system

Abstract

Provide a system. A means for collecting information related to a living body from a database; means for pre-processing the collected biological information; means for training a generative model based on the preprocessed information using a machine learning algorithm; and means for generating novel compounds using the trained generative model; A means for evaluating the effects and side effects of the generated novel compound through simulation; A means for selecting an optimal compound based on the evaluation results; means for generating a report including the selection results; a means for providing the generated report to a user; A system including:

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] Current anticancer drugs often have strong side effects that significantly reduce patients' quality of life (QOL). Furthermore, discovering new anticancer drugs requires a great deal of time and cost, making it necessary to streamline this process. With conventional methods, the process from compound design to efficacy verification is complicated, making it particularly difficult to predict and suppress side effects. Furthermore, there is the issue of how to effectively utilize existing data to rapidly discover new, effective compounds. [Means for solving the problem]

[0005] The present invention provides a system for generating novel compounds by training a generative model using a machine learning algorithm based on biological information collected from a database. This system then designs novel compounds using the generative model trained on data collected and preprocessed by a server, and evaluates their efficacy and side effects through simulation. Furthermore, the system employs a mechanism for selecting optimal compounds based on the evaluation results, generating a report containing the selection results, and providing it to the user, thereby streamlining the process of discovering new anticancer drugs and enabling the rapid provision of treatments that are beneficial to patients and minimize side effects.

[0006] A "database" is an information management system that stores biological information and allows searching and retrieval of information based on specific queries.

[0007] "Preprocessing" is a process of correcting and removing incomplete data and standardizing the format and scale in order to improve data quality.

[0008] A "machine learning algorithm" is a computational method that learns from large amounts of data, automatically finds patterns and rules, and makes predictions and classifications.

[0009] A "generative model" is a model that uses machine learning algorithms to generate new data or compounds based on training data.

[0010] A "novel compound" is a substance with a new molecular structure that has not previously existed and is designed to have a specific effect.

[0011] "Simulation" is a technology that uses a computer to simulate real physical or chemical phenomena and predict the results.

[0012] A "report" is information that visually and documentally summarizes the results of a simulation or evaluation, and is provided in a format that is easy for users to understand.

[0013] "Users" are researchers, doctors, and other users who use the system to obtain information and provide feedback. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14]FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0015] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0016] First, the terms used in the following description will be explained.

[0017] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0018] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0019] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0020] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0021] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0022] [First embodiment]

[0023] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0024] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0025] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0026] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0027] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0028] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0029] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0030] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0031] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0032] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0033] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0034] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0035] This invention is a system for efficiently designing and evaluating anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. The generated reports are provided to users, and further improvements are made based on user feedback.

[0036] Data collection and preprocessing

[0037] 1. Data Collection

[0038] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases.

[0039] For example, the server uses an API to obtain biometric information and stores it in local storage.

[0040] 2. Data Preprocessing

[0041] The server cleanses the collected data, imputes or removes missing values, and standardizes and unifies the format of the data.

[0042] For example, incomplete data is removed, converted to SMILES format, and standardized.

[0043] Training generative AI and compound generation

[0044] 1. Training the AI ​​model

[0045] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder).

[0046] The trained model is used to generate novel compounds.

[0047] 2. Generation of new compounds

[0048] The server uses the trained model to generate three-dimensional structures of new compounds.

[0049] For example, a user provides the input parameters needed to design a molecule with a particular therapeutic effect, and the server generates a new molecular structure based on those parameters.

[0050] Compound evaluation and selection

[0051] 1. Run the simulation

[0052] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of new compounds.

[0053] For example, the interactions of compounds are analyzed through simulations, and their effectiveness is quantified.

[0054] 2. Evaluation and Selection

[0055] The server analyzes the simulation results and selects the optimal compound.

[0056] For example, the server may preferentially select compounds with high efficacy and few side effects.

[0057] Report Generation and User Interface

[0058] 1. Report Generation

[0059] The server generates a report based on the evaluation results and creates a document containing the visual analysis results.

[0060] For example, the server generates a detailed analysis report in PDF or HTML format.

[0061] 2. User Interface

[0062] The terminal displays the generated report to the user, who can use it to plan their next actions.

[0063] For example, a researcher can use a terminal to log into the system and view the latest reports.

[0064] Incorporating user feedback

[0065] 1. Send Feedback

[0066] Users can input feedback on the report content and send it to the server via their terminal. This feedback is then used to improve the model.

[0067] For example, a user can input clinical trial results and opinions about a particular compound, which are then sent to the server and used as future training data.

[0068] In this way, by carrying out an integrated process from data collection to the generation, evaluation, and feedback of new compounds, the present invention makes it possible to streamline the development of new anticancer drugs and quickly provide treatments with reduced side effects.

[0069] The processing flow will be explained below.

[0070] Step 1:

[0071] Data collection

[0072] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[0073] Step 2:

[0074] Data Preprocessing

[0075] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[0076] Step 3:

[0077] Dataset preparation

[0078] The server splits the preprocessed data into training, validation, and test sets, which prevents overfitting of the model and allows for accurate performance evaluation.

[0079] Step 4:

[0080] Building generative AI models

[0081] The server uses a deep learning framework (e.g., TENSORFLOW® or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[0082] Step 5:

[0083] Training generative AI models

[0084] The server uses the preprocessed data to train the generative AI model, evaluates the model's performance at each epoch, and optimizes the model while correcting errors.

[0085] Step 6:

[0086] Generation of new compounds

[0087] Using a trained generative AI model, the server generates new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[0088] Step 7:

[0089] Simulation execution

[0090] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[0091] Step 8:

[0092] Evaluation and Selection

[0093] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[0094] Step 9:

[0095] Report Generation

[0096] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[0097] Step 10:

[0098] Providing reports

[0099] The terminal displays the generated report to the user, who can use it to plan their next actions.

[0100] Step 11:

[0101] Collecting feedback

[0102] Users can enter feedback on the report content, which is then sent to the server via their device and used to improve the model and plan new experiments.

[0103] This specific processing step enables the system to efficiently and effectively advance the discovery and evaluation of new anti-cancer drugs.

[0104] Example 1

[0105] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0106] The current anticancer drug development process is known to be extremely time-consuming and costly. Furthermore, conventional methods make it difficult to adequately predict the efficacy and side effects of compounds, making it difficult to efficiently discover new compounds. Furthermore, there is a lack of a mechanism for quickly incorporating user feedback, which causes delays in the development cycle. The present invention aims to solve these problems and streamline the design and evaluation of new anticancer drugs.

[0107] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0108] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds by simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for the user to input feedback on the report content and send it to the system. This makes the process of designing and evaluating anticancer drugs more efficient, quickly incorporates user feedback, and enables the discovery of more effective anticancer drugs.

[0109] "Means for collecting" refers to the functions and processes for obtaining information related to a living organism from a database.

[0110] "Preprocessing means" refers to the functions and processes for cleansing collected biological information, completing or removing missing values, and standardizing and formatting data.

[0111] "Means for training a generative model" refers to the functionality and process of training a generative model based on preprocessed information using a machine learning algorithm.

[0112] "Means for generating novel compounds" refers to the functions and processes for generating three-dimensional structures of new compounds using a trained generative model.

[0113] "Means for evaluation by simulation" refers to the function and process of predicting and evaluating the effects and side effects of a generated novel compound using molecular dynamics simulation.

[0114] The "means for selecting an optimal compound" refers to a function and process for selecting an optimal compound based on the simulation results.

[0115] "Means for generating a report" refers to the function and process for generating detailed analysis results in document format (PDF, HTML, etc.) based on the selection results.

[0116] "Means of providing to users" refers to the functions and processes that make the generated reports available to users for viewing and access.

[0117] "Means for inputting feedback and sending it to the system" refers to the function and process by which a user inputs opinions and evaluations on the content of a generated report and sends that information to the system.

[0118] This invention is a system for efficiently designing and evaluating anticancer drugs, which collects biological information from a database, generates compounds using machine learning algorithms, and evaluates their efficacy and side effects. Below, we will explain in detail how to specifically implement this system.

[0119] Hardware and software used

[0120] Server: Responsible for data collection, preprocessing, training of machine learning models, generation of new compounds, simulation, evaluation, report generation, and feedback collection. It is desirable for the server to have hardware with high computing power. For example, an API is used to access the database, and local storage is used for data storage.

[0121] Device: Provides an interface for users to view reports and provide feedback. Devices can be regular computers, tablets, smartphones, etc.

[0122] Software used: Python, TensorFlow, and PyTorch are used for machine learning, and specialized software such as GROMACS and AMBER are used for molecular dynamics simulations.

[0123] Data collection and preprocessing

[0124] The server first collects information related to living organisms from a database. Specifically, it obtains chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs. Examples of databases that are collected include NCBI and PubChem. The server uses an API to obtain the data and saves it in local storage. The collected data is then preprocessed. During preprocessing, missing values ​​are filled in, the data is standardized, and the format is unified. This allows the data to be used efficiently in the next step.

[0125] Training generative AI models

[0126] The server trains a machine learning algorithm (e.g., GAN or autoencoder) based on the preprocessed data. The training data includes chemical structures, clinical trial data, genetic information, and more, enabling the generative AI model to handle a wide variety of conditions. Through this training, the server builds a highly accurate model for generating new compounds.

[0127] Generation of new compounds

[0128] When a user provides a prompt to design a molecule with a specific therapeutic effect, the server uses the trained generative AI model to generate the three-dimensional structure of the new compound. For example, a user might enter the following prompt:

[0129] "Design new anticancer drugs that are highly effective against specific cancer cells. Prioritize drugs with fewer side effects."

[0130] This causes the server to generate a new molecular structure based on the provided conditions and proceed to the next evaluation step.

[0131] Compound evaluation and selection

[0132] The server performs molecular dynamics simulations on the generated new compounds to predict and evaluate their effects and side effects. The simulations use molecular dynamics simulators such as GROMACS and AMBER to analyze the interactions and stability of the compounds. The server then selects the optimal compound based on the simulation results.

[0133] Report Generation and User Interface

[0134] The server generates a detailed analysis report based on the evaluation results and creates a document in PDF or HTML format. The generated report is sent to the terminal, where the user can view it. The user can enter feedback on the report content and send it to the server via the terminal. This feedback is used to improve the model and refine the data in the future.

[0135] In this way, a system will be constructed that efficiently designs and evaluates anticancer drugs through a series of processes.

[0136] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0137] Step 1: Data collection

[0138] The server collects information about anticancer drugs from multiple databases.

[0139] Input: Data from databases such as NCBI and PubChem (e.g., chemical structures, clinical trial data, genetic information).

[0140] Processing: Uses API to access the database and retrieve the required information.

[0141] Output: Collected data is saved to local storage.

[0142] Specific operation: The server calls the NCBI API, collects genetic information, and saves it in a local file in JSON format.

[0143] Step 2: Data Preprocessing

[0144] The server pre-processes the collected data.

[0145] Input: Data stored in local storage (chemical structures in JSON format, clinical trial data, and genetic information).

[0146] Processing: Data cleansing (implantation and removal of missing values), data standardization (e.g., conversion to SMILES format), and format unification.

[0147] Output: The preprocessed dataset.

[0148] Specific operation: The server fills in missing values ​​with 0, converts them to SMILES format, and saves them in a CSV file.

[0149] Step 3: Training the AI ​​model

[0150] The server uses the preprocessed data to train the AI ​​model.

[0151] Input: Preprocessed dataset (CSV file).

[0152] Processing: Train a model using a machine learning algorithm (e.g., GAN or autoencoder).

[0153] Output: A trained generative model.

[0154] What it does: The server uses TensorFlow to train a model on a CUDA-enabled GPU and saves the trained model.

[0155] Step 4: Generation of new compounds

[0156] The server uses the trained generative model to generate novel compounds.

[0157] Input: Prompt statements such as drug efficacy and target gene information provided by the user.

[0158] Processing: A prompt sentence is input into the generative model to generate the three-dimensional structure of the compound.

[0159] Output: Three-dimensional structure of the generated new compound (e.g., SMILES format).

[0160] Specific operation: The server parses the prompt sentence, generates a SMILES format of the compound, and saves it.

[0161] Step 5: Run the simulation

[0162] The server performs molecular dynamics simulations to predict the effects and side effects of the resulting compounds.

[0163] Input: Three-dimensional structure of the new compound (SMILES format).

[0164] Processing: Molecular dynamics simulations are performed using GROMACS and AMBER.

[0165] Output: Simulation results (evaluation indicators such as efficacy, side effects, and stability).

[0166] Specific operation: The server runs molecular dynamics simulations using GROMACS, analyzes the results, and saves them as a report.

[0167] Step 6: Evaluation and selection

[0168] The server selects the optimal compound based on the simulation results.

[0169] Input: Simulation results.

[0170] Processing: Compare the evaluation indexes of each compound and select the most suitable compound.

[0171] Output: Detailed information on the selected compound.

[0172] Specific operation: The server applies an algorithm based on the simulation results to select the optimal compound.

[0173] Step 7: Generate reports

[0174] The server generates a report based on the evaluation results.

[0175] Input: Detailed information about the selected compound.

[0176] Processing: Generate detailed analysis reports in PDF and HTML formats.

[0177] Output: The generated analysis report.

[0178] Specific operation: The server embeds information into a report template based on the selection results and saves it as a PDF file.

[0179] Step 8: User Interface

[0180] The terminal displays the generated report to the user.

[0181] Input: The generated analysis report.

[0182] Action: Display the report on the terminal screen.

[0183] Output: Users can view the report.

[0184] Specific operation: The terminal displays the latest report through a web browser, and the user views it.

[0185] Step 9: Send your feedback

[0186] The user inputs feedback on the report content and sends it to the server via the terminal.

[0187] Input: User feedback (opinions and ratings about a particular compound).

[0188] Processing: Send feedback from the device to the server.

[0189] Output: The feedback is saved on the server.

[0190] Specific operation: When a user enters their opinion into the feedback form and clicks the "Submit" button, the information is sent to the server and stored.

[0191] The above is a concrete explanation of the processing flow of this system and each processing step.

[0192] (Application example 1)

[0193] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0194] In conventional anticancer drug development, the design of new compounds, simulation evaluation, and the introduction of optimized manufacturing processes are performed separately, making the overall process extremely time-consuming and costly. Furthermore, while it is necessary to not only generate and evaluate new compounds, but also to automate and optimize the manufacturing process, there is no system in place to consistently perform these tasks. This makes it difficult to quickly bring drugs to market, and efficient methods are needed to improve the quality of treatment.

[0195] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0196] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for optimizing and automating the manufacturing process in the factory, thereby enabling the rapid development of anticancer drugs and the efficiency of the manufacturing process.

[0197] "Information related to living organisms" is a general term for data such as chemical substances that affect living organisms, genetic information, and clinical trial data.

[0198] "Preprocessing" refers to a series of processes that cleanse collected data, fill in or remove missing values, and standardize data and standardize formats.

[0199] "Machine learning algorithms" is a general term for mathematical models and methods for making future predictions and classifications based on past data.

[0200] A "generative model" is a model that is trained by a machine learning algorithm and is responsible for generating new data.

[0201] "Novel compounds" refer to chemical substances with new molecular structures that differ from existing compounds, and in particular refer to candidate substances in the development of pharmaceuticals such as anticancer drugs.

[0202] "Simulation" is a technology for virtually testing and predicting the effects and side effects of new compounds on a computer.

[0203] The "evaluation results" are information such as the efficacy and side effects of new compounds obtained through simulation, and are data for determining the usefulness of the compounds.

[0204] A "report" is a document created based on the evaluation results, and includes information on the analysis results and the optimal compound.

[0205] "Feedback" refers to opinions and additional information provided by users who review a report, and is used to improve the system.

[0206] "Manufacturing process optimization" refers to making adjustments and improvements to operate the manufacturing process efficiently, with the aim of optimizing resource allocation and reducing costs.

[0207] "Automation" refers to the process of using systems or robots to automatically perform processes that require manual operation.

[0208] This invention relates to a system for streamlining the design and evaluation of anticancer drugs, and includes a process of collecting biological information from a database, generating new compounds using machine learning algorithms, and evaluating their effects and side effects through simulation.

[0209] A specific embodiment of the system uses the following hardware and software: The server is a high-performance computer equipped with the latest GPU, and the software uses Python, TensorFlow, Keras, and other necessary scientific calculation libraries. The database has an API for collecting biological information. Simulation software such as GROMACS is also used to perform molecular dynamics simulations.

[0210] System program processing overview

[0211] 1. Data Collection and Preprocessing

[0212] The server collects bio-related information such as chemical structures, clinical trial data, and genetic information related to anticancer drugs from multiple databases via APIs. The collected data is cleansed and undergoes pre-processing such as filling in or removing missing values ​​and standardizing data formats.

[0213] 2. Training the generative AI and generating compounds

[0214] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or autoencoder) and generates novel compounds using the trained generative model, generating new molecular structures based on user-provided input parameters for designing molecules with specific therapeutic effects.

[0215] 3. Compound evaluation and selection

[0216] Molecular dynamics simulations are performed on the generated new compounds to predict their efficacy, side effects, and stability. The simulation results are analyzed to select the optimal compound with high efficacy and few side effects.

[0217] 4. Report Generation and User Interface

[0218] The server generates a report based on the evaluation results for the selected compounds. This report contains visual analysis results and is generated in PDF or HTML format. Users can view this report on their devices and plan their next actions.

[0219] 5. Incorporating User Feedback

[0220] Users can provide feedback on the generated reports, which is then sent to the server, where it is used as training data for future model improvements.

[0221] Specific examples

[0222] For example, if a user wants to create a new compound with high efficacy against a specific type of cancer cell, the server will design a chemical structure based on that request, run a molecular dynamics simulation to evaluate its effectiveness, and provide a report of the results, collecting feedback from the user so that it can be reflected in the next compound creation. This dramatically improves the efficiency of drug development.

[0223] Prompt Sentence Examples

[0224] "Using a generative AI model, propose new anti-cancer drug compounds. Evaluate the efficacy of these compounds through simulations and select the most suitable one."

[0225] As described above, this system significantly improves the speed and efficiency of anticancer drug development by consistently automating processes from data collection to the generation, evaluation, report creation, and feedback collection of new compounds.

[0226] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0227] Step 1:

[0228] The server collects biological information (e.g., chemical structures, clinical trial data, and genetic information) from multiple databases through APIs. It uses the APIs as input to obtain biological data. The data collection process involves sending requests to each database and storing the returned data in local storage. The output is a raw dataset of biological data.

[0229] Step 2:

[0230] The server preprocesses the collected data. First, it cleanses the data by imputing or removing missing values. Then it standardizes the data and unifies the format. The input is the raw data collected in step 1, and the output is the cleansed, standardized data. Specifically, it runs a data cleansing script to remove incomplete data and convert the remaining data into a unified format (for example, SMILES format).

[0231] Step 3:

[0232] The server uses the preprocessed data to train a machine learning algorithm. The input is the standardized data obtained in step 2. In operation, a generative AI model (e.g., a GAN or autoencoder) is trained using a machine learning library such as TensorFlow or Keras. The output is the trained generative model.

[0233] Step 4:

[0234] The server uses the trained model to generate the three-dimensional structure of a new compound. The generative model operates based on a prompt from the user (e.g., "Please design a molecule with a specific medicinal effect."). The input is the trained generative model and the user's prompt, and the output is the three-dimensional structure of the new compound. Specifically, the server inputs the prompt into the model and executes the process of generating a new compound based on it.

[0235] Step 5:

[0236] The server performs molecular dynamics simulations on new compounds to predict their efficacy, side effects, and stability. The input is the three-dimensional structure of the new compound generated in step 4. The operation uses simulation software such as GROMACS to analyze the interactions of the compounds. The output is the simulation results (data on efficacy, side effects, and stability).

[0237] Step 6:

[0238] The server analyzes the simulation results and selects the optimal compound. The input is the simulation results from step 5. The operation compares the evaluation indicators of each compound (drug efficacy, side effects, etc.) and selects the optimal compound. The output is information about the selected optimal compound.

[0239] Step 7:

[0240] The server generates a report based on the evaluation results and creates a document containing the visual analysis results. The input is the selection results from step 6. Specifically, the server uses report generation software to create a document in PDF or HTML format. The output is a report.

[0241] Step 8:

[0242] The user uses a terminal to view the generated report and enter feedback. The input is the generated report and the user's feedback. The operation is to enter feedback through the terminal's UI and send it to the server. The output is the collected feedback data.

[0243] Step 9:

[0244] The server collects feedback from users and uses it to improve the model next time. The input is the feedback data entered by the user. In operation, the feedback data is analyzed and stored as new training data. The output is the improved training data for the next time.

[0245] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0246] This invention is a system for streamlining the design and evaluation of anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. Furthermore, it aims to improve the quality of feedback by combining it with an emotion engine that recognizes the user's emotions.

[0247] Data collection and preprocessing

[0248] 1. Data Collection

[0249] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[0250] 2. Data Preprocessing

[0251] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[0252] Training generative AI and compound generation

[0253] 1. Training the AI ​​model

[0254] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder). The trained model is then used to generate novel compounds.

[0255] 2. Generation of new compounds

[0256] The server uses the trained model to generate new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[0257] Compound evaluation and selection

[0258] 1. Run the simulation

[0259] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[0260] 2. Evaluation and Selection

[0261] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[0262] Report Generation and User Interface

[0263] 1. Report Generation

[0264] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[0265] 2. User Interface

[0266] The terminal displays the generated report to the user, who can use it to plan their next actions.

[0267] Emotion engine integration

[0268] 1. Gathering feedback

[0269] When a user enters feedback on a report, the device recognizes the user's emotions in real time through an emotion engine, which uses technologies such as facial expression analysis and voice analysis to analyze the user's emotional state.

[0270] 2. Feedback Analysis

[0271] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in improving the generative model and designing new experiments.

[0272] 3. Dynamically Adjusting Reports

[0273] The device recognizes the user's emotions in real time and dynamically adjusts the content of the report it presents based on their emotional state. For example, if the user expresses anxiety, it will respond by adding more detailed explanations and illustrations.

[0274] Specific examples

[0275] 1. Specific examples of emotion recognition

[0276] When a user logs in to the system and views a report, the device uses a camera and microphone to monitor the user's facial expressions and voice. For example, if the user frowns, the emotion engine recognizes anxiety or doubt, and the device displays additional explanations in the report.

[0277] 2. Specific examples of feedback

[0278] When a user provides feedback on a report, such as "This result makes me feel anxious," the emotion engine sends that feedback along with the emotion data to the server, which then uses that data to identify areas for improvement in the generative model.

[0279] In this way, by combining an emotion engine, feedback and data analysis that take user emotions into account becomes possible, further improving the efficiency of the anticancer drug design and evaluation process.

[0280] The processing flow will be explained below.

[0281] Step 1:

[0282] Data collection

[0283] The server collects chemical structures, clinical trial data, genetic information, etc. related to anticancer drugs from multiple databases via API. This collection is performed based on a specified query, and the retrieved data is stored in temporary local storage. For example, the server may retrieve chemical structure data with specific medicinal properties from PubChem.

[0284] Step 2:

[0285] Data Preprocessing

[0286] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use. For example, it removes records with missing values ​​and standardizes molecular weight and hydrophobicity index.

[0287] Step 3:

[0288] Dataset preparation

[0289] The server splits the preprocessed data into training, validation, and test sets. This prevents overfitting of the model and allows for accurate performance evaluation. For example, split 70% of the data into training, 20% for validation, and 10% for test.

[0290] Step 4:

[0291] Building generative AI models

[0292] The server uses a deep learning framework (e.g., TensorFlow or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[0293] Step 5:

[0294] Training generative AI models

[0295] The server trains the generative AI model using the prepared dataset. The model's performance is evaluated at each epoch, and the model is optimized while correcting errors. As a concrete example, training is performed for 100 epochs, and early stopping techniques are used to avoid overfitting.

[0296] Step 6:

[0297] Generation of new compounds

[0298] The server generates new molecular structures using a trained generative AI model. During the generation process, the structure is adjusted to satisfy constraints specified by the user based on the drug efficacy and molecular properties. For example, to generate a molecular structure with anti-tumor activity, constraints can be set to include specific structural elements.

[0299] Step 7:

[0300] Simulation execution

[0301] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the generated compounds. Because the simulations require advanced computing resources, parallel processing is used. Specifically, software such as GROMACS is used to analyze the dynamic behavior of compounds.

[0302] Step 8:

[0303] Evaluation and Selection

[0304] The server analyzes the simulation results and selects the best candidate from all the generated compounds. Evaluation criteria include high efficacy, few side effects, and molecular stability. For example, candidate compounds are selected based on IC50 values ​​and ADMET profiles.

[0305] Step 9:

[0306] Report Generation

[0307] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, making it easy for users to understand the analysis results. For example, a report is generated that includes the 3D structure of the selected compound and its predicted drug efficacy profile.

[0308] Step 10:

[0309] Providing reports

[0310] The terminal displays the generated report to the user, who can use it to plan their next action. For example, a researcher can use the terminal to log in to the system and view the latest report in a dashboard format.

[0311] Step 11:

[0312] Collecting feedback

[0313] The user enters feedback on the report content. The device recognizes the user's emotions in real time through the emotion engine and sends the feedback to the server after taking those emotions into consideration. For example, when the user enters feedback, the emotion engine analyzes the user's facial expressions and tone of voice and records them as emotional data.

[0314] Step 12:

[0315] Analyzing and incorporating feedback

[0316] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in identifying areas for improvement in the generative model and in new experimental designs. For example, the parameter settings of a generative model that received a lot of positive feedback are reflected in the next training run.

[0317] Example 2

[0318] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0319] In the conventional anticancer drug design and evaluation process, data collection, preprocessing, model training, new compound generation, simulation, evaluation, and feedback collection and analysis are performed separately, making it difficult to improve overall efficiency and accuracy.In addition, it is difficult to improve the quality of feedback taking into account the user's emotional state and to adjust dynamic reports, resulting in a lack of support for user understanding and decision-making.

[0320] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0321] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, means for collecting feedback from the user and analyzing the user's emotions through an emotion engine, and means for dynamically adjusting the report content according to the user's emotional state. This makes it possible to streamline the anticancer drug design and evaluation process, realize feedback collection and analysis from the user's perspective, and support user understanding and decision-making.

[0322] A "database" is an information source that stores and manages information related to living organisms, such as chemical structures of anticancer drugs, clinical trial data, and genetic information.

[0323] "Information related to living organisms" refers to biological and chemical data, such as chemical structures, clinical trial data, and genetic information, that are necessary for the design and evaluation of anticancer drugs.

[0324] "Preprocessing" refers to the process of cleansing collected biological information, correcting and removing incomplete data, standardizing data, scaling data, and otherwise formatting it into a format that is easy for machine learning algorithms to use.

[0325] A "machine learning algorithm" is a mathematical method or model for pattern recognition and prediction using large amounts of data, and is used in this invention to train the generative model.

[0326] A "generative model" is a model for generating molecular structures of novel compounds using a trained machine learning algorithm.

[0327] "New compounds" are chemical substances with new molecular structures created based on data on existing anticancer drugs, and are the target substances for evaluation of their anticancer effects and side effects.

[0328] "Simulation" refers to the process of conducting molecular dynamics analysis on a computer to predict the effects, side effects, and stability of the generated new compound.

[0329] The "evaluation results" are analytical data on the efficacy, side effects, stability, etc. of the new compound obtained through simulation.

[0330] The "optimal compound" refers to a compound that is judged to have high efficacy, few side effects, and a stable molecular structure based on the evaluation results.

[0331] A "report" is a document summarizing detailed analysis results and visualization data, including the selection of optimal compounds.

[0332] "Users" refer to researchers and medical professionals who use this system to design and evaluate anticancer drugs.

[0333] "Feedback" refers to comments or suggestions that a user inputs on a generated report, and is information that can be used as a reference for improving the system or for the next action.

[0334] The "emotion engine" is a mechanism for recognizing and analyzing the user's emotional state in real time through analysis of the user's facial expressions and voice.

[0335] "Dynamic adjustment" is the process of changing or adding report content in real time according to the user's emotional state, a function that deepens user understanding.

[0336] MODE FOR CARRYING OUT THE INVENTION

[0337] This invention is a system for improving the efficiency of the anticancer drug design and evaluation process, collecting feedback and adjusting reports while taking user emotions into consideration. Specific methods for implementing this system, as well as the hardware and software used, are described below.

[0338] 1. Data Collection

[0339] The server collects biological information such as chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL. The collection is done using an API, and the acquired data is temporarily stored in local storage.

[0340] Examples:

[0341] The server uses the Python "requests" library to call the PubChem API and retrieves chemical structure information in JSON format. The retrieved data is stored in the " / data / pubchem / " directory.

[0342] 2. Data Preprocessing

[0343] The server cleanses the collected biological information, corrects or removes incomplete data, and standardizes chemical structure information into SMILES format and data scale.

[0344] Examples:

[0345] The server creates a data frame using Pandas, replaces missing values ​​with NaN, converts chemical structures to SMILES format using RDKit, and scales the data to the range 0-1 using MinMaxScaler.

[0346] 3. Training the AI ​​model

[0347] The server uses the preprocessed data to train machine learning algorithms, such as generative adversarial networks (GANs) and autoencoders, and the trained models are used to generate future compounds.

[0348] Examples:

[0349] The server uses TensorFlow to build a GAN, split the data into a training set and a test set, trains the model using the training set, and saves the model as / models / compound_generator.h5.

[0350] 4. Generation of new compounds

[0351] It generates new molecular structures using a trained generative model, allowing users to specify constraints based on drug efficacy and molecular properties.

[0352] Examples:

[0353] The server receives a prompt from the user: "Generate the molecular structure of an anticancer drug with high efficacy and few side effects." It uses a trained GAN model to generate a molecule based on the user's request and saves the generated molecular structure in JSON format.

[0354] 5. Run the simulation

[0355] The server runs molecular dynamics simulations to predict the effects and side effects of the new compounds generated, using tools such as GROMACS.

[0356] Examples:

[0357] The server starts GROMACS and performs molecular simulation of the generated compound. The simulation results are analyzed and saved in the " / results / simulation_output" directory.

[0358] 6. Evaluation and Selection

[0359] The server analyzes the simulation results and selects the optimal compound, using criteria such as efficacy, minimal side effects, and molecular stability.

[0360] Examples:

[0361] The server uses Python's "SciPy" to perform statistical analysis of the data and select the optimal compounds. The selection results are saved as "selected_compounds.csv."

[0362] 7. Report Generation

[0363] The server generates a report containing detailed analysis results for the selected compounds, including visualization of the data.

[0364] Examples:

[0365] The server uses Matplotlib to visualize the generated compound structures and simulation results as graphs and figures, and the generated report is saved in PDF format in / reports / compound_report.pdf.

[0366] 8. Report viewing and feedback collection

[0367] The terminal provides the generated report to the user and collects feedback.

[0368] Examples:

[0369] The terminal uses a web browser to display a user interface, allowing users to view reports, and a feedback form allows users to enter comments and suggestions.

[0370] 9. Feedback Analysis

[0371] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights from it.

[0372] Examples:

[0373] The server uses natural language processing (NLP) technology to analyze the feedback text and extract emotional data.

[0374] 10. Dynamically Adjusting Reports

[0375] The terminal recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[0376] Examples:

[0377] The device uses OpenCV and Microsoft® Azure® Cognitive Services to analyze the user's facial expressions and voice, and if the user expresses anxiety, a detailed explanation and additional illustrations are dynamically added to the report.

[0378] The above is a specific embodiment of the invention. By using this system, anticancer drugs can be efficiently designed and evaluated, and feedback from users' perspectives can be collected and analyzed.

[0379] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0380] Step 1: Data collection

[0381] The server collects chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL via APIs and temporarily stores them in local storage.

[0382] Input: Queries from databases such as PubChem and ChEMBL.

[0383] Output: JSON formatted data file of biometric information.

[0384] Specific behavior:

[0385] 1. The server uses the Python "requests" library to send a GET request to the PubChem API.

[0386] 2. Receive the data obtained from the API in JSON format and save it in the " / data / pubchem / " directory of local storage.

[0387] Step 2: Data Preprocessing

[0388] The server cleanses the collected biological information, corrects or removes incomplete data, standardizes chemical structure information into SMILES format, and standardizes the data scale.

[0389] Input: A JSON formatted data file of biometric information.

[0390] Output: The preprocessed data frame.

[0391] Specific behavior:

[0392] 1. The server creates a data frame using "Pandas" and replaces missing values ​​with "NaN".

[0393] 2. The server uses the "RDKit" to convert the chemical structure information into SMILES format.

[0394] 3. The server uses "MinMaxScaler" to scale the data to the range 0 to 1.

[0395] Step 3: Training the AI ​​model

[0396] The server uses the preprocessed data to train a machine learning algorithm (e.g., GAN or autoencoder).

[0397] Input: Preprocessed dataframe.

[0398] Output: A trained generative model file.

[0399] Specific behavior:

[0400] 1. The server builds a GAN model using TensorFlow.

[0401] 2. Split the data into a training set and a test set.

[0402] 3. Train the model on the training set and save the trained model as " / models / compound_generator.h5".

[0403] Step 4: Generation of new compounds

[0404] The server uses the trained generative model to generate new molecular structures.

[0405] Input: A trained generative model, a prompt from the user.

[0406] Output: New compound data in JSON format.

[0407] Specific behavior:

[0408] 1. The server receives a prompt from the user, such as "Please generate the molecular structure of an anti-cancer drug that has high efficacy and few side effects."

[0409] 2. Generate new molecules using the trained model and save the results in JSON format.

[0410] Step 5: Run the simulation

[0411] The server performs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the generated compounds.

[0412] Input: New compound data in JSON format.

[0413] Output: Simulation result data.

[0414] Specific behavior:

[0415] 1. The server sets up a molecular simulation using GROMACS.

[0416] 2. Run simulations and analyze the efficacy, side effects, and stability data of the generated molecules.

[0417] 3. Save the simulation results in the " / results / simulation_output" directory.

[0418] Step 6: Evaluation and selection

[0419] The server analyzes the simulation results and selects the optimal compound.

[0420] Input: Simulation result data.

[0421] Output: A list of optimal compounds.

[0422] Specific behavior:

[0423] 1. The server analyzes the simulation results using Python's "SciPy".

[0424] 2. Select the optimal compound based on evaluation criteria (high efficacy, few side effects, molecular stability).

[0425] 3. Save the list of optimal compounds as "selected_compounds.csv".

[0426] Step 7: Generate reports

[0427] The server generates a report containing detailed analysis results of the selected compounds.

[0428] Input: A list of optimal compounds.

[0429] Output: Report in PDF format.

[0430] Specific behavior:

[0431] 1. The server uses "Matplotlib" to visualize the compound structures and simulation results.

[0432] 2. Include the analysis results and visualization data in the report and save it in PDF format at " / reports / compound_report.pdf".

[0433] Step 8: View reports and gather feedback

[0434] The terminal provides the generated report to the user and collects feedback.

[0435] Input: Report in PDF format.

[0436] Output: User feedback.

[0437] Specific behavior:

[0438] 1. The terminal displays the user interface via a web browser and presents the report.

[0439] 2. The user enters comments and suggestions using the feedback form, and the device collects the feedback.

[0440] Step 9: Analyze the feedback

[0441] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights therefrom.

[0442] Input: User feedback data.

[0443] Output: Feedback analysis results.

[0444] Specific behavior:

[0445] 1. The server uses "Natural Language Processing (NLP)" technology to analyze the feedback text along with sentiment data.

[0446] 2. Extract analytical insights based on sentiment data.

[0447] Step 10: Dynamically Adjusting Reports

[0448] The device recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[0449] Input: Feedback analysis results, real-time user emotion data.

[0450] Output: Dynamically adjusted report.

[0451] Specific behavior:

[0452] 1. The device analyzes the user's facial expressions and voice in real time using OpenCV and Microsoft Azure Cognitive Services.

[0453] 2. Dynamically add detailed explanations and additional illustrations to your reports based on feedback analysis results.

[0454] (Application example 2)

[0455] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0456] In the process of designing and evaluating new anticancer drugs, the key challenges are to efficiently and accurately collect chemical structure data and manufacturing information, preprocess the data, train generative models, generate new compounds, evaluate efficacy and side effects through simulations, select optimal candidates, and collect user feedback. In particular, it is important to analyze operator feedback based on their emotions and reflect it in improving the generative model.

[0457] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0458] In this invention, the server includes: means for collecting information related to living organisms from a database; means for preprocessing the collected information related to living organisms; means for training a generative model based on the preprocessed information using a machine learning algorithm; means for generating new compounds using the trained generative model; means for evaluating the effects and side effects of the generated new compounds through simulation; means for selecting an optimal compound based on the evaluation results; means for generating a report including the selection results; means for providing the generated report to a user; means for a robot to collect chemical structure data and manufacturing information in real time and cleanse the data; means for a robot to generate new compounds using a machine learning model based on the preprocessed data; means for a robot to evaluate the generated compounds using molecular dynamics simulation; and means for analyzing user emotions and collecting feedback using an emotion engine. This enables the design and evaluation process of new anticancer drugs to be carried out efficiently and accurately, and makes it possible to reflect feedback based on the operator's emotions in improving the generative model.

[0459] A "database" is a system that stores information systematically and allows specific data to be searched and retrieved efficiently.

[0460] "Biologically relevant information" refers to data related to biological research and medical applications, such as chemical structure data, clinical trial data, and genetic information.

[0461] "Preprocessing" refers to the process of cleansing collected data, preparing it in the required format, and standardizing it so that it can be used by machine learning models.

[0462] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications. Common examples include neural networks, decision trees, and support vector machines.

[0463] A "generative model" is a machine learning model that generates new data from given data. Examples include GANs (generative adversarial networks) and VAEs (variational autoencoders).

[0464] "Novel compounds" are chemical substances that have not previously existed and are generated using machine learning models.

[0465] "Simulation" refers to a computational method that mimics real-world phenomena in a virtual environment, such as molecular dynamics simulation.

[0466] "Evaluation" refers to predicting and analyzing the efficacy, side effects, stability, etc. of the compound produced through simulation.

[0467] A "report" is a written or electronic document that summarizes evaluation results and analysis data.

[0468] A "robot" is a mechanical device that performs physical tasks autonomously or semi-autonomously.

[0469] "Cleansing" refers to the process of processing missing values ​​in data and removing unnecessary data.

[0470] An "emotion engine" is a system that recognizes and analyzes a user's emotional state in real time. It includes facial expression analysis and voice analysis technologies.

[0471] "Feedback" refers to opinions and impressions provided by users, which are used to improve the system.

[0472] "Molecular dynamics simulation" is a simulation method that reproduces the movement and interactions of molecules using a computer.

[0473] In this invention, we develop a comprehensive system for the efficient design and evaluation of anticancer drugs. The system consists of the following main components:

[0474] 1. Data Collection and Preprocessing:

[0475] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from a database. It accesses the database via an API to obtain the necessary information. This data is temporarily stored in local storage.

[0476] The server cleanses the collected data, converts it into a consistent format (e.g., SMILES), and scales and standardizes the data to make it suitable for AI models.

[0477] 2. Training the generative AI model and generating compounds:

[0478] The server uses the preprocessed data to train machine learning algorithms (e.g., generative adversarial networks (GANs) or variational autoencoders (VAEs)). The trained models are then used to generate novel compounds.

[0479] The server uses the trained generative model to generate novel compounds, adjusting their structure based on constraints related to drug efficacy and molecular properties.

[0480] 3. Compound evaluation and selection:

[0481] The server runs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the compounds generated. These simulations require advanced computing resources and utilize parallel processing techniques.

[0482] The server analyzes the simulation results and selects the best candidates, taking into account criteria such as efficacy, minimal side effects, and molecular stability.

[0483] 4. Report Generation and User Interface:

[0484] The server creates a report based on the selection results, which includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[0485] The terminal displays the generated report to the user, who can use it to plan their next actions.

[0486] 5. Feedback collection and sentiment analysis:

[0487] The device uses a camera and microphone to recognize the user's emotions in real time when they enter feedback on the report content. The emotion engine analyzes their facial expressions and voice to determine their emotional state.

[0488] The server analyzes the feedback based on the emotion data obtained from the emotion engine and identifies areas for improvement in the generative model. This feedback improves the accuracy of the model.

[0489] Examples:

[0490] For example, a factory robot collects chemical structure data and manufacturing information in real time on the production line and cleanses the data. Based on the preprocessed data, the robot uses machine learning models to generate new compounds and evaluates them using molecular dynamics simulations. Operators provide feedback, and emotion analysis is performed via a camera and microphone. This feedback information is used to improve the system and reflect it in new compound designs.

[0491] Example prompt sentence:

[0492] data_sources = ["path / to / chemical_structure.csv", "path / to / clinical_trial.csv"]

[0493] model_path = "path / to / anticancer_drug_model.h5"

[0494] Creating a robot instance

[0495] robot = AntiCancerDrugDesignRobot(data_sources, model_path)

[0496] Data collection and preprocessing

[0497] preprocessed_data = robot.collect_and_preprocess_data()

[0498] Generation of new compounds

[0499] generated_compounds = robot.generate_new_compounds(preprocessed_data)

[0500] Running molecular simulations

[0501] simulation_results = robot.run_simulation(generated_compounds)

[0502] Emotional Feedback and Analysis

[0503] feedback_text = "These results are reassuring."

[0504] emotion_analysis = robot.analyze_emotions_and_feedback(feedback_text)

[0505] print("Emotion Analysis:", emotion_analysis)

[0506] This system not only streamlines the process of designing and evaluating anti-cancer drugs, but also enables more accurate results by incorporating emotion-based feedback.

[0507] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0508] Step 1:

[0509] Data collection

[0510] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from multiple databases. It accesses the databases based on queries specified through the API to obtain the required data. This data is temporarily stored in local storage. The input is the raw data obtained from the API, and the output is the raw data stored in local storage.

[0511] Step 2:

[0512] Data Preprocessing

[0513] The server cleanses the collected data and corrects or removes incomplete data. In particular, it converts chemical structure information into a consistent format (e.g., SMILES format) and standardizes the data, making it easier for AI models to use. The input is the raw data stored in local storage, and the output is the cleansed and standardized data.

[0514] Step 3:

[0515] Training an AI model

[0516] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or VAE). The model learns patterns in the data and gains the ability to generate novel compounds. This training requires extensive computing resources. The input is the preprocessed data, and the output is a trained generative model.

[0517] Step 4:

[0518] Generation of new compounds

[0519] The server uses the trained generative model to generate new molecular structures. During this generation process, the model is adjusted to satisfy constraints specified by the user based on drug efficacy and molecular properties. The input is the trained generative model and the constraints set by the user, and the output is data on the generated new compound.

[0520] Step 5:

[0521] Simulation execution

[0522] The server runs molecular dynamics simulations of the generated new compounds to predict their efficacy, side effects, and stability. This simulation is made more efficient through parallel processing. The input is data on the new compounds, and the output is the simulation results.

[0523] Step 6:

[0524] Evaluation and Selection

[0525] The server analyzes the simulation results and selects the optimal candidate from all the compounds generated. Evaluation criteria include efficacy, few side effects, and molecular stability. The input is the simulation results, and the output is data on the selected optimal compound.

[0526] Step 7:

[0527] Report Generation

[0528] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results. The input is the data for the selected optimal compounds, and the output is the generated report.

[0529] Step 8:

[0530] Report View

[0531] The terminal displays the report to the user, who can refer to it and plan their next action. The input is the generated report, and the output is the report screen that the user sees.

[0532] Step 9:

[0533] Feedback collection

[0534] The device collects feedback on the user's report and uses an emotion engine to analyze the user's emotional state. It uses a camera and microphone to analyze facial expressions and voice to recognize emotions in real time. The input is the user's feedback and emotion data, and the output is the analyzed emotion data.

[0535] Step 10:

[0536] Feedback Analysis

[0537] The server analyzes the feedback content based on the emotion data obtained from the emotion engine and identifies areas for improvement in the generative model. This insight is used as training data for the next generative model. The input is the analyzed emotion data, and the output is the identified areas for improvement.

[0538] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0539] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0540] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0541] [Second embodiment]

[0542] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0543] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0544] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0545] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0546] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0547] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0548] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0549] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0550] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0551] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0552] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0553] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0554] This invention is a system for efficiently designing and evaluating anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. The generated reports are provided to users, and further improvements are made based on user feedback.

[0555] Data collection and preprocessing

[0556] 1. Data Collection

[0557] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases.

[0558] For example, the server uses an API to obtain biometric information and stores it in local storage.

[0559] 2. Data Preprocessing

[0560] The server cleanses the collected data, imputes or removes missing values, and standardizes and unifies the format of the data.

[0561] For example, incomplete data is removed, converted to SMILES format, and standardized.

[0562] Training generative AI and compound generation

[0563] 1. Training the AI ​​model

[0564] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder).

[0565] The trained model is used to generate novel compounds.

[0566] 2. Generation of new compounds

[0567] The server uses the trained model to generate three-dimensional structures of new compounds.

[0568] For example, a user provides the input parameters needed to design a molecule with a particular therapeutic effect, and the server generates a new molecular structure based on those parameters.

[0569] Compound evaluation and selection

[0570] 1. Run the simulation

[0571] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of new compounds.

[0572] For example, the interactions of compounds are analyzed through simulations, and their effectiveness is quantified.

[0573] 2. Evaluation and Selection

[0574] The server analyzes the simulation results and selects the optimal compound.

[0575] For example, the server may preferentially select compounds with high efficacy and few side effects.

[0576] Report Generation and User Interface

[0577] 1. Report Generation

[0578] The server generates a report based on the evaluation results and creates a document containing the visual analysis results.

[0579] For example, the server generates a detailed analysis report in PDF or HTML format.

[0580] 2. User Interface

[0581] The terminal displays the generated report to the user, who can use it to plan their next actions.

[0582] For example, a researcher can use a terminal to log into the system and view the latest reports.

[0583] Incorporating user feedback

[0584] 1. Send Feedback

[0585] Users can input feedback on the report content and send it to the server via their terminal. This feedback is then used to improve the model.

[0586] For example, a user can input clinical trial results and opinions about a particular compound, which are then sent to the server and used as future training data.

[0587] In this way, by carrying out an integrated process from data collection to the generation, evaluation, and feedback of new compounds, the present invention makes it possible to streamline the development of new anticancer drugs and quickly provide treatments with reduced side effects.

[0588] The processing flow will be explained below.

[0589] Step 1:

[0590] Data collection

[0591] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[0592] Step 2:

[0593] Data Preprocessing

[0594] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[0595] Step 3:

[0596] Dataset preparation

[0597] The server splits the preprocessed data into training, validation, and test sets, which prevents overfitting of the model and allows for accurate performance evaluation.

[0598] Step 4:

[0599] Building generative AI models

[0600] The server uses a deep learning framework (e.g., TensorFlow or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[0601] Step 5:

[0602] Training generative AI models

[0603] The server uses the preprocessed data to train the generative AI model, evaluates the model's performance at each epoch, and optimizes the model while correcting errors.

[0604] Step 6:

[0605] Generation of new compounds

[0606] Using a trained generative AI model, the server generates new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[0607] Step 7:

[0608] Simulation execution

[0609] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[0610] Step 8:

[0611] Evaluation and Selection

[0612] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[0613] Step 9:

[0614] Report Generation

[0615] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[0616] Step 10:

[0617] Providing reports

[0618] The terminal displays the generated report to the user, who can use it to plan their next actions.

[0619] Step 11:

[0620] Collecting feedback

[0621] Users can enter feedback on the report content, which is then sent to the server via their device and used to improve the model and plan new experiments.

[0622] This specific processing step enables the system to efficiently and effectively advance the discovery and evaluation of new anti-cancer drugs.

[0623] Example 1

[0624] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0625] The current anticancer drug development process is known to be extremely time-consuming and costly. Furthermore, conventional methods make it difficult to adequately predict the efficacy and side effects of compounds, making it difficult to efficiently discover new compounds. Furthermore, there is a lack of a mechanism for quickly incorporating user feedback, which causes delays in the development cycle. The present invention aims to solve these problems and streamline the design and evaluation of new anticancer drugs.

[0626] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0627] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds by simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for the user to input feedback on the report content and send it to the system. This makes the process of designing and evaluating anticancer drugs more efficient, quickly incorporates user feedback, and enables the discovery of more effective anticancer drugs.

[0628] "Means for collecting" refers to the functions and processes for obtaining information related to a living organism from a database.

[0629] "Preprocessing means" refers to the functions and processes for cleansing collected biological information, completing or removing missing values, and standardizing and formatting data.

[0630] "Means for training a generative model" refers to the functionality and process of training a generative model based on preprocessed information using a machine learning algorithm.

[0631] "Means for generating novel compounds" refers to the functions and processes for generating three-dimensional structures of new compounds using a trained generative model.

[0632] "Means for evaluation by simulation" refers to the function and process of predicting and evaluating the effects and side effects of a generated novel compound using molecular dynamics simulation.

[0633] The "means for selecting an optimal compound" refers to a function and process for selecting an optimal compound based on the simulation results.

[0634] "Means for generating a report" refers to the function and process for generating detailed analysis results in document format (PDF, HTML, etc.) based on the selection results.

[0635] "Means of providing to users" refers to the functions and processes that make the generated reports available to users for viewing and access.

[0636] "Means for inputting feedback and sending it to the system" refers to the function and process by which a user inputs opinions and evaluations on the content of a generated report and sends that information to the system.

[0637] This invention is a system for efficiently designing and evaluating anticancer drugs, which collects biological information from a database, generates compounds using machine learning algorithms, and evaluates their efficacy and side effects. Below, we will explain in detail how to specifically implement this system.

[0638] Hardware and software used

[0639] Server: Responsible for data collection, preprocessing, training of machine learning models, generation of new compounds, simulation, evaluation, report generation, and feedback collection. It is desirable for the server to have hardware with high computing power. For example, an API is used to access the database, and local storage is used for data storage.

[0640] Device: Provides an interface for users to view reports and provide feedback. Devices can be regular computers, tablets, smartphones, etc.

[0641] Software used: Python, TensorFlow, and PyTorch are used for machine learning, and specialized software such as GROMACS and AMBER are used for molecular dynamics simulations.

[0642] Data collection and preprocessing

[0643] The server first collects information related to living organisms from a database. Specifically, it obtains chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs. Examples of databases that are collected include NCBI and PubChem. The server uses an API to obtain the data and saves it in local storage. The collected data is then preprocessed. During preprocessing, missing values ​​are filled in, the data is standardized, and the format is unified. This allows the data to be used efficiently in the next step.

[0644] Training generative AI models

[0645] The server trains a machine learning algorithm (e.g., GAN or autoencoder) based on the preprocessed data. The training data includes chemical structures, clinical trial data, genetic information, and more, enabling the generative AI model to handle a wide variety of conditions. Through this training, the server builds a highly accurate model for generating new compounds.

[0646] Generation of new compounds

[0647] When a user provides a prompt to design a molecule with a specific therapeutic effect, the server uses the trained generative AI model to generate the three-dimensional structure of the new compound. For example, a user might enter the following prompt:

[0648] "Design new anticancer drugs that are highly effective against specific cancer cells. Prioritize drugs with fewer side effects."

[0649] This causes the server to generate a new molecular structure based on the provided conditions and proceed to the next evaluation step.

[0650] Compound evaluation and selection

[0651] The server performs molecular dynamics simulations on the generated new compounds to predict and evaluate their effects and side effects. The simulations use molecular dynamics simulators such as GROMACS and AMBER to analyze the interactions and stability of the compounds. The server then selects the optimal compound based on the simulation results.

[0652] Report Generation and User Interface

[0653] The server generates a detailed analysis report based on the evaluation results and creates a document in PDF or HTML format. The generated report is sent to the terminal, where the user can view it. The user can enter feedback on the report content and send it to the server via the terminal. This feedback is used to improve the model and refine the data in the future.

[0654] In this way, a system will be constructed that efficiently designs and evaluates anticancer drugs through a series of processes.

[0655] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0656] Step 1: Data collection

[0657] The server collects information about anticancer drugs from multiple databases.

[0658] Input: Data from databases such as NCBI and PubChem (e.g., chemical structures, clinical trial data, genetic information).

[0659] Processing: Uses API to access the database and retrieve the required information.

[0660] Output: Collected data is saved to local storage.

[0661] Specific operation: The server calls the NCBI API, collects genetic information, and saves it in a local file in JSON format.

[0662] Step 2: Data Preprocessing

[0663] The server pre-processes the collected data.

[0664] Input: Data stored in local storage (chemical structures in JSON format, clinical trial data, and genetic information).

[0665] Processing: Data cleansing (implantation and removal of missing values), data standardization (e.g., conversion to SMILES format), and format unification.

[0666] Output: The preprocessed dataset.

[0667] Specific operation: The server fills in missing values ​​with 0, converts them to SMILES format, and saves them in a CSV file.

[0668] Step 3: Training the AI ​​model

[0669] The server uses the preprocessed data to train the AI ​​model.

[0670] Input: Preprocessed dataset (CSV file).

[0671] Processing: Train a model using a machine learning algorithm (e.g., GAN or autoencoder).

[0672] Output: A trained generative model.

[0673] What it does: The server uses TensorFlow to train a model on a CUDA-enabled GPU and saves the trained model.

[0674] Step 4: Generation of new compounds

[0675] The server uses the trained generative model to generate novel compounds.

[0676] Input: Prompt statements such as drug efficacy and target gene information provided by the user.

[0677] Processing: A prompt sentence is input into the generative model to generate the three-dimensional structure of the compound.

[0678] Output: Three-dimensional structure of the generated new compound (e.g., SMILES format).

[0679] Specific operation: The server parses the prompt sentence, generates a SMILES format of the compound, and saves it.

[0680] Step 5: Run the simulation

[0681] The server performs molecular dynamics simulations to predict the effects and side effects of the resulting compounds.

[0682] Input: Three-dimensional structure of the new compound (SMILES format).

[0683] Processing: Molecular dynamics simulations are performed using GROMACS and AMBER.

[0684] Output: Simulation results (evaluation indicators such as efficacy, side effects, and stability).

[0685] Specific operation: The server runs molecular dynamics simulations using GROMACS, analyzes the results, and saves them as a report.

[0686] Step 6: Evaluation and selection

[0687] The server selects the optimal compound based on the simulation results.

[0688] Input: Simulation results.

[0689] Processing: Compare the evaluation indexes of each compound and select the most suitable compound.

[0690] Output: Detailed information on the selected compound.

[0691] Specific operation: The server applies an algorithm based on the simulation results to select the optimal compound.

[0692] Step 7: Generate reports

[0693] The server generates a report based on the evaluation results.

[0694] Input: Detailed information about the selected compound.

[0695] Processing: Generate detailed analysis reports in PDF and HTML formats.

[0696] Output: The generated analysis report.

[0697] Specific operation: The server embeds information into a report template based on the selection results and saves it as a PDF file.

[0698] Step 8: User Interface

[0699] The terminal displays the generated report to the user.

[0700] Input: The generated analysis report.

[0701] Action: Display the report on the terminal screen.

[0702] Output: Users can view the report.

[0703] Specific operation: The terminal displays the latest report through a web browser, and the user views it.

[0704] Step 9: Send your feedback

[0705] The user inputs feedback on the report content and sends it to the server via the terminal.

[0706] Input: User feedback (opinions and ratings about a particular compound).

[0707] Processing: Send feedback from the device to the server.

[0708] Output: The feedback is saved on the server.

[0709] Specific operation: When a user enters their opinion into the feedback form and clicks the "Submit" button, the information is sent to the server and stored.

[0710] The above is a concrete explanation of the processing flow of this system and each processing step.

[0711] (Application example 1)

[0712] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0713] In conventional anticancer drug development, the design of new compounds, simulation evaluation, and the introduction of optimized manufacturing processes are performed separately, making the overall process extremely time-consuming and costly. Furthermore, while it is necessary to not only generate and evaluate new compounds, but also to automate and optimize the manufacturing process, there is no system in place to consistently perform these tasks. This makes it difficult to quickly bring drugs to market, and efficient methods are needed to improve the quality of treatment.

[0714] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0715] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for optimizing and automating the manufacturing process in the factory, thereby enabling the rapid development of anticancer drugs and the efficiency of the manufacturing process.

[0716] "Information related to living organisms" is a general term for data such as chemical substances that affect living organisms, genetic information, and clinical trial data.

[0717] "Preprocessing" refers to a series of processes that cleanse collected data, fill in or remove missing values, and standardize data and standardize formats.

[0718] "Machine learning algorithms" is a general term for mathematical models and methods for making future predictions and classifications based on past data.

[0719] A "generative model" is a model that is trained by a machine learning algorithm and is responsible for generating new data.

[0720] "Novel compounds" refer to chemical substances with new molecular structures that differ from existing compounds, and in particular refer to candidate substances in the development of pharmaceuticals such as anticancer drugs.

[0721] "Simulation" is a technology for virtually testing and predicting the effects and side effects of new compounds on a computer.

[0722] The "evaluation results" are information such as the efficacy and side effects of new compounds obtained through simulation, and are data for determining the usefulness of the compounds.

[0723] A "report" is a document created based on the evaluation results, and includes information on the analysis results and the optimal compound.

[0724] "Feedback" refers to opinions and additional information provided by users who review a report, and is used to improve the system.

[0725] "Manufacturing process optimization" refers to making adjustments and improvements to operate the manufacturing process efficiently, with the aim of optimizing resource allocation and reducing costs.

[0726] "Automation" refers to the process of using systems or robots to automatically perform processes that require manual operation.

[0727] This invention relates to a system for streamlining the design and evaluation of anticancer drugs, and includes a process of collecting biological information from a database, generating new compounds using machine learning algorithms, and evaluating their effects and side effects through simulation.

[0728] A specific embodiment of the system uses the following hardware and software: The server is a high-performance computer equipped with the latest GPU, and the software uses Python, TensorFlow, Keras, and other necessary scientific calculation libraries. The database has an API for collecting biological information. Simulation software such as GROMACS is also used to perform molecular dynamics simulations.

[0729] System program processing overview

[0730] 1. Data Collection and Preprocessing

[0731] The server collects bio-related information such as chemical structures, clinical trial data, and genetic information related to anticancer drugs from multiple databases via APIs. The collected data is cleansed and undergoes pre-processing such as filling in or removing missing values ​​and standardizing data formats.

[0732] 2. Training the generative AI and generating compounds

[0733] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or autoencoder) and generates novel compounds using the trained generative model, generating new molecular structures based on user-provided input parameters for designing molecules with specific therapeutic effects.

[0734] 3. Compound evaluation and selection

[0735] Molecular dynamics simulations are performed on the generated new compounds to predict their efficacy, side effects, and stability. The simulation results are analyzed to select the optimal compound with high efficacy and few side effects.

[0736] 4. Report Generation and User Interface

[0737] The server generates a report based on the evaluation results for the selected compounds. This report contains visual analysis results and is generated in PDF or HTML format. Users can view this report on their devices and plan their next actions.

[0738] 5. Incorporating User Feedback

[0739] Users can provide feedback on the generated reports, which is then sent to the server, where it is used as training data for future model improvements.

[0740] Specific examples

[0741] For example, if a user wants to create a new compound with high efficacy against a specific type of cancer cell, the server will design a chemical structure based on that request, run a molecular dynamics simulation to evaluate its effectiveness, and provide a report of the results, collecting feedback from the user so that it can be reflected in the next compound creation. This dramatically improves the efficiency of drug development.

[0742] Prompt Sentence Examples

[0743] "Using a generative AI model, propose new anti-cancer drug compounds. Evaluate the efficacy of these compounds through simulations and select the most suitable one."

[0744] As described above, this system significantly improves the speed and efficiency of anticancer drug development by consistently automating processes from data collection to the generation, evaluation, report creation, and feedback collection of new compounds.

[0745] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0746] Step 1:

[0747] The server collects biological information (e.g., chemical structures, clinical trial data, and genetic information) from multiple databases through APIs. It uses the APIs as input to obtain biological data. The data collection process involves sending requests to each database and storing the returned data in local storage. The output is a raw dataset of biological data.

[0748] Step 2:

[0749] The server preprocesses the collected data. First, it cleanses the data by imputing or removing missing values. Then it standardizes the data and unifies the format. The input is the raw data collected in step 1, and the output is the cleansed, standardized data. Specifically, it runs a data cleansing script to remove incomplete data and convert the remaining data into a unified format (for example, SMILES format).

[0750] Step 3:

[0751] The server uses the preprocessed data to train a machine learning algorithm. The input is the standardized data obtained in step 2. In operation, a generative AI model (e.g., a GAN or autoencoder) is trained using a machine learning library such as TensorFlow or Keras. The output is the trained generative model.

[0752] Step 4:

[0753] The server uses the trained model to generate the three-dimensional structure of a new compound. The generative model operates based on a prompt from the user (e.g., "Please design a molecule with a specific medicinal effect."). The input is the trained generative model and the user's prompt, and the output is the three-dimensional structure of the new compound. Specifically, the server inputs the prompt into the model and executes the process of generating a new compound based on it.

[0754] Step 5:

[0755] The server performs molecular dynamics simulations on new compounds to predict their efficacy, side effects, and stability. The input is the three-dimensional structure of the new compound generated in step 4. The operation uses simulation software such as GROMACS to analyze the interactions of the compounds. The output is the simulation results (data on efficacy, side effects, and stability).

[0756] Step 6:

[0757] The server analyzes the simulation results and selects the optimal compound. The input is the simulation results from step 5. The operation compares the evaluation indicators of each compound (drug efficacy, side effects, etc.) and selects the optimal compound. The output is information about the selected optimal compound.

[0758] Step 7:

[0759] The server generates a report based on the evaluation results and creates a document containing the visual analysis results. The input is the selection results from step 6. Specifically, the server uses report generation software to create a document in PDF or HTML format. The output is a report.

[0760] Step 8:

[0761] The user uses a terminal to view the generated report and enter feedback. The input is the generated report and the user's feedback. The operation is to enter feedback through the terminal's UI and send it to the server. The output is the collected feedback data.

[0762] Step 9:

[0763] The server collects feedback from users and uses it to improve the model next time. The input is the feedback data entered by the user. In operation, the feedback data is analyzed and stored as new training data. The output is the improved training data for the next time.

[0764] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0765] This invention is a system for streamlining the design and evaluation of anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. Furthermore, it aims to improve the quality of feedback by combining it with an emotion engine that recognizes the user's emotions.

[0766] Data collection and preprocessing

[0767] 1. Data Collection

[0768] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[0769] 2. Data Preprocessing

[0770] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[0771] Training generative AI and compound generation

[0772] 1. Training the AI ​​model

[0773] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder). The trained model is then used to generate novel compounds.

[0774] 2. Generation of new compounds

[0775] The server uses the trained model to generate new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[0776] Compound evaluation and selection

[0777] 1. Run the simulation

[0778] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[0779] 2. Evaluation and Selection

[0780] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[0781] Report Generation and User Interface

[0782] 1. Report Generation

[0783] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[0784] 2. User Interface

[0785] The terminal displays the generated report to the user, who can use it to plan their next actions.

[0786] Emotion engine integration

[0787] 1. Gathering feedback

[0788] When a user enters feedback on a report, the device recognizes the user's emotions in real time through an emotion engine, which uses technologies such as facial expression analysis and voice analysis to analyze the user's emotional state.

[0789] 2. Feedback Analysis

[0790] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in improving the generative model and designing new experiments.

[0791] 3. Dynamically Adjusting Reports

[0792] The device recognizes the user's emotions in real time and dynamically adjusts the content of the report it presents based on their emotional state. For example, if the user expresses anxiety, it will respond by adding more detailed explanations and illustrations.

[0793] Specific examples

[0794] 1. Specific examples of emotion recognition

[0795] When a user logs in to the system and views a report, the device uses a camera and microphone to monitor the user's facial expressions and voice. For example, if the user frowns, the emotion engine recognizes anxiety or doubt, and the device displays additional explanations in the report.

[0796] 2. Specific examples of feedback

[0797] When a user provides feedback on a report, such as "This result makes me feel anxious," the emotion engine sends that feedback along with the emotion data to the server, which then uses that data to identify areas for improvement in the generative model.

[0798] In this way, by combining an emotion engine, feedback and data analysis that take user emotions into account becomes possible, further improving the efficiency of the anticancer drug design and evaluation process.

[0799] The processing flow will be explained below.

[0800] Step 1:

[0801] Data collection

[0802] The server collects chemical structures, clinical trial data, genetic information, etc. related to anticancer drugs from multiple databases via API. This collection is performed based on a specified query, and the retrieved data is stored in temporary local storage. For example, the server may retrieve chemical structure data with specific medicinal properties from PubChem.

[0803] Step 2:

[0804] Data Preprocessing

[0805] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use. For example, it removes records with missing values ​​and standardizes molecular weight and hydrophobicity index.

[0806] Step 3:

[0807] Dataset preparation

[0808] The server splits the preprocessed data into training, validation, and test sets. This prevents overfitting of the model and allows for accurate performance evaluation. For example, split 70% of the data into training, 20% for validation, and 10% for test.

[0809] Step 4:

[0810] Building generative AI models

[0811] The server uses a deep learning framework (e.g., TensorFlow or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[0812] Step 5:

[0813] Training generative AI models

[0814] The server trains the generative AI model using the prepared dataset. The model's performance is evaluated at each epoch, and the model is optimized while correcting errors. As a concrete example, training is performed for 100 epochs, and early stopping techniques are used to avoid overfitting.

[0815] Step 6:

[0816] Generation of new compounds

[0817] The server generates new molecular structures using a trained generative AI model. During the generation process, the structure is adjusted to satisfy constraints specified by the user based on the drug efficacy and molecular properties. For example, to generate a molecular structure with anti-tumor activity, constraints can be set to include specific structural elements.

[0818] Step 7:

[0819] Simulation execution

[0820] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the generated compounds. Because the simulations require advanced computing resources, parallel processing is used. Specifically, software such as GROMACS is used to analyze the dynamic behavior of compounds.

[0821] Step 8:

[0822] Evaluation and Selection

[0823] The server analyzes the simulation results and selects the best candidate from all the generated compounds. Evaluation criteria include high efficacy, few side effects, and molecular stability. For example, candidate compounds are selected based on IC50 values ​​and ADMET profiles.

[0824] Step 9:

[0825] Report Generation

[0826] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, making it easy for users to understand the analysis results. For example, a report is generated that includes the 3D structure of the selected compound and its predicted drug efficacy profile.

[0827] Step 10:

[0828] Providing reports

[0829] The terminal displays the generated report to the user, who can use it to plan their next action. For example, a researcher can use the terminal to log in to the system and view the latest report in a dashboard format.

[0830] Step 11:

[0831] Collecting feedback

[0832] The user enters feedback on the report content. The device recognizes the user's emotions in real time through the emotion engine and sends the feedback to the server after taking those emotions into consideration. For example, when the user enters feedback, the emotion engine analyzes the user's facial expressions and tone of voice and records them as emotional data.

[0833] Step 12:

[0834] Analyzing and incorporating feedback

[0835] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in identifying areas for improvement in the generative model and in new experimental designs. For example, the parameter settings of a generative model that received a lot of positive feedback are reflected in the next training run.

[0836] Example 2

[0837] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0838] In the conventional anticancer drug design and evaluation process, data collection, preprocessing, model training, new compound generation, simulation, evaluation, and feedback collection and analysis are performed separately, making it difficult to improve overall efficiency and accuracy.In addition, it is difficult to improve the quality of feedback taking into account the user's emotional state and to adjust dynamic reports, resulting in a lack of support for user understanding and decision-making.

[0839] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0840] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, means for collecting feedback from the user and analyzing the user's emotions through an emotion engine, and means for dynamically adjusting the report content according to the user's emotional state. This makes it possible to streamline the anticancer drug design and evaluation process, realize feedback collection and analysis from the user's perspective, and support user understanding and decision-making.

[0841] A "database" is an information source that stores and manages information related to living organisms, such as chemical structures of anticancer drugs, clinical trial data, and genetic information.

[0842] "Information related to living organisms" refers to biological and chemical data, such as chemical structures, clinical trial data, and genetic information, that are necessary for the design and evaluation of anticancer drugs.

[0843] "Preprocessing" refers to the process of cleansing collected biological information, correcting and removing incomplete data, standardizing data, scaling data, and otherwise formatting it into a format that is easy for machine learning algorithms to use.

[0844] A "machine learning algorithm" is a mathematical method or model for pattern recognition and prediction using large amounts of data, and is used in this invention to train the generative model.

[0845] A "generative model" is a model for generating molecular structures of novel compounds using a trained machine learning algorithm.

[0846] "New compounds" are chemical substances with new molecular structures created based on data on existing anticancer drugs, and are the target substances for evaluation of their anticancer effects and side effects.

[0847] "Simulation" refers to the process of conducting molecular dynamics analysis on a computer to predict the effects, side effects, and stability of the generated new compound.

[0848] The "evaluation results" are analytical data on the efficacy, side effects, stability, etc. of the new compound obtained through simulation.

[0849] The "optimal compound" refers to a compound that is judged to have high efficacy, few side effects, and a stable molecular structure based on the evaluation results.

[0850] A "report" is a document summarizing detailed analysis results and visualization data, including the selection of optimal compounds.

[0851] "Users" refer to researchers and medical professionals who use this system to design and evaluate anticancer drugs.

[0852] "Feedback" refers to comments or suggestions that a user inputs on a generated report, and is information that can be used as a reference for improving the system or for the next action.

[0853] The "emotion engine" is a mechanism for recognizing and analyzing the user's emotional state in real time through analysis of the user's facial expressions and voice.

[0854] "Dynamic adjustment" is the process of changing or adding report content in real time according to the user's emotional state, a function that deepens user understanding.

[0855] MODE FOR CARRYING OUT THE INVENTION

[0856] This invention is a system for improving the efficiency of the anticancer drug design and evaluation process, collecting feedback and adjusting reports while taking user emotions into consideration. Specific methods for implementing this system, as well as the hardware and software used, are described below.

[0857] 1. Data Collection

[0858] The server collects biological information such as chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL. The collection is done using an API, and the acquired data is temporarily stored in local storage.

[0859] Examples:

[0860] The server uses the Python "requests" library to call the PubChem API and retrieves chemical structure information in JSON format. The retrieved data is stored in the " / data / pubchem / " directory.

[0861] 2. Data Preprocessing

[0862] The server cleanses the collected biological information, corrects or removes incomplete data, and standardizes chemical structure information into SMILES format and data scale.

[0863] Examples:

[0864] The server creates a data frame using Pandas, replaces missing values ​​with NaN, converts chemical structures to SMILES format using RDKit, and scales the data to the range 0-1 using MinMaxScaler.

[0865] 3. Training the AI ​​model

[0866] The server uses the preprocessed data to train machine learning algorithms, such as generative adversarial networks (GANs) and autoencoders, and the trained models are used to generate future compounds.

[0867] Examples:

[0868] The server uses TensorFlow to build a GAN, split the data into a training set and a test set, trains the model using the training set, and saves the model as / models / compound_generator.h5.

[0869] 4. Generation of new compounds

[0870] It generates new molecular structures using a trained generative model, allowing users to specify constraints based on drug efficacy and molecular properties.

[0871] Examples:

[0872] The server receives a prompt from the user: "Generate the molecular structure of an anticancer drug with high efficacy and few side effects." It uses a trained GAN model to generate a molecule based on the user's request and saves the generated molecular structure in JSON format.

[0873] 5. Run the simulation

[0874] The server runs molecular dynamics simulations to predict the effects and side effects of the new compounds generated, using tools such as GROMACS.

[0875] Examples:

[0876] The server starts GROMACS and performs molecular simulation of the generated compound. The simulation results are analyzed and saved in the " / results / simulation_output" directory.

[0877] 6. Evaluation and Selection

[0878] The server analyzes the simulation results and selects the optimal compound, using criteria such as efficacy, minimal side effects, and molecular stability.

[0879] Examples:

[0880] The server uses Python's "SciPy" to perform statistical analysis of the data and select the optimal compounds. The selection results are saved as "selected_compounds.csv."

[0881] 7. Report Generation

[0882] The server generates a report containing detailed analysis results for the selected compounds, including visualization of the data.

[0883] Examples:

[0884] The server uses Matplotlib to visualize the generated compound structures and simulation results as graphs and figures, and the generated report is saved in PDF format in / reports / compound_report.pdf.

[0885] 8. Report viewing and feedback collection

[0886] The terminal provides the generated report to the user and collects feedback.

[0887] Examples:

[0888] The terminal uses a web browser to display a user interface, allowing users to view reports, and a feedback form allows users to enter comments and suggestions.

[0889] 9. Feedback Analysis

[0890] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights from it.

[0891] Examples:

[0892] The server uses natural language processing (NLP) technology to analyze the feedback text and extract emotional data.

[0893] 10. Dynamically Adjusting Reports

[0894] The terminal recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[0895] Examples:

[0896] The device uses OpenCV and Microsoft Azure Cognitive Services to analyze the user's facial expressions and voice, and if the user expresses anxiety, it dynamically adds a detailed explanation and additional illustrations to the report.

[0897] The above is a specific embodiment of the invention. By using this system, anticancer drugs can be efficiently designed and evaluated, and feedback from users' perspectives can be collected and analyzed.

[0898] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0899] Step 1: Data collection

[0900] The server collects chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL via APIs and temporarily stores them in local storage.

[0901] Input: Queries from databases such as PubChem and ChEMBL.

[0902] Output: JSON formatted data file of biometric information.

[0903] Specific behavior:

[0904] 1. The server uses the Python "requests" library to send a GET request to the PubChem API.

[0905] 2. Receive the data obtained from the API in JSON format and save it in the " / data / pubchem / " directory of local storage.

[0906] Step 2: Data Preprocessing

[0907] The server cleanses the collected biological information, corrects or removes incomplete data, standardizes chemical structure information into SMILES format, and standardizes the data scale.

[0908] Input: A JSON formatted data file of biometric information.

[0909] Output: The preprocessed data frame.

[0910] Specific behavior:

[0911] 1. The server creates a data frame using "Pandas" and replaces missing values ​​with "NaN".

[0912] 2. The server uses the "RDKit" to convert the chemical structure information into SMILES format.

[0913] 3. The server uses "MinMaxScaler" to scale the data to the range 0 to 1.

[0914] Step 3: Training the AI ​​model

[0915] The server uses the preprocessed data to train a machine learning algorithm (e.g., GAN or autoencoder).

[0916] Input: Preprocessed dataframe.

[0917] Output: A trained generative model file.

[0918] Specific behavior:

[0919] 1. The server builds a GAN model using TensorFlow.

[0920] 2. Split the data into a training set and a test set.

[0921] 3. Train the model on the training set and save the trained model as " / models / compound_generator.h5".

[0922] Step 4: Generation of new compounds

[0923] The server uses the trained generative model to generate new molecular structures.

[0924] Input: A trained generative model, a prompt from the user.

[0925] Output: New compound data in JSON format.

[0926] Specific behavior:

[0927] 1. The server receives a prompt from the user, such as "Please generate the molecular structure of an anti-cancer drug that has high efficacy and few side effects."

[0928] 2. Generate new molecules using the trained model and save the results in JSON format.

[0929] Step 5: Run the simulation

[0930] The server performs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the generated compounds.

[0931] Input: New compound data in JSON format.

[0932] Output: Simulation result data.

[0933] Specific behavior:

[0934] 1. The server sets up a molecular simulation using GROMACS.

[0935] 2. Run simulations and analyze the efficacy, side effects, and stability data of the generated molecules.

[0936] 3. Save the simulation results in the " / results / simulation_output" directory.

[0937] Step 6: Evaluation and selection

[0938] The server analyzes the simulation results and selects the optimal compound.

[0939] Input: Simulation result data.

[0940] Output: A list of optimal compounds.

[0941] Specific behavior:

[0942] 1. The server analyzes the simulation results using Python's "SciPy".

[0943] 2. Select the optimal compound based on evaluation criteria (high efficacy, few side effects, molecular stability).

[0944] 3. Save the list of optimal compounds as "selected_compounds.csv".

[0945] Step 7: Generate reports

[0946] The server generates a report containing detailed analysis results of the selected compounds.

[0947] Input: A list of optimal compounds.

[0948] Output: Report in PDF format.

[0949] Specific behavior:

[0950] 1. The server uses "Matplotlib" to visualize the compound structures and simulation results.

[0951] 2. Include the analysis results and visualization data in the report and save it in PDF format at " / reports / compound_report.pdf".

[0952] Step 8: View reports and gather feedback

[0953] The terminal provides the generated report to the user and collects feedback.

[0954] Input: Report in PDF format.

[0955] Output: User feedback.

[0956] Specific behavior:

[0957] 1. The terminal displays the user interface via a web browser and presents the report.

[0958] 2. The user enters comments and suggestions using the feedback form, and the device collects the feedback.

[0959] Step 9: Analyze the feedback

[0960] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights therefrom.

[0961] Input: User feedback data.

[0962] Output: Feedback analysis results.

[0963] Specific behavior:

[0964] 1. The server uses "Natural Language Processing (NLP)" technology to analyze the feedback text along with sentiment data.

[0965] 2. Extract analytical insights based on sentiment data.

[0966] Step 10: Dynamically Adjusting Reports

[0967] The device recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[0968] Input: Feedback analysis results, real-time user emotion data.

[0969] Output: Dynamically adjusted report.

[0970] Specific behavior:

[0971] 1. The device analyzes the user's facial expressions and voice in real time using OpenCV and Microsoft Azure Cognitive Services.

[0972] 2. Dynamically add detailed explanations and additional illustrations to your reports based on feedback analysis results.

[0973] (Application example 2)

[0974] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0975] In the process of designing and evaluating new anticancer drugs, the key challenges are to efficiently and accurately collect chemical structure data and manufacturing information, preprocess the data, train generative models, generate new compounds, evaluate efficacy and side effects through simulations, select optimal candidates, and collect user feedback. In particular, it is important to analyze operator feedback based on their emotions and reflect it in improving the generative model.

[0976] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0977] In this invention, the server includes: means for collecting information related to living organisms from a database; means for preprocessing the collected information related to living organisms; means for training a generative model based on the preprocessed information using a machine learning algorithm; means for generating new compounds using the trained generative model; means for evaluating the effects and side effects of the generated new compounds through simulation; means for selecting an optimal compound based on the evaluation results; means for generating a report including the selection results; means for providing the generated report to a user; means for a robot to collect chemical structure data and manufacturing information in real time and cleanse the data; means for a robot to generate new compounds using a machine learning model based on the preprocessed data; means for a robot to evaluate the generated compounds using molecular dynamics simulation; and means for analyzing user emotions and collecting feedback using an emotion engine. This enables the design and evaluation process of new anticancer drugs to be carried out efficiently and accurately, and makes it possible to reflect feedback based on the operator's emotions in improving the generative model.

[0978] A "database" is a system that stores information systematically and allows specific data to be searched and retrieved efficiently.

[0979] "Biologically relevant information" refers to data related to biological research and medical applications, such as chemical structure data, clinical trial data, and genetic information.

[0980] "Preprocessing" refers to the process of cleansing collected data, preparing it in the required format, and standardizing it so that it can be used by machine learning models.

[0981] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications. Common examples include neural networks, decision trees, and support vector machines.

[0982] A "generative model" is a machine learning model that generates new data from given data. Examples include GANs (generative adversarial networks) and VAEs (variational autoencoders).

[0983] "Novel compounds" are chemical substances that have not previously existed and are generated using machine learning models.

[0984] "Simulation" refers to a computational method that mimics real-world phenomena in a virtual environment, such as molecular dynamics simulation.

[0985] "Evaluation" refers to predicting and analyzing the efficacy, side effects, stability, etc. of the compound produced through simulation.

[0986] A "report" is a written or electronic document that summarizes evaluation results and analysis data.

[0987] A "robot" is a mechanical device that performs physical tasks autonomously or semi-autonomously.

[0988] "Cleansing" refers to the process of processing missing values ​​in data and removing unnecessary data.

[0989] An "emotion engine" is a system that recognizes and analyzes a user's emotional state in real time. It includes facial expression analysis and voice analysis technologies.

[0990] "Feedback" refers to opinions and impressions provided by users, which are used to improve the system.

[0991] "Molecular dynamics simulation" is a simulation method that reproduces the movement and interactions of molecules using a computer.

[0992] In this invention, we develop a comprehensive system for the efficient design and evaluation of anticancer drugs. The system consists of the following main components:

[0993] 1. Data Collection and Preprocessing:

[0994] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from a database. It accesses the database via an API to obtain the necessary information. This data is temporarily stored in local storage.

[0995] The server cleanses the collected data, converts it into a consistent format (e.g., SMILES), and scales and standardizes the data to make it suitable for AI models.

[0996] 2. Training the generative AI model and generating compounds:

[0997] The server uses the preprocessed data to train machine learning algorithms (e.g., generative adversarial networks (GANs) or variational autoencoders (VAEs)). The trained models are then used to generate novel compounds.

[0998] The server uses the trained generative model to generate novel compounds, adjusting their structure based on constraints related to drug efficacy and molecular properties.

[0999] 3. Compound evaluation and selection:

[1000] The server runs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the compounds generated. These simulations require advanced computing resources and utilize parallel processing techniques.

[1001] The server analyzes the simulation results and selects the best candidates, taking into account criteria such as efficacy, minimal side effects, and molecular stability.

[1002] 4. Report Generation and User Interface:

[1003] The server creates a report based on the selection results, which includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[1004] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1005] 5. Feedback collection and sentiment analysis:

[1006] The device uses a camera and microphone to recognize the user's emotions in real time when they enter feedback on the report content. The emotion engine analyzes their facial expressions and voice to determine their emotional state.

[1007] The server analyzes the feedback based on the emotion data obtained from the emotion engine and identifies areas for improvement in the generative model. This feedback improves the accuracy of the model.

[1008] Examples:

[1009] For example, a factory robot collects chemical structure data and manufacturing information in real time on the production line and cleanses the data. Based on the preprocessed data, the robot uses machine learning models to generate new compounds and evaluates them using molecular dynamics simulations. Operators provide feedback, and emotion analysis is performed via a camera and microphone. This feedback information is used to improve the system and reflect it in new compound designs.

[1010] Example prompt sentence:

[1011] data_sources = ["path / to / chemical_structure.csv", "path / to / clinical_trial.csv"]

[1012] model_path = "path / to / anticancer_drug_model.h5"

[1013] Creating a robot instance

[1014] robot = AntiCancerDrugDesignRobot(data_sources, model_path)

[1015] Data collection and preprocessing

[1016] preprocessed_data = robot.collect_and_preprocess_data()

[1017] Generation of new compounds

[1018] generated_compounds = robot.generate_new_compounds(preprocessed_data)

[1019] Running molecular simulations

[1020] simulation_results = robot.run_simulation(generated_compounds)

[1021] Emotional Feedback and Analysis

[1022] feedback_text = "These results are reassuring."

[1023] emotion_analysis = robot.analyze_emotions_and_feedback(feedback_text)

[1024] print("Emotion Analysis:", emotion_analysis)

[1025] This system not only streamlines the process of designing and evaluating anti-cancer drugs, but also enables more accurate results by incorporating emotion-based feedback.

[1026] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1027] Step 1:

[1028] Data collection

[1029] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from multiple databases. It accesses the databases based on queries specified through the API to obtain the required data. This data is temporarily stored in local storage. The input is the raw data obtained from the API, and the output is the raw data stored in local storage.

[1030] Step 2:

[1031] Data Preprocessing

[1032] The server cleanses the collected data and corrects or removes incomplete data. In particular, it converts chemical structure information into a consistent format (e.g., SMILES format) and standardizes the data, making it easier for AI models to use. The input is the raw data stored in local storage, and the output is the cleansed and standardized data.

[1033] Step 3:

[1034] Training an AI model

[1035] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or VAE). The model learns patterns in the data and gains the ability to generate novel compounds. This training requires extensive computing resources. The input is the preprocessed data, and the output is a trained generative model.

[1036] Step 4:

[1037] Generation of new compounds

[1038] The server uses the trained generative model to generate new molecular structures. During this generation process, the model is adjusted to satisfy constraints specified by the user based on drug efficacy and molecular properties. The input is the trained generative model and the constraints set by the user, and the output is data on the generated new compound.

[1039] Step 5:

[1040] Simulation execution

[1041] The server runs molecular dynamics simulations of the generated new compounds to predict their efficacy, side effects, and stability. This simulation is made more efficient through parallel processing. The input is data on the new compounds, and the output is the simulation results.

[1042] Step 6:

[1043] Evaluation and Selection

[1044] The server analyzes the simulation results and selects the optimal candidate from all the compounds generated. Evaluation criteria include efficacy, few side effects, and molecular stability. The input is the simulation results, and the output is data on the selected optimal compound.

[1045] Step 7:

[1046] Report Generation

[1047] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results. The input is the data for the selected optimal compounds, and the output is the generated report.

[1048] Step 8:

[1049] Report View

[1050] The terminal displays the report to the user, who can refer to it and plan their next action. The input is the generated report, and the output is the report screen that the user sees.

[1051] Step 9:

[1052] Feedback collection

[1053] The device collects feedback on the user's report and uses an emotion engine to analyze the user's emotional state. It uses a camera and microphone to analyze facial expressions and voice to recognize emotions in real time. The input is the user's feedback and emotion data, and the output is the analyzed emotion data.

[1054] Step 10:

[1055] Feedback Analysis

[1056] The server analyzes the feedback content based on the emotion data obtained from the emotion engine and identifies areas for improvement in the generative model. This insight is used as training data for the next generative model. The input is the analyzed emotion data, and the output is the identified areas for improvement.

[1057] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1058] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1059] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[1060] [Third embodiment]

[1061] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1062] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[1063] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1064] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[1065] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1066] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1067] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1068] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1069] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1070] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1071] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1072] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[1073] This invention is a system for efficiently designing and evaluating anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. The generated reports are provided to users, and further improvements are made based on user feedback.

[1074] Data collection and preprocessing

[1075] 1. Data Collection

[1076] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases.

[1077] For example, the server uses an API to obtain biometric information and stores it in local storage.

[1078] 2. Data Preprocessing

[1079] The server cleanses the collected data, imputes or removes missing values, and standardizes and unifies the format of the data.

[1080] For example, incomplete data is removed, converted to SMILES format, and standardized.

[1081] Training generative AI and compound generation

[1082] 1. Training the AI ​​model

[1083] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder).

[1084] The trained model is used to generate novel compounds.

[1085] 2. Generation of new compounds

[1086] The server uses the trained model to generate three-dimensional structures of new compounds.

[1087] For example, a user provides the input parameters needed to design a molecule with a particular therapeutic effect, and the server generates a new molecular structure based on those parameters.

[1088] Compound evaluation and selection

[1089] 1. Run the simulation

[1090] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of new compounds.

[1091] For example, the interactions of compounds are analyzed through simulations, and their effectiveness is quantified.

[1092] 2. Evaluation and Selection

[1093] The server analyzes the simulation results and selects the optimal compound.

[1094] For example, the server may preferentially select compounds with high efficacy and few side effects.

[1095] Report Generation and User Interface

[1096] 1. Report Generation

[1097] The server generates a report based on the evaluation results and creates a document containing the visual analysis results.

[1098] For example, the server generates a detailed analysis report in PDF or HTML format.

[1099] 2. User Interface

[1100] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1101] For example, a researcher can use a terminal to log into the system and view the latest reports.

[1102] Incorporating user feedback

[1103] 1. Send Feedback

[1104] Users can input feedback on the report content and send it to the server via their terminal. This feedback is then used to improve the model.

[1105] For example, a user can input clinical trial results and opinions about a particular compound, which are then sent to the server and used as future training data.

[1106] In this way, by carrying out an integrated process from data collection to the generation, evaluation, and feedback of new compounds, the present invention makes it possible to streamline the development of new anticancer drugs and quickly provide treatments with reduced side effects.

[1107] The processing flow will be explained below.

[1108] Step 1:

[1109] Data collection

[1110] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[1111] Step 2:

[1112] Data Preprocessing

[1113] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[1114] Step 3:

[1115] Dataset preparation

[1116] The server splits the preprocessed data into training, validation, and test sets, which prevents overfitting of the model and allows for accurate performance evaluation.

[1117] Step 4:

[1118] Building generative AI models

[1119] The server uses a deep learning framework (e.g., TensorFlow or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[1120] Step 5:

[1121] Training generative AI models

[1122] The server uses the preprocessed data to train the generative AI model, evaluates the model's performance at each epoch, and optimizes the model while correcting errors.

[1123] Step 6:

[1124] Generation of new compounds

[1125] Using a trained generative AI model, the server generates new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[1126] Step 7:

[1127] Simulation execution

[1128] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[1129] Step 8:

[1130] Evaluation and Selection

[1131] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[1132] Step 9:

[1133] Report Generation

[1134] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[1135] Step 10:

[1136] Providing reports

[1137] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1138] Step 11:

[1139] Collecting feedback

[1140] Users can enter feedback on the report content, which is then sent to the server via their device and used to improve the model and plan new experiments.

[1141] This specific processing step enables the system to efficiently and effectively advance the discovery and evaluation of new anti-cancer drugs.

[1142] Example 1

[1143] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1144] The current anticancer drug development process is known to be extremely time-consuming and costly. Furthermore, conventional methods make it difficult to adequately predict the efficacy and side effects of compounds, making it difficult to efficiently discover new compounds. Furthermore, there is a lack of a mechanism for quickly incorporating user feedback, which causes delays in the development cycle. The present invention aims to solve these problems and streamline the design and evaluation of new anticancer drugs.

[1145] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1146] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds by simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for the user to input feedback on the report content and send it to the system. This makes the process of designing and evaluating anticancer drugs more efficient, quickly incorporates user feedback, and enables the discovery of more effective anticancer drugs.

[1147] "Means for collecting" refers to the functions and processes for obtaining information related to a living organism from a database.

[1148] "Preprocessing means" refers to the functions and processes for cleansing collected biological information, completing or removing missing values, and standardizing and formatting data.

[1149] "Means for training a generative model" refers to the functionality and process of training a generative model based on preprocessed information using a machine learning algorithm.

[1150] "Means for generating novel compounds" refers to the functions and processes for generating three-dimensional structures of new compounds using a trained generative model.

[1151] "Means for evaluation by simulation" refers to the function and process of predicting and evaluating the effects and side effects of a generated novel compound using molecular dynamics simulation.

[1152] The "means for selecting an optimal compound" refers to a function and process for selecting an optimal compound based on the simulation results.

[1153] "Means for generating a report" refers to the function and process for generating detailed analysis results in document format (PDF, HTML, etc.) based on the selection results.

[1154] "Means of providing to users" refers to the functions and processes that make the generated reports available to users for viewing and access.

[1155] "Means for inputting feedback and sending it to the system" refers to the function and process by which a user inputs opinions and evaluations on the content of a generated report and sends that information to the system.

[1156] This invention is a system for efficiently designing and evaluating anticancer drugs, which collects biological information from a database, generates compounds using machine learning algorithms, and evaluates their efficacy and side effects. Below, we will explain in detail how to specifically implement this system.

[1157] Hardware and software used

[1158] Server: Responsible for data collection, preprocessing, training of machine learning models, generation of new compounds, simulation, evaluation, report generation, and feedback collection. It is desirable for the server to have hardware with high computing power. For example, an API is used to access the database, and local storage is used for data storage.

[1159] Device: Provides an interface for users to view reports and provide feedback. Devices can be regular computers, tablets, smartphones, etc.

[1160] Software used: Python, TensorFlow, and PyTorch are used for machine learning, and specialized software such as GROMACS and AMBER are used for molecular dynamics simulations.

[1161] Data collection and preprocessing

[1162] The server first collects information related to living organisms from a database. Specifically, it obtains chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs. Examples of databases that are collected include NCBI and PubChem. The server uses an API to obtain the data and saves it in local storage. The collected data is then preprocessed. During preprocessing, missing values ​​are filled in, the data is standardized, and the format is unified. This allows the data to be used efficiently in the next step.

[1163] Training generative AI models

[1164] The server trains a machine learning algorithm (e.g., GAN or autoencoder) based on the preprocessed data. The training data includes chemical structures, clinical trial data, genetic information, and more, enabling the generative AI model to handle a wide variety of conditions. Through this training, the server builds a highly accurate model for generating new compounds.

[1165] Generation of new compounds

[1166] When a user provides a prompt to design a molecule with a specific therapeutic effect, the server uses the trained generative AI model to generate the three-dimensional structure of the new compound. For example, a user might enter the following prompt:

[1167] "Design new anticancer drugs that are highly effective against specific cancer cells. Prioritize drugs with fewer side effects."

[1168] This causes the server to generate a new molecular structure based on the provided conditions and proceed to the next evaluation step.

[1169] Compound evaluation and selection

[1170] The server performs molecular dynamics simulations on the generated new compounds to predict and evaluate their effects and side effects. The simulations use molecular dynamics simulators such as GROMACS and AMBER to analyze the interactions and stability of the compounds. The server then selects the optimal compound based on the simulation results.

[1171] Report Generation and User Interface

[1172] The server generates a detailed analysis report based on the evaluation results and creates a document in PDF or HTML format. The generated report is sent to the terminal, where the user can view it. The user can enter feedback on the report content and send it to the server via the terminal. This feedback is used to improve the model and refine the data in the future.

[1173] In this way, a system will be constructed that efficiently designs and evaluates anticancer drugs through a series of processes.

[1174] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1175] Step 1: Data collection

[1176] The server collects information about anticancer drugs from multiple databases.

[1177] Input: Data from databases such as NCBI and PubChem (e.g., chemical structures, clinical trial data, genetic information).

[1178] Processing: Uses API to access the database and retrieve the required information.

[1179] Output: Collected data is saved to local storage.

[1180] Specific operation: The server calls the NCBI API, collects genetic information, and saves it in a local file in JSON format.

[1181] Step 2: Data Preprocessing

[1182] The server pre-processes the collected data.

[1183] Input: Data stored in local storage (chemical structures in JSON format, clinical trial data, and genetic information).

[1184] Processing: Data cleansing (implantation and removal of missing values), data standardization (e.g., conversion to SMILES format), and format unification.

[1185] Output: The preprocessed dataset.

[1186] Specific operation: The server fills in missing values ​​with 0, converts them to SMILES format, and saves them in a CSV file.

[1187] Step 3: Training the AI ​​model

[1188] The server uses the preprocessed data to train the AI ​​model.

[1189] Input: Preprocessed dataset (CSV file).

[1190] Processing: Train a model using a machine learning algorithm (e.g., GAN or autoencoder).

[1191] Output: A trained generative model.

[1192] What it does: The server uses TensorFlow to train a model on a CUDA-enabled GPU and saves the trained model.

[1193] Step 4: Generation of new compounds

[1194] The server uses the trained generative model to generate novel compounds.

[1195] Input: Prompt statements such as drug efficacy and target gene information provided by the user.

[1196] Processing: A prompt sentence is input into the generative model to generate the three-dimensional structure of the compound.

[1197] Output: Three-dimensional structure of the generated new compound (e.g., SMILES format).

[1198] Specific operation: The server parses the prompt sentence, generates a SMILES format of the compound, and saves it.

[1199] Step 5: Run the simulation

[1200] The server performs molecular dynamics simulations to predict the effects and side effects of the resulting compounds.

[1201] Input: Three-dimensional structure of the new compound (SMILES format).

[1202] Processing: Molecular dynamics simulations are performed using GROMACS and AMBER.

[1203] Output: Simulation results (evaluation indicators such as efficacy, side effects, and stability).

[1204] Specific operation: The server runs molecular dynamics simulations using GROMACS, analyzes the results, and saves them as a report.

[1205] Step 6: Evaluation and selection

[1206] The server selects the optimal compound based on the simulation results.

[1207] Input: Simulation results.

[1208] Processing: Compare the evaluation indexes of each compound and select the most suitable compound.

[1209] Output: Detailed information on the selected compound.

[1210] Specific operation: The server applies an algorithm based on the simulation results to select the optimal compound.

[1211] Step 7: Generate reports

[1212] The server generates a report based on the evaluation results.

[1213] Input: Detailed information about the selected compound.

[1214] Processing: Generate detailed analysis reports in PDF and HTML formats.

[1215] Output: The generated analysis report.

[1216] Specific operation: The server embeds information into a report template based on the selection results and saves it as a PDF file.

[1217] Step 8: User Interface

[1218] The terminal displays the generated report to the user.

[1219] Input: The generated analysis report.

[1220] Action: Display the report on the terminal screen.

[1221] Output: Users can view the report.

[1222] Specific operation: The terminal displays the latest report through a web browser, and the user views it.

[1223] Step 9: Send your feedback

[1224] The user inputs feedback on the report content and sends it to the server via the terminal.

[1225] Input: User feedback (opinions and ratings about a particular compound).

[1226] Processing: Send feedback from the device to the server.

[1227] Output: The feedback is saved on the server.

[1228] Specific operation: When a user enters their opinion into the feedback form and clicks the "Submit" button, the information is sent to the server and stored.

[1229] The above is a concrete explanation of the processing flow of this system and each processing step.

[1230] (Application example 1)

[1231] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1232] In conventional anticancer drug development, the design of new compounds, simulation evaluation, and the introduction of optimized manufacturing processes are performed separately, making the overall process extremely time-consuming and costly. Furthermore, while it is necessary to not only generate and evaluate new compounds, but also to automate and optimize the manufacturing process, there is no system in place to consistently perform these tasks. This makes it difficult to quickly bring drugs to market, and efficient methods are needed to improve the quality of treatment.

[1233] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1234] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for optimizing and automating the manufacturing process in the factory, thereby enabling the rapid development of anticancer drugs and the efficiency of the manufacturing process.

[1235] "Information related to living organisms" is a general term for data such as chemical substances that affect living organisms, genetic information, and clinical trial data.

[1236] "Preprocessing" refers to a series of processes that cleanse collected data, fill in or remove missing values, and standardize data and standardize formats.

[1237] "Machine learning algorithms" is a general term for mathematical models and methods for making future predictions and classifications based on past data.

[1238] A "generative model" is a model that is trained by a machine learning algorithm and is responsible for generating new data.

[1239] "Novel compounds" refer to chemical substances with new molecular structures that differ from existing compounds, and in particular refer to candidate substances in the development of pharmaceuticals such as anticancer drugs.

[1240] "Simulation" is a technology for virtually testing and predicting the effects and side effects of new compounds on a computer.

[1241] The "evaluation results" are information such as the efficacy and side effects of new compounds obtained through simulation, and are data for determining the usefulness of the compounds.

[1242] A "report" is a document created based on the evaluation results, and includes information on the analysis results and the optimal compound.

[1243] "Feedback" refers to opinions and additional information provided by users who review a report, and is used to improve the system.

[1244] "Manufacturing process optimization" refers to making adjustments and improvements to operate the manufacturing process efficiently, with the aim of optimizing resource allocation and reducing costs.

[1245] "Automation" refers to the process of using systems or robots to automatically perform processes that require manual operation.

[1246] This invention relates to a system for streamlining the design and evaluation of anticancer drugs, and includes a process of collecting biological information from a database, generating new compounds using machine learning algorithms, and evaluating their effects and side effects through simulation.

[1247] A specific embodiment of the system uses the following hardware and software: The server is a high-performance computer equipped with the latest GPU, and the software uses Python, TensorFlow, Keras, and other necessary scientific calculation libraries. The database has an API for collecting biological information. Simulation software such as GROMACS is also used to perform molecular dynamics simulations.

[1248] System program processing overview

[1249] 1. Data Collection and Preprocessing

[1250] The server collects bio-related information such as chemical structures, clinical trial data, and genetic information related to anticancer drugs from multiple databases via APIs. The collected data is cleansed and undergoes pre-processing such as filling in or removing missing values ​​and standardizing data formats.

[1251] 2. Training the generative AI and generating compounds

[1252] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or autoencoder) and generates novel compounds using the trained generative model, generating new molecular structures based on user-provided input parameters for designing molecules with specific therapeutic effects.

[1253] 3. Compound evaluation and selection

[1254] Molecular dynamics simulations are performed on the generated new compounds to predict their efficacy, side effects, and stability. The simulation results are analyzed to select the optimal compound with high efficacy and few side effects.

[1255] 4. Report Generation and User Interface

[1256] The server generates a report based on the evaluation results for the selected compounds. This report contains visual analysis results and is generated in PDF or HTML format. Users can view this report on their devices and plan their next actions.

[1257] 5. Incorporating User Feedback

[1258] Users can provide feedback on the generated reports, which is then sent to the server, where it is used as training data for future model improvements.

[1259] Specific examples

[1260] For example, if a user wants to create a new compound with high efficacy against a specific type of cancer cell, the server will design a chemical structure based on that request, run a molecular dynamics simulation to evaluate its effectiveness, and provide a report of the results, collecting feedback from the user so that it can be reflected in the next compound creation. This dramatically improves the efficiency of drug development.

[1261] Prompt Sentence Examples

[1262] "Using a generative AI model, propose new anti-cancer drug compounds. Evaluate the efficacy of these compounds through simulations and select the most suitable one."

[1263] As described above, this system significantly improves the speed and efficiency of anticancer drug development by consistently automating processes from data collection to the generation, evaluation, report creation, and feedback collection of new compounds.

[1264] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1265] Step 1:

[1266] The server collects biological information (e.g., chemical structures, clinical trial data, and genetic information) from multiple databases through APIs. It uses the APIs as input to obtain biological data. The data collection process involves sending requests to each database and storing the returned data in local storage. The output is a raw dataset of biological data.

[1267] Step 2:

[1268] The server preprocesses the collected data. First, it cleanses the data by imputing or removing missing values. Then it standardizes the data and unifies the format. The input is the raw data collected in step 1, and the output is the cleansed, standardized data. Specifically, it runs a data cleansing script to remove incomplete data and convert the remaining data into a unified format (for example, SMILES format).

[1269] Step 3:

[1270] The server uses the preprocessed data to train a machine learning algorithm. The input is the standardized data obtained in step 2. In operation, a generative AI model (e.g., a GAN or autoencoder) is trained using a machine learning library such as TensorFlow or Keras. The output is the trained generative model.

[1271] Step 4:

[1272] The server uses the trained model to generate the three-dimensional structure of a new compound. The generative model operates based on a prompt from the user (e.g., "Please design a molecule with a specific medicinal effect."). The input is the trained generative model and the user's prompt, and the output is the three-dimensional structure of the new compound. Specifically, the server inputs the prompt into the model and executes the process of generating a new compound based on it.

[1273] Step 5:

[1274] The server performs molecular dynamics simulations on new compounds to predict their efficacy, side effects, and stability. The input is the three-dimensional structure of the new compound generated in step 4. The operation uses simulation software such as GROMACS to analyze the interactions of the compounds. The output is the simulation results (data on efficacy, side effects, and stability).

[1275] Step 6:

[1276] The server analyzes the simulation results and selects the optimal compound. The input is the simulation results from step 5. The operation compares the evaluation indicators of each compound (drug efficacy, side effects, etc.) and selects the optimal compound. The output is information about the selected optimal compound.

[1277] Step 7:

[1278] The server generates a report based on the evaluation results and creates a document containing the visual analysis results. The input is the selection results from step 6. Specifically, the server uses report generation software to create a document in PDF or HTML format. The output is a report.

[1279] Step 8:

[1280] The user uses a terminal to view the generated report and enter feedback. The input is the generated report and the user's feedback. The operation is to enter feedback through the terminal's UI and send it to the server. The output is the collected feedback data.

[1281] Step 9:

[1282] The server collects feedback from users and uses it to improve the model next time. The input is the feedback data entered by the user. In operation, the feedback data is analyzed and stored as new training data. The output is the improved training data for the next time.

[1283] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1284] This invention is a system for streamlining the design and evaluation of anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. Furthermore, it aims to improve the quality of feedback by combining it with an emotion engine that recognizes the user's emotions.

[1285] Data collection and preprocessing

[1286] 1. Data Collection

[1287] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[1288] 2. Data Preprocessing

[1289] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[1290] Training generative AI and compound generation

[1291] 1. Training the AI ​​model

[1292] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder). The trained model is then used to generate novel compounds.

[1293] 2. Generation of new compounds

[1294] The server uses the trained model to generate new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[1295] Compound evaluation and selection

[1296] 1. Run the simulation

[1297] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[1298] 2. Evaluation and Selection

[1299] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[1300] Report Generation and User Interface

[1301] 1. Report Generation

[1302] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[1303] 2. User Interface

[1304] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1305] Emotion engine integration

[1306] 1. Gathering feedback

[1307] When a user enters feedback on a report, the device recognizes the user's emotions in real time through an emotion engine, which uses technologies such as facial expression analysis and voice analysis to analyze the user's emotional state.

[1308] 2. Feedback Analysis

[1309] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in improving the generative model and designing new experiments.

[1310] 3. Dynamically Adjusting Reports

[1311] The device recognizes the user's emotions in real time and dynamically adjusts the content of the report it presents based on their emotional state. For example, if the user expresses anxiety, it will respond by adding more detailed explanations and illustrations.

[1312] Specific examples

[1313] 1. Specific examples of emotion recognition

[1314] When a user logs in to the system and views a report, the device uses a camera and microphone to monitor the user's facial expressions and voice. For example, if the user frowns, the emotion engine recognizes anxiety or doubt, and the device displays additional explanations in the report.

[1315] 2. Specific examples of feedback

[1316] When a user provides feedback on a report, such as "This result makes me feel anxious," the emotion engine sends that feedback along with the emotion data to the server, which then uses that data to identify areas for improvement in the generative model.

[1317] In this way, by combining an emotion engine, feedback and data analysis that take user emotions into account becomes possible, further improving the efficiency of the anticancer drug design and evaluation process.

[1318] The processing flow will be explained below.

[1319] Step 1:

[1320] Data collection

[1321] The server collects chemical structures, clinical trial data, genetic information, etc. related to anticancer drugs from multiple databases via API. This collection is performed based on a specified query, and the retrieved data is stored in temporary local storage. For example, the server may retrieve chemical structure data with specific medicinal properties from PubChem.

[1322] Step 2:

[1323] Data Preprocessing

[1324] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use. For example, it removes records with missing values ​​and standardizes molecular weight and hydrophobicity index.

[1325] Step 3:

[1326] Dataset preparation

[1327] The server splits the preprocessed data into training, validation, and test sets. This prevents overfitting of the model and allows for accurate performance evaluation. For example, split 70% of the data into training, 20% for validation, and 10% for test.

[1328] Step 4:

[1329] Building generative AI models

[1330] The server uses a deep learning framework (e.g., TensorFlow or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[1331] Step 5:

[1332] Training generative AI models

[1333] The server trains the generative AI model using the prepared dataset. The model's performance is evaluated at each epoch, and the model is optimized while correcting errors. As a concrete example, training is performed for 100 epochs, and early stopping techniques are used to avoid overfitting.

[1334] Step 6:

[1335] Generation of new compounds

[1336] The server generates new molecular structures using a trained generative AI model. During the generation process, the structure is adjusted to satisfy constraints specified by the user based on the drug efficacy and molecular properties. For example, to generate a molecular structure with anti-tumor activity, constraints can be set to include specific structural elements.

[1337] Step 7:

[1338] Simulation execution

[1339] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the generated compounds. Because the simulations require advanced computing resources, parallel processing is used. Specifically, software such as GROMACS is used to analyze the dynamic behavior of compounds.

[1340] Step 8:

[1341] Evaluation and Selection

[1342] The server analyzes the simulation results and selects the best candidate from all the generated compounds. Evaluation criteria include high efficacy, few side effects, and molecular stability. For example, candidate compounds are selected based on IC50 values ​​and ADMET profiles.

[1343] Step 9:

[1344] Report Generation

[1345] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, making it easy for users to understand the analysis results. For example, a report is generated that includes the 3D structure of the selected compound and its predicted drug efficacy profile.

[1346] Step 10:

[1347] Providing reports

[1348] The terminal displays the generated report to the user, who can use it to plan their next action. For example, a researcher can use the terminal to log in to the system and view the latest report in a dashboard format.

[1349] Step 11:

[1350] Collecting feedback

[1351] The user enters feedback on the report content. The device recognizes the user's emotions in real time through the emotion engine and sends the feedback to the server after taking those emotions into consideration. For example, when the user enters feedback, the emotion engine analyzes the user's facial expressions and tone of voice and records them as emotional data.

[1352] Step 12:

[1353] Analyzing and incorporating feedback

[1354] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in identifying areas for improvement in the generative model and in new experimental designs. For example, the parameter settings of a generative model that received a lot of positive feedback are reflected in the next training run.

[1355] Example 2

[1356] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1357] In the conventional anticancer drug design and evaluation process, data collection, preprocessing, model training, new compound generation, simulation, evaluation, and feedback collection and analysis are performed separately, making it difficult to improve overall efficiency and accuracy.In addition, it is difficult to improve the quality of feedback taking into account the user's emotional state and to adjust dynamic reports, resulting in a lack of support for user understanding and decision-making.

[1358] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1359] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, means for collecting feedback from the user and analyzing the user's emotions through an emotion engine, and means for dynamically adjusting the report content according to the user's emotional state. This makes it possible to streamline the anticancer drug design and evaluation process, realize feedback collection and analysis from the user's perspective, and support user understanding and decision-making.

[1360] A "database" is an information source that stores and manages information related to living organisms, such as chemical structures of anticancer drugs, clinical trial data, and genetic information.

[1361] "Information related to living organisms" refers to biological and chemical data, such as chemical structures, clinical trial data, and genetic information, that are necessary for the design and evaluation of anticancer drugs.

[1362] "Preprocessing" refers to the process of cleansing collected biological information, correcting and removing incomplete data, standardizing data, scaling data, and otherwise formatting it into a format that is easy for machine learning algorithms to use.

[1363] A "machine learning algorithm" is a mathematical method or model for pattern recognition and prediction using large amounts of data, and is used in this invention to train the generative model.

[1364] A "generative model" is a model for generating molecular structures of novel compounds using a trained machine learning algorithm.

[1365] "New compounds" are chemical substances with new molecular structures created based on data on existing anticancer drugs, and are the target substances for evaluation of their anticancer effects and side effects.

[1366] "Simulation" refers to the process of conducting molecular dynamics analysis on a computer to predict the effects, side effects, and stability of the generated new compound.

[1367] The "evaluation results" are analytical data on the efficacy, side effects, stability, etc. of the new compound obtained through simulation.

[1368] The "optimal compound" refers to a compound that is judged to have high efficacy, few side effects, and a stable molecular structure based on the evaluation results.

[1369] A "report" is a document summarizing detailed analysis results and visualization data, including the selection of optimal compounds.

[1370] "Users" refer to researchers and medical professionals who use this system to design and evaluate anticancer drugs.

[1371] "Feedback" refers to comments or suggestions that a user inputs on a generated report, and is information that can be used as a reference for improving the system or for the next action.

[1372] The "emotion engine" is a mechanism for recognizing and analyzing the user's emotional state in real time through analysis of the user's facial expressions and voice.

[1373] "Dynamic adjustment" is the process of changing or adding report content in real time according to the user's emotional state, a function that deepens user understanding.

[1374] MODE FOR CARRYING OUT THE INVENTION

[1375] This invention is a system for improving the efficiency of the anticancer drug design and evaluation process, collecting feedback and adjusting reports while taking user emotions into consideration. Specific methods for implementing this system, as well as the hardware and software used, are described below.

[1376] 1. Data Collection

[1377] The server collects biological information such as chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL. The collection is done using an API, and the acquired data is temporarily stored in local storage.

[1378] Examples:

[1379] The server uses the Python "requests" library to call the PubChem API and retrieves chemical structure information in JSON format. The retrieved data is stored in the " / data / pubchem / " directory.

[1380] 2. Data Preprocessing

[1381] The server cleanses the collected biological information, corrects or removes incomplete data, and standardizes chemical structure information into SMILES format and data scale.

[1382] Examples:

[1383] The server creates a data frame using Pandas, replaces missing values ​​with NaN, converts chemical structures to SMILES format using RDKit, and scales the data to the range 0-1 using MinMaxScaler.

[1384] 3. Training the AI ​​model

[1385] The server uses the preprocessed data to train machine learning algorithms, such as generative adversarial networks (GANs) and autoencoders, and the trained models are used to generate future compounds.

[1386] Examples:

[1387] The server uses TensorFlow to build a GAN, split the data into a training set and a test set, trains the model using the training set, and saves the model as / models / compound_generator.h5.

[1388] 4. Generation of new compounds

[1389] It generates new molecular structures using a trained generative model, allowing users to specify constraints based on drug efficacy and molecular properties.

[1390] Examples:

[1391] The server receives a prompt from the user: "Generate the molecular structure of an anticancer drug with high efficacy and few side effects." It uses a trained GAN model to generate a molecule based on the user's request and saves the generated molecular structure in JSON format.

[1392] 5. Run the simulation

[1393] The server runs molecular dynamics simulations to predict the effects and side effects of the new compounds generated, using tools such as GROMACS.

[1394] Examples:

[1395] The server starts GROMACS and performs molecular simulation of the generated compound. The simulation results are analyzed and saved in the " / results / simulation_output" directory.

[1396] 6. Evaluation and Selection

[1397] The server analyzes the simulation results and selects the optimal compound, using criteria such as efficacy, minimal side effects, and molecular stability.

[1398] Examples:

[1399] The server uses Python's "SciPy" to perform statistical analysis of the data and select the optimal compounds. The selection results are saved as "selected_compounds.csv."

[1400] 7. Report Generation

[1401] The server generates a report containing detailed analysis results for the selected compounds, including visualization of the data.

[1402] Examples:

[1403] The server uses Matplotlib to visualize the generated compound structures and simulation results as graphs and figures, and the generated report is saved in PDF format in / reports / compound_report.pdf.

[1404] 8. Report viewing and feedback collection

[1405] The terminal provides the generated report to the user and collects feedback.

[1406] Examples:

[1407] The terminal uses a web browser to display a user interface, allowing users to view reports, and a feedback form allows users to enter comments and suggestions.

[1408] 9. Feedback Analysis

[1409] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights from it.

[1410] Examples:

[1411] The server uses natural language processing (NLP) technology to analyze the feedback text and extract emotional data.

[1412] 10. Dynamically Adjusting Reports

[1413] The terminal recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[1414] Examples:

[1415] The device uses OpenCV and Microsoft Azure Cognitive Services to analyze the user's facial expressions and voice, and if the user expresses anxiety, it dynamically adds a detailed explanation and additional illustrations to the report.

[1416] The above is a specific embodiment of the invention. By using this system, anticancer drugs can be efficiently designed and evaluated, and feedback from users' perspectives can be collected and analyzed.

[1417] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1418] Step 1: Data collection

[1419] The server collects chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL via APIs and temporarily stores them in local storage.

[1420] Input: Queries from databases such as PubChem and ChEMBL.

[1421] Output: JSON formatted data file of biometric information.

[1422] Specific behavior:

[1423] 1. The server uses the Python "requests" library to send a GET request to the PubChem API.

[1424] 2. Receive the data obtained from the API in JSON format and save it in the " / data / pubchem / " directory of local storage.

[1425] Step 2: Data Preprocessing

[1426] The server cleanses the collected biological information, corrects or removes incomplete data, standardizes chemical structure information into SMILES format, and standardizes the data scale.

[1427] Input: A JSON formatted data file of biometric information.

[1428] Output: The preprocessed data frame.

[1429] Specific behavior:

[1430] 1. The server creates a data frame using "Pandas" and replaces missing values ​​with "NaN".

[1431] 2. The server uses the "RDKit" to convert the chemical structure information into SMILES format.

[1432] 3. The server uses "MinMaxScaler" to scale the data to the range 0 to 1.

[1433] Step 3: Training the AI ​​model

[1434] The server uses the preprocessed data to train a machine learning algorithm (e.g., GAN or autoencoder).

[1435] Input: Preprocessed dataframe.

[1436] Output: A trained generative model file.

[1437] Specific behavior:

[1438] 1. The server builds a GAN model using TensorFlow.

[1439] 2. Split the data into a training set and a test set.

[1440] 3. Train the model on the training set and save the trained model as " / models / compound_generator.h5".

[1441] Step 4: Generation of new compounds

[1442] The server uses the trained generative model to generate new molecular structures.

[1443] Input: A trained generative model, a prompt from the user.

[1444] Output: New compound data in JSON format.

[1445] Specific behavior:

[1446] 1. The server receives a prompt from the user, such as "Please generate the molecular structure of an anti-cancer drug that has high efficacy and few side effects."

[1447] 2. Generate new molecules using the trained model and save the results in JSON format.

[1448] Step 5: Run the simulation

[1449] The server performs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the generated compounds.

[1450] Input: New compound data in JSON format.

[1451] Output: Simulation result data.

[1452] Specific behavior:

[1453] 1. The server sets up a molecular simulation using GROMACS.

[1454] 2. Run simulations and analyze the efficacy, side effects, and stability data of the generated molecules.

[1455] 3. Save the simulation results in the " / results / simulation_output" directory.

[1456] Step 6: Evaluation and selection

[1457] The server analyzes the simulation results and selects the optimal compound.

[1458] Input: Simulation result data.

[1459] Output: A list of optimal compounds.

[1460] Specific behavior:

[1461] 1. The server analyzes the simulation results using Python's "SciPy".

[1462] 2. Select the optimal compound based on evaluation criteria (high efficacy, few side effects, molecular stability).

[1463] 3. Save the list of optimal compounds as "selected_compounds.csv".

[1464] Step 7: Generate reports

[1465] The server generates a report containing detailed analysis results of the selected compounds.

[1466] Input: A list of optimal compounds.

[1467] Output: Report in PDF format.

[1468] Specific behavior:

[1469] 1. The server uses "Matplotlib" to visualize the compound structures and simulation results.

[1470] 2. Include the analysis results and visualization data in the report and save it in PDF format at " / reports / compound_report.pdf".

[1471] Step 8: View reports and gather feedback

[1472] The terminal provides the generated report to the user and collects feedback.

[1473] Input: Report in PDF format.

[1474] Output: User feedback.

[1475] Specific behavior:

[1476] 1. The terminal displays the user interface via a web browser and presents the report.

[1477] 2. The user enters comments and suggestions using the feedback form, and the device collects the feedback.

[1478] Step 9: Analyze the feedback

[1479] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights therefrom.

[1480] Input: User feedback data.

[1481] Output: Feedback analysis results.

[1482] Specific behavior:

[1483] 1. The server uses "Natural Language Processing (NLP)" technology to analyze the feedback text along with sentiment data.

[1484] 2. Extract analytical insights based on sentiment data.

[1485] Step 10: Dynamically Adjusting Reports

[1486] The device recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[1487] Input: Feedback analysis results, real-time user emotion data.

[1488] Output: Dynamically adjusted report.

[1489] Specific behavior:

[1490] 1. The device analyzes the user's facial expressions and voice in real time using OpenCV and Microsoft Azure Cognitive Services.

[1491] 2. Dynamically add detailed explanations and additional illustrations to your reports based on feedback analysis results.

[1492] (Application example 2)

[1493] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1494] In the process of designing and evaluating new anticancer drugs, the key challenges are to efficiently and accurately collect chemical structure data and manufacturing information, preprocess the data, train generative models, generate new compounds, evaluate efficacy and side effects through simulations, select optimal candidates, and collect user feedback. In particular, it is important to analyze operator feedback based on their emotions and reflect it in improving the generative model.

[1495] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1496] In this invention, the server includes: means for collecting information related to living organisms from a database; means for preprocessing the collected information related to living organisms; means for training a generative model based on the preprocessed information using a machine learning algorithm; means for generating new compounds using the trained generative model; means for evaluating the effects and side effects of the generated new compounds through simulation; means for selecting an optimal compound based on the evaluation results; means for generating a report including the selection results; means for providing the generated report to a user; means for a robot to collect chemical structure data and manufacturing information in real time and cleanse the data; means for a robot to generate new compounds using a machine learning model based on the preprocessed data; means for a robot to evaluate the generated compounds using molecular dynamics simulation; and means for analyzing user emotions and collecting feedback using an emotion engine. This enables the design and evaluation process of new anticancer drugs to be carried out efficiently and accurately, and makes it possible to reflect feedback based on the operator's emotions in improving the generative model.

[1497] A "database" is a system that stores information systematically and allows specific data to be searched and retrieved efficiently.

[1498] "Biologically relevant information" refers to data related to biological research and medical applications, such as chemical structure data, clinical trial data, and genetic information.

[1499] "Preprocessing" refers to the process of cleansing collected data, preparing it in the required format, and standardizing it so that it can be used by machine learning models.

[1500] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications. Common examples include neural networks, decision trees, and support vector machines.

[1501] A "generative model" is a machine learning model that generates new data from given data. Examples include GANs (generative adversarial networks) and VAEs (variational autoencoders).

[1502] "Novel compounds" are chemical substances that have not previously existed and are generated using machine learning models.

[1503] "Simulation" refers to a computational method that mimics real-world phenomena in a virtual environment, such as molecular dynamics simulation.

[1504] "Evaluation" refers to predicting and analyzing the efficacy, side effects, stability, etc. of the compound produced through simulation.

[1505] A "report" is a written or electronic document that summarizes evaluation results and analysis data.

[1506] A "robot" is a mechanical device that performs physical tasks autonomously or semi-autonomously.

[1507] "Cleansing" refers to the process of processing missing values ​​in data and removing unnecessary data.

[1508] An "emotion engine" is a system that recognizes and analyzes a user's emotional state in real time. It includes facial expression analysis and voice analysis technologies.

[1509] "Feedback" refers to opinions and impressions provided by users, which are used to improve the system.

[1510] "Molecular dynamics simulation" is a simulation method that reproduces the movement and interactions of molecules using a computer.

[1511] In this invention, we develop a comprehensive system for the efficient design and evaluation of anticancer drugs. The system consists of the following main components:

[1512] 1. Data Collection and Preprocessing:

[1513] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from a database. It accesses the database via an API to obtain the necessary information. This data is temporarily stored in local storage.

[1514] The server cleanses the collected data, converts it into a consistent format (e.g., SMILES), and scales and standardizes the data to make it suitable for AI models.

[1515] 2. Training the generative AI model and generating compounds:

[1516] The server uses the preprocessed data to train machine learning algorithms (e.g., generative adversarial networks (GANs) or variational autoencoders (VAEs)). The trained models are then used to generate novel compounds.

[1517] The server uses the trained generative model to generate novel compounds, adjusting their structure based on constraints related to drug efficacy and molecular properties.

[1518] 3. Compound evaluation and selection:

[1519] The server runs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the compounds generated. These simulations require advanced computing resources and utilize parallel processing techniques.

[1520] The server analyzes the simulation results and selects the best candidates, taking into account criteria such as efficacy, minimal side effects, and molecular stability.

[1521] 4. Report Generation and User Interface:

[1522] The server creates a report based on the selection results, which includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[1523] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1524] 5. Feedback collection and sentiment analysis:

[1525] The device uses a camera and microphone to recognize the user's emotions in real time when they enter feedback on the report content. The emotion engine analyzes their facial expressions and voice to determine their emotional state.

[1526] The server analyzes the feedback based on the emotion data obtained from the emotion engine and identifies areas for improvement in the generative model. This feedback improves the accuracy of the model.

[1527] Examples:

[1528] For example, a factory robot collects chemical structure data and manufacturing information in real time on the production line and cleanses the data. Based on the preprocessed data, the robot uses machine learning models to generate new compounds and evaluates them using molecular dynamics simulations. Operators provide feedback, and emotion analysis is performed via a camera and microphone. This feedback information is used to improve the system and reflect it in new compound designs.

[1529] Example prompt sentence:

[1530] data_sources = ["path / to / chemical_structure.csv", "path / to / clinical_trial.csv"]

[1531] model_path = "path / to / anticancer_drug_model.h5"

[1532] Creating a robot instance

[1533] robot = AntiCancerDrugDesignRobot(data_sources, model_path)

[1534] Data collection and preprocessing

[1535] preprocessed_data = robot.collect_and_preprocess_data()

[1536] Generation of new compounds

[1537] generated_compounds = robot.generate_new_compounds(preprocessed_data)

[1538] Running molecular simulations

[1539] simulation_results = robot.run_simulation(generated_compounds)

[1540] Emotional Feedback and Analysis

[1541] feedback_text = "These results are reassuring."

[1542] emotion_analysis = robot.analyze_emotions_and_feedback(feedback_text)

[1543] print("Emotion Analysis:", emotion_analysis)

[1544] This system not only streamlines the process of designing and evaluating anti-cancer drugs, but also enables more accurate results by incorporating emotion-based feedback.

[1545] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1546] Step 1:

[1547] Data collection

[1548] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from multiple databases. It accesses the databases based on queries specified through the API to obtain the required data. This data is temporarily stored in local storage. The input is the raw data obtained from the API, and the output is the raw data stored in local storage.

[1549] Step 2:

[1550] Data Preprocessing

[1551] The server cleanses the collected data and corrects or removes incomplete data. In particular, it converts chemical structure information into a consistent format (e.g., SMILES format) and standardizes the data, making it easier for AI models to use. The input is the raw data stored in local storage, and the output is the cleansed and standardized data.

[1552] Step 3:

[1553] Training an AI model

[1554] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or VAE). The model learns patterns in the data and gains the ability to generate novel compounds. This training requires extensive computing resources. The input is the preprocessed data, and the output is a trained generative model.

[1555] Step 4:

[1556] Generation of new compounds

[1557] The server uses the trained generative model to generate new molecular structures. During this generation process, the model is adjusted to satisfy constraints specified by the user based on drug efficacy and molecular properties. The input is the trained generative model and the constraints set by the user, and the output is data on the generated new compound.

[1558] Step 5:

[1559] Simulation execution

[1560] The server runs molecular dynamics simulations of the generated new compounds to predict their efficacy, side effects, and stability. This simulation is made more efficient through parallel processing. The input is data on the new compounds, and the output is the simulation results.

[1561] Step 6:

[1562] Evaluation and Selection

[1563] The server analyzes the simulation results and selects the optimal candidate from all the compounds generated. Evaluation criteria include efficacy, few side effects, and molecular stability. The input is the simulation results, and the output is data on the selected optimal compound.

[1564] Step 7:

[1565] Report Generation

[1566] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results. The input is the data for the selected optimal compounds, and the output is the generated report.

[1567] Step 8:

[1568] Report View

[1569] The terminal displays the report to the user, who can refer to it and plan their next action. The input is the generated report, and the output is the report screen that the user sees.

[1570] Step 9:

[1571] Feedback collection

[1572] The device collects feedback on the user's report and uses an emotion engine to analyze the user's emotional state. It uses a camera and microphone to analyze facial expressions and voice to recognize emotions in real time. The input is the user's feedback and emotion data, and the output is the analyzed emotion data.

[1573] Step 10:

[1574] Feedback Analysis

[1575] The server analyzes the feedback content based on the emotion data obtained from the emotion engine and identifies areas for improvement in the generative model. This insight is used as training data for the next generative model. The input is the analyzed emotion data, and the output is the identified areas for improvement.

[1576] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1577] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1578] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1579] [Fourth embodiment]

[1580] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1581] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1582] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1583] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1584] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1585] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1586] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1587] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1588] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1589] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1590] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1591] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1592] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1593] This invention is a system for efficiently designing and evaluating anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. The generated reports are provided to users, and further improvements are made based on user feedback.

[1594] Data collection and preprocessing

[1595] 1. Data Collection

[1596] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases.

[1597] For example, the server uses an API to obtain biometric information and stores it in local storage.

[1598] 2. Data Preprocessing

[1599] The server cleanses the collected data, imputes or removes missing values, and standardizes and unifies the format of the data.

[1600] For example, incomplete data is removed, converted to SMILES format, and standardized.

[1601] Training generative AI and compound generation

[1602] 1. Training the AI ​​model

[1603] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder).

[1604] The trained model is used to generate novel compounds.

[1605] 2. Generation of new compounds

[1606] The server uses the trained model to generate three-dimensional structures of new compounds.

[1607] For example, a user provides the input parameters needed to design a molecule with a particular therapeutic effect, and the server generates a new molecular structure based on those parameters.

[1608] Compound evaluation and selection

[1609] 1. Run the simulation

[1610] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of new compounds.

[1611] For example, the interactions of compounds are analyzed through simulations, and their effectiveness is quantified.

[1612] 2. Evaluation and Selection

[1613] The server analyzes the simulation results and selects the optimal compound.

[1614] For example, the server may preferentially select compounds with high efficacy and few side effects.

[1615] Report Generation and User Interface

[1616] 1. Report Generation

[1617] The server generates a report based on the evaluation results and creates a document containing the visual analysis results.

[1618] For example, the server generates a detailed analysis report in PDF or HTML format.

[1619] 2. User Interface

[1620] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1621] For example, a researcher can use a terminal to log into the system and view the latest reports.

[1622] Incorporating user feedback

[1623] 1. Send Feedback

[1624] Users can input feedback on the report content and send it to the server via their terminal. This feedback is then used to improve the model.

[1625] For example, a user can input clinical trial results and opinions about a particular compound, which are then sent to the server and used as future training data.

[1626] In this way, by carrying out an integrated process from data collection to the generation, evaluation, and feedback of new compounds, the present invention makes it possible to streamline the development of new anticancer drugs and quickly provide treatments with reduced side effects.

[1627] The processing flow will be explained below.

[1628] Step 1:

[1629] Data collection

[1630] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[1631] Step 2:

[1632] Data Preprocessing

[1633] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[1634] Step 3:

[1635] Dataset preparation

[1636] The server splits the preprocessed data into training, validation, and test sets, which prevents overfitting of the model and allows for accurate performance evaluation.

[1637] Step 4:

[1638] Building generative AI models

[1639] The server uses a deep learning framework (e.g., TensorFlow or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[1640] Step 5:

[1641] Training generative AI models

[1642] The server uses the preprocessed data to train the generative AI model, evaluates the model's performance at each epoch, and optimizes the model while correcting errors.

[1643] Step 6:

[1644] Generation of new compounds

[1645] Using a trained generative AI model, the server generates new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[1646] Step 7:

[1647] Simulation execution

[1648] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[1649] Step 8:

[1650] Evaluation and Selection

[1651] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[1652] Step 9:

[1653] Report Generation

[1654] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[1655] Step 10:

[1656] Providing reports

[1657] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1658] Step 11:

[1659] Collecting feedback

[1660] Users can enter feedback on the report content, which is then sent to the server via their device and used to improve the model and plan new experiments.

[1661] This specific processing step enables the system to efficiently and effectively advance the discovery and evaluation of new anti-cancer drugs.

[1662] Example 1

[1663] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1664] The current anticancer drug development process is known to be extremely time-consuming and costly. Furthermore, conventional methods make it difficult to adequately predict the efficacy and side effects of compounds, making it difficult to efficiently discover new compounds. Furthermore, there is a lack of a mechanism for quickly incorporating user feedback, which causes delays in the development cycle. The present invention aims to solve these problems and streamline the design and evaluation of new anticancer drugs.

[1665] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1666] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds by simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for the user to input feedback on the report content and send it to the system. This makes the process of designing and evaluating anticancer drugs more efficient, quickly incorporates user feedback, and enables the discovery of more effective anticancer drugs.

[1667] "Means for collecting" refers to the functions and processes for obtaining information related to a living organism from a database.

[1668] "Preprocessing means" refers to the functions and processes for cleansing collected biological information, completing or removing missing values, and standardizing and formatting data.

[1669] "Means for training a generative model" refers to the functionality and process of training a generative model based on preprocessed information using a machine learning algorithm.

[1670] "Means for generating novel compounds" refers to the functions and processes for generating three-dimensional structures of new compounds using a trained generative model.

[1671] "Means for evaluation by simulation" refers to the function and process of predicting and evaluating the effects and side effects of a generated novel compound using molecular dynamics simulation.

[1672] The "means for selecting an optimal compound" refers to a function and process for selecting an optimal compound based on the simulation results.

[1673] "Means for generating a report" refers to the function and process for generating detailed analysis results in document format (PDF, HTML, etc.) based on the selection results.

[1674] "Means of providing to users" refers to the functions and processes that make the generated reports available to users for viewing and access.

[1675] "Means for inputting feedback and sending it to the system" refers to the function and process by which a user inputs opinions and evaluations on the content of a generated report and sends that information to the system.

[1676] This invention is a system for efficiently designing and evaluating anticancer drugs, which collects biological information from a database, generates compounds using machine learning algorithms, and evaluates their efficacy and side effects. Below, we will explain in detail how to specifically implement this system.

[1677] Hardware and software used

[1678] Server: Responsible for data collection, preprocessing, training of machine learning models, generation of new compounds, simulation, evaluation, report generation, and feedback collection. It is desirable for the server to have hardware with high computing power. For example, an API is used to access the database, and local storage is used for data storage.

[1679] Device: Provides an interface for users to view reports and provide feedback. Devices can be regular computers, tablets, smartphones, etc.

[1680] Software used: Python, TensorFlow, and PyTorch are used for machine learning, and specialized software such as GROMACS and AMBER are used for molecular dynamics simulations.

[1681] Data collection and preprocessing

[1682] The server first collects information related to living organisms from a database. Specifically, it obtains chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs. Examples of databases that are collected include NCBI and PubChem. The server uses an API to obtain the data and saves it in local storage. The collected data is then preprocessed. During preprocessing, missing values ​​are filled in, the data is standardized, and the format is unified. This allows the data to be used efficiently in the next step.

[1683] Training generative AI models

[1684] The server trains a machine learning algorithm (e.g., GAN or autoencoder) based on the preprocessed data. The training data includes chemical structures, clinical trial data, genetic information, and more, enabling the generative AI model to handle a wide variety of conditions. Through this training, the server builds a highly accurate model for generating new compounds.

[1685] Generation of new compounds

[1686] When a user provides a prompt to design a molecule with a specific therapeutic effect, the server uses the trained generative AI model to generate the three-dimensional structure of the new compound. For example, a user might enter the following prompt:

[1687] "Design new anticancer drugs that are highly effective against specific cancer cells. Prioritize drugs with fewer side effects."

[1688] This causes the server to generate a new molecular structure based on the provided conditions and proceed to the next evaluation step.

[1689] Compound evaluation and selection

[1690] The server performs molecular dynamics simulations on the generated new compounds to predict and evaluate their effects and side effects. The simulations use molecular dynamics simulators such as GROMACS and AMBER to analyze the interactions and stability of the compounds. The server then selects the optimal compound based on the simulation results.

[1691] Report Generation and User Interface

[1692] The server generates a detailed analysis report based on the evaluation results and creates a document in PDF or HTML format. The generated report is sent to the terminal, where the user can view it. The user can enter feedback on the report content and send it to the server via the terminal. This feedback is used to improve the model and refine the data in the future.

[1693] In this way, a system will be constructed that efficiently designs and evaluates anticancer drugs through a series of processes.

[1694] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1695] Step 1: Data collection

[1696] The server collects information about anticancer drugs from multiple databases.

[1697] Input: Data from databases such as NCBI and PubChem (e.g., chemical structures, clinical trial data, genetic information).

[1698] Processing: Uses API to access the database and retrieve the required information.

[1699] Output: Collected data is saved to local storage.

[1700] Specific operation: The server calls the NCBI API, collects genetic information, and saves it in a local file in JSON format.

[1701] Step 2: Data Preprocessing

[1702] The server pre-processes the collected data.

[1703] Input: Data stored in local storage (chemical structures in JSON format, clinical trial data, and genetic information).

[1704] Processing: Data cleansing (implantation and removal of missing values), data standardization (e.g., conversion to SMILES format), and format unification.

[1705] Output: The preprocessed dataset.

[1706] Specific operation: The server fills in missing values ​​with 0, converts them to SMILES format, and saves them in a CSV file.

[1707] Step 3: Training the AI ​​model

[1708] The server uses the preprocessed data to train the AI ​​model.

[1709] Input: Preprocessed dataset (CSV file).

[1710] Processing: Train a model using a machine learning algorithm (e.g., GAN or autoencoder).

[1711] Output: A trained generative model.

[1712] What it does: The server uses TensorFlow to train a model on a CUDA-enabled GPU and saves the trained model.

[1713] Step 4: Generation of new compounds

[1714] The server uses the trained generative model to generate novel compounds.

[1715] Input: Prompt statements such as drug efficacy and target gene information provided by the user.

[1716] Processing: A prompt sentence is input into the generative model to generate the three-dimensional structure of the compound.

[1717] Output: Three-dimensional structure of the generated new compound (e.g., SMILES format).

[1718] Specific operation: The server parses the prompt sentence, generates a SMILES format of the compound, and saves it.

[1719] Step 5: Run the simulation

[1720] The server performs molecular dynamics simulations to predict the effects and side effects of the resulting compounds.

[1721] Input: Three-dimensional structure of the new compound (SMILES format).

[1722] Processing: Molecular dynamics simulations are performed using GROMACS and AMBER.

[1723] Output: Simulation results (evaluation indicators such as efficacy, side effects, and stability).

[1724] Specific operation: The server runs molecular dynamics simulations using GROMACS, analyzes the results, and saves them as a report.

[1725] Step 6: Evaluation and selection

[1726] The server selects the optimal compound based on the simulation results.

[1727] Input: Simulation results.

[1728] Processing: Compare the evaluation indexes of each compound and select the most suitable compound.

[1729] Output: Detailed information on the selected compound.

[1730] Specific operation: The server applies an algorithm based on the simulation results to select the optimal compound.

[1731] Step 7: Generate reports

[1732] The server generates a report based on the evaluation results.

[1733] Input: Detailed information about the selected compound.

[1734] Processing: Generate detailed analysis reports in PDF and HTML formats.

[1735] Output: The generated analysis report.

[1736] Specific operation: The server embeds information into a report template based on the selection results and saves it as a PDF file.

[1737] Step 8: User Interface

[1738] The terminal displays the generated report to the user.

[1739] Input: The generated analysis report.

[1740] Action: Display the report on the terminal screen.

[1741] Output: Users can view the report.

[1742] Specific operation: The terminal displays the latest report through a web browser, and the user views it.

[1743] Step 9: Send your feedback

[1744] The user inputs feedback on the report content and sends it to the server via the terminal.

[1745] Input: User feedback (opinions and ratings about a particular compound).

[1746] Processing: Send feedback from the device to the server.

[1747] Output: The feedback is saved on the server.

[1748] Specific operation: When a user enters their opinion into the feedback form and clicks the "Submit" button, the information is sent to the server and stored.

[1749] The above is a concrete explanation of the processing flow of this system and each processing step.

[1750] (Application example 1)

[1751] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1752] In conventional anticancer drug development, the design of new compounds, simulation evaluation, and the introduction of optimized manufacturing processes are performed separately, making the overall process extremely time-consuming and costly. Furthermore, while it is necessary to not only generate and evaluate new compounds, but also to automate and optimize the manufacturing process, there is no system in place to consistently perform these tasks. This makes it difficult to quickly bring drugs to market, and efficient methods are needed to improve the quality of treatment.

[1753] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1754] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, and means for optimizing and automating the manufacturing process in the factory, thereby enabling the rapid development of anticancer drugs and the efficiency of the manufacturing process.

[1755] "Information related to living organisms" is a general term for data such as chemical substances that affect living organisms, genetic information, and clinical trial data.

[1756] "Preprocessing" refers to a series of processes that cleanse collected data, fill in or remove missing values, and standardize data and standardize formats.

[1757] "Machine learning algorithms" is a general term for mathematical models and methods for making future predictions and classifications based on past data.

[1758] A "generative model" is a model that is trained by a machine learning algorithm and is responsible for generating new data.

[1759] "Novel compounds" refer to chemical substances with new molecular structures that differ from existing compounds, and in particular refer to candidate substances in the development of pharmaceuticals such as anticancer drugs.

[1760] "Simulation" is a technology for virtually testing and predicting the effects and side effects of new compounds on a computer.

[1761] The "evaluation results" are information such as the efficacy and side effects of new compounds obtained through simulation, and are data for determining the usefulness of the compounds.

[1762] A "report" is a document created based on the evaluation results, and includes information on the analysis results and the optimal compound.

[1763] "Feedback" refers to opinions and additional information provided by users who review a report, and is used to improve the system.

[1764] "Manufacturing process optimization" refers to making adjustments and improvements to operate the manufacturing process efficiently, with the aim of optimizing resource allocation and reducing costs.

[1765] "Automation" refers to the process of using systems or robots to automatically perform processes that require manual operation.

[1766] This invention relates to a system for streamlining the design and evaluation of anticancer drugs, and includes a process of collecting biological information from a database, generating new compounds using machine learning algorithms, and evaluating their effects and side effects through simulation.

[1767] A specific embodiment of the system uses the following hardware and software: The server is a high-performance computer equipped with the latest GPU, and the software uses Python, TensorFlow, Keras, and other necessary scientific calculation libraries. The database has an API for collecting biological information. Simulation software such as GROMACS is also used to perform molecular dynamics simulations.

[1768] System program processing overview

[1769] 1. Data Collection and Preprocessing

[1770] The server collects bio-related information such as chemical structures, clinical trial data, and genetic information related to anticancer drugs from multiple databases via APIs. The collected data is cleansed and undergoes pre-processing such as filling in or removing missing values ​​and standardizing data formats.

[1771] 2. Training the generative AI and generating compounds

[1772] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or autoencoder) and generates novel compounds using the trained generative model, generating new molecular structures based on user-provided input parameters for designing molecules with specific therapeutic effects.

[1773] 3. Compound evaluation and selection

[1774] Molecular dynamics simulations are performed on the generated new compounds to predict their efficacy, side effects, and stability. The simulation results are analyzed to select the optimal compound with high efficacy and few side effects.

[1775] 4. Report Generation and User Interface

[1776] The server generates a report based on the evaluation results for the selected compounds. This report contains visual analysis results and is generated in PDF or HTML format. Users can view this report on their devices and plan their next actions.

[1777] 5. Incorporating User Feedback

[1778] Users can provide feedback on the generated reports, which is then sent to the server, where it is used as training data for future model improvements.

[1779] Specific examples

[1780] For example, if a user wants to create a new compound with high efficacy against a specific type of cancer cell, the server will design a chemical structure based on that request, run a molecular dynamics simulation to evaluate its effectiveness, and provide a report of the results, collecting feedback from the user so that it can be reflected in the next compound creation. This dramatically improves the efficiency of drug development.

[1781] Prompt Sentence Examples

[1782] "Using a generative AI model, propose new anti-cancer drug compounds. Evaluate the efficacy of these compounds through simulations and select the most suitable one."

[1783] As described above, this system significantly improves the speed and efficiency of anticancer drug development by consistently automating processes from data collection to the generation, evaluation, report creation, and feedback collection of new compounds.

[1784] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1785] Step 1:

[1786] The server collects biological information (e.g., chemical structures, clinical trial data, and genetic information) from multiple databases through APIs. It uses the APIs as input to obtain biological data. The data collection process involves sending requests to each database and storing the returned data in local storage. The output is a raw dataset of biological data.

[1787] Step 2:

[1788] The server preprocesses the collected data. First, it cleanses the data by imputing or removing missing values. Then it standardizes the data and unifies the format. The input is the raw data collected in step 1, and the output is the cleansed, standardized data. Specifically, it runs a data cleansing script to remove incomplete data and convert the remaining data into a unified format (for example, SMILES format).

[1789] Step 3:

[1790] The server uses the preprocessed data to train a machine learning algorithm. The input is the standardized data obtained in step 2. In operation, a generative AI model (e.g., a GAN or autoencoder) is trained using a machine learning library such as TensorFlow or Keras. The output is the trained generative model.

[1791] Step 4:

[1792] The server uses the trained model to generate the three-dimensional structure of a new compound. The generative model operates based on a prompt from the user (e.g., "Please design a molecule with a specific medicinal effect."). The input is the trained generative model and the user's prompt, and the output is the three-dimensional structure of the new compound. Specifically, the server inputs the prompt into the model and executes the process of generating a new compound based on it.

[1793] Step 5:

[1794] The server performs molecular dynamics simulations on new compounds to predict their efficacy, side effects, and stability. The input is the three-dimensional structure of the new compound generated in step 4. The operation uses simulation software such as GROMACS to analyze the interactions of the compounds. The output is the simulation results (data on efficacy, side effects, and stability).

[1795] Step 6:

[1796] The server analyzes the simulation results and selects the optimal compound. The input is the simulation results from step 5. The operation compares the evaluation indicators of each compound (drug efficacy, side effects, etc.) and selects the optimal compound. The output is information about the selected optimal compound.

[1797] Step 7:

[1798] The server generates a report based on the evaluation results and creates a document containing the visual analysis results. The input is the selection results from step 6. Specifically, the server uses report generation software to create a document in PDF or HTML format. The output is a report.

[1799] Step 8:

[1800] The user uses a terminal to view the generated report and enter feedback. The input is the generated report and the user's feedback. The operation is to enter feedback through the terminal's UI and send it to the server. The output is the collected feedback data.

[1801] Step 9:

[1802] The server collects feedback from users and uses it to improve the model next time. The input is the feedback data entered by the user. In operation, the feedback data is analyzed and stored as new training data. The output is the improved training data for the next time.

[1803] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1804] This invention is a system for streamlining the design and evaluation of anticancer drugs. It collects and preprocesses biological information from a database, generates new compounds using machine learning algorithms, and evaluates their effects and side effects through simulations. Furthermore, it aims to improve the quality of feedback by combining it with an emotion engine that recognizes the user's emotions.

[1805] Data collection and preprocessing

[1806] 1. Data Collection

[1807] The server collects chemical structures, clinical trial data, genetic information, and other information related to anticancer drugs from multiple databases via APIs. This collection is performed based on specified queries, and the acquired data is temporarily stored in local storage.

[1808] 2. Data Preprocessing

[1809] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use.

[1810] Training generative AI and compound generation

[1811] 1. Training the AI ​​model

[1812] The server uses the preprocessed data to train a machine learning algorithm (e.g., a GAN or autoencoder). The trained model is then used to generate novel compounds.

[1813] 2. Generation of new compounds

[1814] The server uses the trained model to generate new molecular structures, adjusting them to meet user-specified constraints based on drug efficacy and molecular properties.

[1815] Compound evaluation and selection

[1816] 1. Run the simulation

[1817] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the compounds created. Because the simulations require advanced computing resources, parallel processing is often used.

[1818] 2. Evaluation and Selection

[1819] The server analyzes the simulation results and selects the best candidate from all the compounds generated, based on evaluation criteria such as efficacy, few side effects, and molecular stability.

[1820] Report Generation and User Interface

[1821] 1. Report Generation

[1822] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[1823] 2. User Interface

[1824] The terminal displays the generated report to the user, who can use it to plan their next actions.

[1825] Emotion engine integration

[1826] 1. Gathering feedback

[1827] When a user enters feedback on a report, the device recognizes the user's emotions in real time through an emotion engine, which uses technologies such as facial expression analysis and voice analysis to analyze the user's emotional state.

[1828] 2. Feedback Analysis

[1829] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in improving the generative model and designing new experiments.

[1830] 3. Dynamically Adjusting Reports

[1831] The device recognizes the user's emotions in real time and dynamically adjusts the content of the report it presents based on their emotional state. For example, if the user expresses anxiety, it will respond by adding more detailed explanations and illustrations.

[1832] Specific examples

[1833] 1. Specific examples of emotion recognition

[1834] When a user logs in to the system and views a report, the device uses a camera and microphone to monitor the user's facial expressions and voice. For example, if the user frowns, the emotion engine recognizes anxiety or doubt, and the device displays additional explanations in the report.

[1835] 2. Specific examples of feedback

[1836] When a user provides feedback on a report, such as "This result makes me feel anxious," the emotion engine sends that feedback along with the emotion data to the server, which then uses that data to identify areas for improvement in the generative model.

[1837] In this way, by combining an emotion engine, feedback and data analysis that take user emotions into account becomes possible, further improving the efficiency of the anticancer drug design and evaluation process.

[1838] The processing flow will be explained below.

[1839] Step 1:

[1840] Data collection

[1841] The server collects chemical structures, clinical trial data, genetic information, etc. related to anticancer drugs from multiple databases via API. This collection is performed based on a specified query, and the retrieved data is stored in temporary local storage. For example, the server may retrieve chemical structure data with specific medicinal properties from PubChem.

[1842] Step 2:

[1843] Data Preprocessing

[1844] The server cleanses the collected data, correcting or removing incomplete data and converting chemical structure information into a unified format (e.g., SMILES format). It also standardizes the scale of the data and prepares it in a format that is easy for AI models to use. For example, it removes records with missing values ​​and standardizes molecular weight and hydrophobicity index.

[1845] Step 3:

[1846] Dataset preparation

[1847] The server splits the preprocessed data into training, validation, and test sets. This prevents overfitting of the model and allows for accurate performance evaluation. For example, split 70% of the data into training, 20% for validation, and 10% for test.

[1848] Step 4:

[1849] Building generative AI models

[1850] The server uses a deep learning framework (e.g., TensorFlow or PyTorch) to build a generative AI model (e.g., a GAN or autoencoder), which is then used to generate novel compounds.

[1851] Step 5:

[1852] Training generative AI models

[1853] The server trains the generative AI model using the prepared dataset. The model's performance is evaluated at each epoch, and the model is optimized while correcting errors. As a concrete example, training is performed for 100 epochs, and early stopping techniques are used to avoid overfitting.

[1854] Step 6:

[1855] Generation of new compounds

[1856] The server generates new molecular structures using a trained generative AI model. During the generation process, the structure is adjusted to satisfy constraints specified by the user based on the drug efficacy and molecular properties. For example, to generate a molecular structure with anti-tumor activity, constraints can be set to include specific structural elements.

[1857] Step 7:

[1858] Simulation execution

[1859] The server performs molecular dynamics simulations to predict the efficacy, side effects, and stability of the generated compounds. Because the simulations require advanced computing resources, parallel processing is used. Specifically, software such as GROMACS is used to analyze the dynamic behavior of compounds.

[1860] Step 8:

[1861] Evaluation and Selection

[1862] The server analyzes the simulation results and selects the best candidate from all the generated compounds. Evaluation criteria include high efficacy, few side effects, and molecular stability. For example, candidate compounds are selected based on IC50 values ​​and ADMET profiles.

[1863] Step 9:

[1864] Report Generation

[1865] The server creates a report based on the selection results. The report includes detailed analysis results and visualized data for the selected compounds, making it easy for users to understand the analysis results. For example, a report is generated that includes the 3D structure of the selected compound and its predicted drug efficacy profile.

[1866] Step 10:

[1867] Providing reports

[1868] The terminal displays the generated report to the user, who can use it to plan their next action. For example, a researcher can use the terminal to log in to the system and view the latest report in a dashboard format.

[1869] Step 11:

[1870] Collecting feedback

[1871] The user enters feedback on the report content. The device recognizes the user's emotions in real time through the emotion engine and sends the feedback to the server after taking those emotions into consideration. For example, when the user enters feedback, the emotion engine analyzes the user's facial expressions and tone of voice and records them as emotional data.

[1872] Step 12:

[1873] Analyzing and incorporating feedback

[1874] The server analyzes the feedback based on the emotion data obtained from the emotion engine and extracts valuable insights backed by the user's emotions. These insights are reflected in identifying areas for improvement in the generative model and in new experimental designs. For example, the parameter settings of a generative model that received a lot of positive feedback are reflected in the next training run.

[1875] Example 2

[1876] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1877] In the conventional anticancer drug design and evaluation process, data collection, preprocessing, model training, new compound generation, simulation, evaluation, and feedback collection and analysis are performed separately, making it difficult to improve overall efficiency and accuracy.In addition, it is difficult to improve the quality of feedback taking into account the user's emotional state and to adjust dynamic reports, resulting in a lack of support for user understanding and decision-making.

[1878] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1879] In this invention, the server includes means for collecting information related to living organisms from a database, means for preprocessing the collected information related to living organisms, means for training a generative model based on the preprocessed information using a machine learning algorithm, means for generating new compounds using the trained generative model, means for evaluating the effects and side effects of the generated new compounds through simulation, means for selecting an optimal compound based on the evaluation results, means for generating a report including the selection results, means for providing the generated report to a user, means for collecting feedback from the user and analyzing the user's emotions through an emotion engine, and means for dynamically adjusting the report content according to the user's emotional state. This makes it possible to streamline the anticancer drug design and evaluation process, realize feedback collection and analysis from the user's perspective, and support user understanding and decision-making.

[1880] A "database" is an information source that stores and manages information related to living organisms, such as chemical structures of anticancer drugs, clinical trial data, and genetic information.

[1881] "Information related to living organisms" refers to biological and chemical data, such as chemical structures, clinical trial data, and genetic information, that are necessary for the design and evaluation of anticancer drugs.

[1882] "Preprocessing" refers to the process of cleansing collected biological information, correcting and removing incomplete data, standardizing data, scaling data, and otherwise formatting it into a format that is easy for machine learning algorithms to use.

[1883] A "machine learning algorithm" is a mathematical method or model for pattern recognition and prediction using large amounts of data, and is used in this invention to train the generative model.

[1884] A "generative model" is a model for generating molecular structures of novel compounds using a trained machine learning algorithm.

[1885] "New compounds" are chemical substances with new molecular structures created based on data on existing anticancer drugs, and are the target substances for evaluation of their anticancer effects and side effects.

[1886] "Simulation" refers to the process of conducting molecular dynamics analysis on a computer to predict the effects, side effects, and stability of the generated new compound.

[1887] The "evaluation results" are analytical data on the efficacy, side effects, stability, etc. of the new compound obtained through simulation.

[1888] The "optimal compound" refers to a compound that is judged to have high efficacy, few side effects, and a stable molecular structure based on the evaluation results.

[1889] A "report" is a document summarizing detailed analysis results and visualization data, including the selection of optimal compounds.

[1890] "Users" refer to researchers and medical professionals who use this system to design and evaluate anticancer drugs.

[1891] "Feedback" refers to comments or suggestions that a user inputs on a generated report, and is information that can be used as a reference for improving the system or for the next action.

[1892] The "emotion engine" is a mechanism for recognizing and analyzing the user's emotional state in real time through analysis of the user's facial expressions and voice.

[1893] "Dynamic adjustment" is the process of changing or adding report content in real time according to the user's emotional state, a function that deepens user understanding.

[1894] MODE FOR CARRYING OUT THE INVENTION

[1895] This invention is a system for improving the efficiency of the anticancer drug design and evaluation process, collecting feedback and adjusting reports while taking user emotions into consideration. Specific methods for implementing this system, as well as the hardware and software used, are described below.

[1896] 1. Data Collection

[1897] The server collects biological information such as chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL. The collection is done using an API, and the acquired data is temporarily stored in local storage.

[1898] Examples:

[1899] The server uses the Python "requests" library to call the PubChem API and retrieves chemical structure information in JSON format. The retrieved data is stored in the " / data / pubchem / " directory.

[1900] 2. Data Preprocessing

[1901] The server cleanses the collected biological information, corrects or removes incomplete data, and standardizes chemical structure information into SMILES format and data scale.

[1902] Examples:

[1903] The server creates a data frame using Pandas, replaces missing values ​​with NaN, converts chemical structures to SMILES format using RDKit, and scales the data to the range 0-1 using MinMaxScaler.

[1904] 3. Training the AI ​​model

[1905] The server uses the preprocessed data to train machine learning algorithms, such as generative adversarial networks (GANs) and autoencoders, and the trained models are used to generate future compounds.

[1906] Examples:

[1907] The server uses TensorFlow to build a GAN, split the data into a training set and a test set, trains the model using the training set, and saves the model as / models / compound_generator.h5.

[1908] 4. Generation of new compounds

[1909] It generates new molecular structures using a trained generative model, allowing users to specify constraints based on drug efficacy and molecular properties.

[1910] Examples:

[1911] The server receives a prompt from the user: "Generate the molecular structure of an anticancer drug with high efficacy and few side effects." It uses a trained GAN model to generate a molecule based on the user's request and saves the generated molecular structure in JSON format.

[1912] 5. Run the simulation

[1913] The server runs molecular dynamics simulations to predict the effects and side effects of the new compounds generated, using tools such as GROMACS.

[1914] Examples:

[1915] The server starts GROMACS and performs molecular simulation of the generated compound. The simulation results are analyzed and saved in the " / results / simulation_output" directory.

[1916] 6. Evaluation and Selection

[1917] The server analyzes the simulation results and selects the optimal compound, using criteria such as efficacy, minimal side effects, and molecular stability.

[1918] Examples:

[1919] The server uses Python's "SciPy" to perform statistical analysis of the data and select the optimal compounds. The selection results are saved as "selected_compounds.csv."

[1920] 7. Report Generation

[1921] The server generates a report containing detailed analysis results for the selected compounds, including visualization of the data.

[1922] Examples:

[1923] The server uses Matplotlib to visualize the generated compound structures and simulation results as graphs and figures, and the generated report is saved in PDF format in / reports / compound_report.pdf.

[1924] 8. Report viewing and feedback collection

[1925] The terminal provides the generated report to the user and collects feedback.

[1926] Examples:

[1927] The terminal uses a web browser to display a user interface, allowing users to view reports, and a feedback form allows users to enter comments and suggestions.

[1928] 9. Feedback Analysis

[1929] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights from it.

[1930] Examples:

[1931] The server uses natural language processing (NLP) technology to analyze the feedback text and extract emotional data.

[1932] 10. Dynamically Adjusting Reports

[1933] The terminal recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[1934] Examples:

[1935] The device uses OpenCV and Microsoft Azure Cognitive Services to analyze the user's facial expressions and voice, and if the user expresses anxiety, it dynamically adds a detailed explanation and additional illustrations to the report.

[1936] The above is a specific embodiment of the invention. By using this system, anticancer drugs can be efficiently designed and evaluated, and feedback from users' perspectives can be collected and analyzed.

[1937] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1938] Step 1: Data collection

[1939] The server collects chemical structures, clinical trial data, and genetic information about anticancer drugs from databases such as PubChem and ChEMBL via APIs and temporarily stores them in local storage.

[1940] Input: Queries from databases such as PubChem and ChEMBL.

[1941] Output: JSON formatted data file of biometric information.

[1942] Specific behavior:

[1943] 1. The server uses the Python "requests" library to send a GET request to the PubChem API.

[1944] 2. Receive the data obtained from the API in JSON format and save it in the " / data / pubchem / " directory of local storage.

[1945] Step 2: Data Preprocessing

[1946] The server cleanses the collected biological information, corrects or removes incomplete data, standardizes chemical structure information into SMILES format, and standardizes the data scale.

[1947] Input: A JSON formatted data file of biometric information.

[1948] Output: The preprocessed data frame.

[1949] Specific behavior:

[1950] 1. The server creates a data frame using "Pandas" and replaces missing values ​​with "NaN".

[1951] 2. The server uses the "RDKit" to convert the chemical structure information into SMILES format.

[1952] 3. The server uses "MinMaxScaler" to scale the data to the range 0 to 1.

[1953] Step 3: Training the AI ​​model

[1954] The server uses the preprocessed data to train a machine learning algorithm (e.g., GAN or autoencoder).

[1955] Input: Preprocessed dataframe.

[1956] Output: A trained generative model file.

[1957] Specific behavior:

[1958] 1. The server builds a GAN model using TensorFlow.

[1959] 2. Split the data into a training set and a test set.

[1960] 3. Train the model on the training set and save the trained model as " / models / compound_generator.h5".

[1961] Step 4: Generation of new compounds

[1962] The server uses the trained generative model to generate new molecular structures.

[1963] Input: A trained generative model, a prompt from the user.

[1964] Output: New compound data in JSON format.

[1965] Specific behavior:

[1966] 1. The server receives a prompt from the user, such as "Please generate the molecular structure of an anti-cancer drug that has high efficacy and few side effects."

[1967] 2. Generate new molecules using the trained model and save the results in JSON format.

[1968] Step 5: Run the simulation

[1969] The server performs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the generated compounds.

[1970] Input: New compound data in JSON format.

[1971] Output: Simulation result data.

[1972] Specific behavior:

[1973] 1. The server sets up a molecular simulation using GROMACS.

[1974] 2. Run simulations and analyze the efficacy, side effects, and stability data of the generated molecules.

[1975] 3. Save the simulation results in the " / results / simulation_output" directory.

[1976] Step 6: Evaluation and selection

[1977] The server analyzes the simulation results and selects the optimal compound.

[1978] Input: Simulation result data.

[1979] Output: A list of optimal compounds.

[1980] Specific behavior:

[1981] 1. The server analyzes the simulation results using Python's "SciPy".

[1982] 2. Select the optimal compound based on evaluation criteria (high efficacy, few side effects, molecular stability).

[1983] 3. Save the list of optimal compounds as "selected_compounds.csv".

[1984] Step 7: Generate reports

[1985] The server generates a report containing detailed analysis results of the selected compounds.

[1986] Input: A list of optimal compounds.

[1987] Output: Report in PDF format.

[1988] Specific behavior:

[1989] 1. The server uses "Matplotlib" to visualize the compound structures and simulation results.

[1990] 2. Include the analysis results and visualization data in the report and save it in PDF format at " / reports / compound_report.pdf".

[1991] Step 8: View reports and gather feedback

[1992] The terminal provides the generated report to the user and collects feedback.

[1993] Input: Report in PDF format.

[1994] Output: User feedback.

[1995] Specific behavior:

[1996] 1. The terminal displays the user interface via a web browser and presents the report.

[1997] 2. The user enters comments and suggestions using the feedback form, and the device collects the feedback.

[1998] Step 9: Analyze the feedback

[1999] The server uses an emotion engine to analyze the feedback to understand the user's emotional state and extract insights therefrom.

[2000] Input: User feedback data.

[2001] Output: Feedback analysis results.

[2002] Specific behavior:

[2003] 1. The server uses "Natural Language Processing (NLP)" technology to analyze the feedback text along with sentiment data.

[2004] 2. Extract analytical insights based on sentiment data.

[2005] Step 10: Dynamically Adjusting Reports

[2006] The device recognizes the user's emotions in real time and dynamically adjusts the report content according to the user's emotional state.

[2007] Input: Feedback analysis results, real-time user emotion data.

[2008] Output: Dynamically adjusted report.

[2009] Specific behavior:

[2010] 1. The device analyzes the user's facial expressions and voice in real time using OpenCV and Microsoft Azure Cognitive Services.

[2011] 2. Dynamically add detailed explanations and additional illustrations to your reports based on feedback analysis results.

[2012] (Application example 2)

[2013] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[2014] In the process of designing and evaluating new anticancer drugs, the key challenges are to efficiently and accurately collect chemical structure data and manufacturing information, preprocess the data, train generative models, generate new compounds, evaluate efficacy and side effects through simulations, select optimal candidates, and collect user feedback. In particular, it is important to analyze operator feedback based on their emotions and reflect it in improving the generative model.

[2015] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[2016] In this invention, the server includes: means for collecting information related to living organisms from a database; means for preprocessing the collected information related to living organisms; means for training a generative model based on the preprocessed information using a machine learning algorithm; means for generating new compounds using the trained generative model; means for evaluating the effects and side effects of the generated new compounds through simulation; means for selecting an optimal compound based on the evaluation results; means for generating a report including the selection results; means for providing the generated report to a user; means for a robot to collect chemical structure data and manufacturing information in real time and cleanse the data; means for a robot to generate new compounds using a machine learning model based on the preprocessed data; means for a robot to evaluate the generated compounds using molecular dynamics simulation; and means for analyzing user emotions and collecting feedback using an emotion engine. This enables the design and evaluation process of new anticancer drugs to be carried out efficiently and accurately, and makes it possible to reflect feedback based on the operator's emotions in improving the generative model.

[2017] A "database" is a system that stores information systematically and allows specific data to be searched and retrieved efficiently.

[2018] "Biologically relevant information" refers to data related to biological research and medical applications, such as chemical structure data, clinical trial data, and genetic information.

[2019] "Preprocessing" refers to the process of cleansing collected data, preparing it in the required format, and standardizing it so that it can be used by machine learning models.

[2020] A "machine learning algorithm" is a mathematical model that learns from data and makes predictions or classifications. Common examples include neural networks, decision trees, and support vector machines.

[2021] A "generative model" is a machine learning model that generates new data from given data. Examples include GANs (generative adversarial networks) and VAEs (variational autoencoders).

[2022] "Novel compounds" are chemical substances that have not previously existed and are generated using machine learning models.

[2023] "Simulation" refers to a computational method that mimics real-world phenomena in a virtual environment, such as molecular dynamics simulation.

[2024] "Evaluation" refers to predicting and analyzing the efficacy, side effects, stability, etc. of the compound produced through simulation.

[2025] A "report" is a written or electronic document that summarizes evaluation results and analysis data.

[2026] A "robot" is a mechanical device that performs physical tasks autonomously or semi-autonomously.

[2027] "Cleansing" refers to the process of processing missing values ​​in data and removing unnecessary data.

[2028] An "emotion engine" is a system that recognizes and analyzes a user's emotional state in real time. It includes facial expression analysis and voice analysis technologies.

[2029] "Feedback" refers to opinions and impressions provided by users, which are used to improve the system.

[2030] "Molecular dynamics simulation" is a simulation method that reproduces the movement and interactions of molecules using a computer.

[2031] In this invention, we develop a comprehensive system for the efficient design and evaluation of anticancer drugs. The system consists of the following main components:

[2032] 1. Data Collection and Preprocessing:

[2033] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from a database. It accesses the database via an API to obtain the necessary information. This data is temporarily stored in local storage.

[2034] The server cleanses the collected data, converts it into a consistent format (e.g., SMILES), and scales and standardizes the data to make it suitable for AI models.

[2035] 2. Training the generative AI model and generating compounds:

[2036] The server uses the preprocessed data to train machine learning algorithms (e.g., generative adversarial networks (GANs) or variational autoencoders (VAEs)). The trained models are then used to generate novel compounds.

[2037] The server uses the trained generative model to generate novel compounds, adjusting their structure based on constraints related to drug efficacy and molecular properties.

[2038] 3. Compound evaluation and selection:

[2039] The server runs molecular dynamics simulations to evaluate the efficacy, side effects, and stability of the compounds generated. These simulations require advanced computing resources and utilize parallel processing techniques.

[2040] The server analyzes the simulation results and selects the best candidates, taking into account criteria such as efficacy, minimal side effects, and molecular stability.

[2041] 4. Report Generation and User Interface:

[2042] The server creates a report based on the selection results, which includes detailed analysis results and visualized data for the selected compounds, allowing users to easily understand the analysis results.

[2043] The terminal displays the generated report to the user, who can use it to plan their next actions.

[2044] 5. Feedback collection and sentiment analysis:

[2045] The device uses a camera and microphone to recognize the user's emotions in real time when they enter feedback on the report content. The emotion engine analyzes their facial expressions and voice to determine their emotional state.

[2046] The server analyzes the feedback based on the emotion data obtained from the emotion engine and identifies areas for improvement in the generative model. This feedback improves the accuracy of the model.

[2047] Examples:

[2048] For example, a factory robot collects chemical structure data and manufacturing information in real time on the production line and cleanses the data. Based on the preprocessed data, the robot uses machine learning models to generate new compounds and evaluates them using molecular dynamics simulations. Operators provide feedback, and emotion analysis is performed via a camera and microphone. This feedback information is used to improve the system and reflect it in new compound designs.

[2049] Example prompt sentence:

[2050] data_sources = ["path / to / chemical_structure.csv", "path / to / clinical_trial.csv"]

[2051] model_path = "path / to / anticancer_drug_model.h5"

[2052] Creating a robot instance

[2053] robot = AntiCancerDrugDesignRobot(data_sources, model_path)

[2054] Data collection and preprocessing

[2055] preprocessed_data = robot.collect_and_preprocess_data()

[2056] Generation of new compounds

[2057] generated_compounds = robot.generate_new_compounds(preprocessed_data)

[2058] Running molecular simulations

[2059] simulation_results = robot.run_simulation(generated_compounds)

[2060] Emotional Feedback and Analysis

[2061] feedback_text = "These results are reassuring."

[2062] emotion_analysis = robot.analyze_emotions_and_feedback(feedback_text)

[2063] print("Emotion Analysis:", emotion_analysis)

[2064] This system not only streamlines the process of designing and evaluating anti-cancer drugs, but also enables more accurate results by incorporating emotion-based feedback.

[2065] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[2066] Step 1:

[2067] Data collection

[2068] The server collects biological information (chemical structure data, clinical trial data, genetic information, etc.) from multiple databases. It accesses the databases based on queries specified through the API to obtain the required data. This data is temporarily stored in local storage. The input is the raw data obtained from the API, and the output is the raw data stored in local storage.

[2069] Step 2:

[2070] Data Preprocessing

[2071] The server cleanses the collected data and corrects or removes incomplete data. In particular, it converts chemical structure information into a consistent format (e.g., SMILES format) and standardizes the data, making it easier for AI models to use. The input is the raw data stored in local storage, and the output is the cleansed and standardized data.

[2072] Step 3:

[2073] Training an AI model

[2074] The server uses the preprocessed data to train a machine learning algorithm (such as a GAN or VAE). The model learns patterns in the data and gains the ability to generate novel compounds. This training requires extensive computing resources. The input is the preprocessed data, and the output is a trained generative model.

[2075] Step 4:

[2076] Generation of n...

Claims

1. a means for collecting information relating to the living body from a database; means for pre-processing the collected biological information; means for training a generative model based on the preprocessed information using a machine learning algorithm; and means for generating novel compounds using the trained generative model; A means for evaluating the effects and side effects of the generated novel compound through simulation; A means for selecting an optimal compound based on the evaluation results; means for generating a report including the selection results; a means for providing the generated report to a user; A system including:

2. 10. The system of claim 1, further comprising means for inputting feedback from a user regarding the generated report.

3. 2. The system according to claim 1, further comprising means for performing a molecular dynamics simulation of the generated new compound.

Images