conversion" as described in claim 1, "molecule drawing" and "molecule preview" can all be realized, see claim 2 and claim 3.

6. The compound search sub-function is characterized by the use of the online free ChEMBL database, which can retrieve and display relevant molecular information.

7. According to claim 1, the essential feature of this platform is that it brings together commonly used databases, tools and related algorithm implementations for small molecule drug research and development. Please refer to the instructions for details of the functions and algorithms.

Images