Method and system for selecting and predicting genes related to biological characteristics

Abstract

A method for selecting and predicting genes related to biological characteristics, according to an embodiment of the present invention, is a method by which a computing system including a memory and a processor determines, on the basis of bio-data, a plurality of genes related to a specific biological characteristic, comprising the steps of: acquiring a single-cell gene expression profile of a tissue sample for a higher-level biological characteristic including a target that is the specific biological characteristic; generating gene correlation information in which a plurality of genes co-expressing the acquired single-cell gene expression profile are clustered; determining gene importance for the target through a first machine learning model on the basis of the generated gene correlation information; and determining, as important genes, a plurality of genes for a task of the target on the basis of the determined gene importance of the target.

Description

Method and system for screening and predicting genes related to biological characteristics

[0001] The present invention relates to a method and system for selecting and predicting genes related to biological characteristics. More specifically, a high-dimensional weighted gene co-expression network analysis can be performed based on biodata, including single-cell gene expression profiles, to determine relevant genes using a machine learning model.

[0002] And the present invention relates to a method and system for selecting and predicting biological characteristics-related genes, which implements a method for predicting the results of a task by re-determining genes highly related to the target by learning a second machine learning model, which is a target prediction model, based on reconstructed gene expression information based on determined related genes.

[0003] Recently, research on bio-platform technologies, such as personalized medicine, RNA drug candidate discovery, and RNA-based gene therapy development, is actively underway using artificial intelligence.

[0004] Among these, personalized medicine is a medical approach that provides customized treatment to each patient based on the patient's unique genetic makeup, medical history, etc.

[0005] In the field of cancer treatment, personalized medicine analyzes a patient's tumor at the tumor microenvironment level and identifies cancer-related genes to provide effective treatment methods for each patient.

[0006] In other words, personalized medicine can be a means of providing effective treatment options to patients with rare cancers or those with difficult-to-treat cancers for whom treatment options are limited by traditional treatments.

[0007] However, conventional genetic screening and prediction methods utilize RNA sequencing data, which makes it difficult to identify tumor microenvironment and cell heterogeneity, making it difficult to identify biological mechanisms and functions.

[0008] In addition, single-cell RNA sequencing data used in conventional gene selection and prediction methods have limitations in that preprocessing is difficult because all genes with low relevance to the target to be analyzed are analyzed, and the prediction results of machine learning models are not very accurate.

[0009] Therefore, there is a need for the introduction of new technologies that can more accurately analyze the tumor microenvironment and cellular heterogeneity, efficiently determine only genes highly related to the target, and accurately predict tasks.

[0010] (Prior patent document 1) KR 2022-0074088 A

[0011] One embodiment of the present invention has been devised to solve the problems of the prior art as described above, and its purpose is to provide a method and system for selecting and predicting genes related to biological characteristics.

[0012] According to various embodiments of the present disclosure, a method and system for selecting and predicting genes related to biological characteristics are provided, which generate a gene module with a plurality of genes having similar biological mechanisms and functions based on bio data including single-cell gene expression profiles, determine genes with high gene importance using a first machine learning model, and predict a task for a target using a second machine learning model.

[0013] In particular, the present invention aims to provide a method for learning a machine learning model capable of determining genes highly related to biological characteristics by designing a machine learning model that is not limited by knowledge through existing research in biological characteristics for which prior research is insufficient, and a method for predicting a task using a machine learning model generated by learning the same.

[0014] However, the technical problems to be solved by the present invention and embodiments of the present invention are not limited to the technical problems described above, and other technical problems may exist.

[0015] A method for selecting and predicting genes related to a biological characteristic according to an embodiment of the present invention is a method for determining a plurality of genes related to a specific biological characteristic based on bio data by a computing system including a memory and a processor, the method comprising: obtaining a single-cell gene expression profile of a tissue sample for a higher-order biological characteristic including a target, which is the specific biological characteristic; generating gene correlation information by clustering a plurality of genes co-expressing the obtained single-cell gene expression profile; determining gene importance for a target through a first machine learning model according to the generated gene correlation information; and determining a plurality of genes for a task of the target as important genes according to the determined gene importance of the target.

[0016] Additionally, the step of obtaining the single-cell gene expression profile includes the step of obtaining the single-cell gene expression profile of the tissue sample for the first type of cancer that is the target and the tissue samples for each of the remaining multiple cancer types.

[0017] In addition, the step of generating the gene correlation information includes a step of performing network analysis on the single cell gene expression profile to calculate a correlation between genes as a weight, and a step of clustering a plurality of co-expressed genes according to the calculated weights between the genes.

[0018] Additionally, the step of generating the genetic correlation information includes the step of matching genes to each node and assigning weights between genes to edges between nodes.

[0019] In addition, the step of generating the gene correlation information includes a step of calculating a correlation coefficient for the correlation between genes based on the single cell gene expression profile, a step of generating a network with weights assigned based on the calculated correlation coefficient between genes, and a step of grouping nodes in the generated network based on a clustering algorithm to classify co-expressed genes into gene modules.

[0020] Additionally, the step of generating the gene correlation information includes a step of performing high-dimensional weighted gene co-expression network analysis (hdWGCNA) based on the single-cell gene expression profile.

[0021] In addition, the step of determining the gene importance for the target through the first machine learning model includes the step of generating first learning data by adding information about the tissue sample as a label based on the gene correlation information to the single cell gene expression profile.

[0022] In addition, the step of generating the first learning data includes the step of adding the type and characteristics of the tissue sample generating the single-cell gene expression profile as a label, and the step of reflecting the gene expression information of the single-cell gene expression profile into the network of the gene correlation information.

[0023] In addition, the step of generating the first learning data includes the step of adding the type and characteristics of the tissue sample generating the single-cell gene expression profile as a label, and the step of reflecting the gene expression information of the single-cell gene expression profile into the network of the gene correlation information.

[0024] In addition, the step of determining the gene importance through the first machine learning model includes the step of calculating the gene importance by converting the correlation to the target into a numerical value based on the frequency of the gene used to predict the target through the tree-based first machine learning model with the first learning data.

[0025] In addition, the step of determining the important gene includes at least one step of determining a gene whose gene importance is equal to or greater than a preset value as the important gene, and a step of determining a top preset number of n genes based on the gene importance as the important gene.

[0026] In addition, the method further includes a step of generating second learning data by reconstructing a single-cell gene expression profile for the above-mentioned important genes, and a step of learning a second machine learning model based on the second learning data to perform a task for the target.

[0027] Additionally, the step of reconstructing the single-cell gene expression profile to generate second learning data includes a step of extracting gene expression information for the important genes to generate second learning data.

[0028] In addition, the method further includes a step of performing a task for the target through the second machine learning model, and then determining some of the important genes as secondary important genes through weights given to the important genes by the second machine learning model when performing the task.

[0029] In addition, the method includes a step of extracting gene expression information for the determined second important gene to generate third learning data, and a step of re-performing the task by training the second machine learning model with the generated third learning data.

[0030] Meanwhile, a system for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention comprises: at least one memory; and at least one processor for performing selection and prediction of genes related to biological characteristics by reading out at least one application stored in the memory; wherein instructions of the processor include instructions for performing the steps of: obtaining a single-cell gene expression profile of a tissue sample for a higher biological characteristic including a target, which is the specific biological characteristic; generating gene correlation information by clustering a plurality of genes co-expressing the obtained single-cell gene expression profile; determining gene importance for the target through a first machine learning model according to the generated gene correlation information; and determining a plurality of genes for a task of the target as important genes according to the determined gene importance of the target.

[0031] In addition, the present invention further includes a Field Programmable Gate Array (FPGA) implementation for a predetermined artificial neural network, comprising: a plurality of neurons configured as an array including at least one register, at least one programmable logic, and at least one input interface; a plurality of synaptic circuits storing synaptic weights that adjust connection strengths between the plurality of neurons; and at least one routing network that controls data flow between the plurality of neurons; wherein each of the plurality of neurons is connected to at least one other neuron through the routing network to set a transmission path of the weights.

[0032] Also, a plurality of neurons organized in an array including at least one register, at least one microprocessor, and at least one input; and a plurality of synaptic circuits storing synaptic weights that adjust connection strengths between the plurality of neurons; wherein each of the plurality of neurons further includes an application specific integrated circuit (ASIC) for a predetermined artificial neural network connected to at least one other neuron through one of the plurality of synaptic circuits.

[0033] A method and system for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention effectively reflects complex cell interactions and cell heterogeneity within a tumor microenvironment by determining a plurality of important genes using the first and second machine learning models, thereby having the effect of helping in the study of cell-to-cell interactions and biological mechanisms.

[0034] In addition, the method and system for selecting and predicting biological characteristics-related genes according to an embodiment of the present invention precisely analyzes the heterogeneity of each single cell and cell-to-cell interactions within a tumor microenvironment based on the single-cell gene expression profile itself, thereby identifying gene expression variations and increasing the accuracy of predicting cancer progression and treatment responsiveness.

[0035] In addition, the method and system for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention have the effect of increasing the possibility of understanding the mechanism of a disease and / or discovering new treatment targets by analyzing a plurality of genes co-expressed within a tumor microenvironment through network analysis, thereby identifying a gene network interacting in a biological process.

[0036] In addition, the method and system for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention have the effect of performing high-dimensional weighted gene co-expression network analysis to identify correlations between multiple genes, thereby identifying multiple important genes and understanding the biological mechanisms and / or functions of the genes.

[0037] In addition, the method and system for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention can utilize lower-dimensional data than high-dimensional genetic data by determining genes using the first machine learning model, thereby reducing the cost and time required for the preprocessing process.

[0038] In addition, the method and system for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention have the effect of easily processing noise in high-dimensional genetic data by determining genes according to gene importance, thereby increasing the prediction accuracy of a machine learning model.

[0039] In addition, the method and system for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention have the effect of improving the prediction performance results of a machine learning model and identifying common important genes that can explain various biological phenomena.

[0040] In addition, the method and system for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention effectively reflects complex cell interactions and cell heterogeneity within a tumor microenvironment by analyzing single-cell gene expression patterns, thereby having the effect of understanding cell-to-cell interactions and biological mechanisms.

[0041] However, the effects that can be obtained from the present invention are not limited to the effects mentioned above, and other effects that are not mentioned can be clearly understood from the description below.

[0042] FIG. 1 illustrates an example of a block diagram of a computing system that implements a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0043] FIG. 2 illustrates an example of a block diagram of a computing device that implements a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0044] FIG. 3 illustrates an example block diagram of another aspect of a computing device implementing a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0045] Figure 4 illustrates a flowchart of a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0046] FIG. 5 illustrates a diagram of a method for generating a single-cell gene expression profile according to one embodiment of the present invention.

[0047] FIG. 6 illustrates a diagram projecting a plurality of genes and / or single cells into a low-dimensional space by clustering them according to one embodiment of the present invention.

[0048] Figure 7 illustrates a network analysis diagram of a single cell gene expression profile according to one embodiment of the present invention.

[0049] FIG. 8 illustrates a table of targets and tasks for a second machine learning model according to one embodiment of the present invention.

[0050] Figure 9 is a comparative performance analysis table showing that the isolated cell gene expression profile network derived through the present invention has superior AUROC performance in multiple pan-cancer downstream tasks.

[0051] Figure 10 is a diagram visualizing the feature importance of the top 10 important genes for each downstream task and the common gene overlap between tasks, showing that the system of the present invention consistently captures key biomarkers across tasks.

[0052] The present invention is capable of various modifications and embodiments. Therefore, specific embodiments are illustrated in the drawings and described in detail in the detailed description. The effects and features of the present invention, as well as the methods for achieving them, will become clear with reference to the embodiments described in detail below together with the drawings. However, the present invention is not limited to the embodiments disclosed below and can be implemented in various forms. In the following embodiments, the terms "first," "second," etc. are not used in a limiting sense but are used for the purpose of distinguishing one component from another. Furthermore, the singular expression includes the plural expression unless the context clearly indicates otherwise. Furthermore, terms such as "include" or "have" indicate the presence of a feature or component described in the specification, and do not preemptively exclude the possibility that one or more other features or components may be added. Furthermore, in the drawings, the sizes of components may be exaggerated or reduced for convenience of explanation. For example, the size and thickness of each component shown in the drawings are arbitrarily shown for convenience of explanation, and thus the present invention is not necessarily limited to what is shown.

[0053] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. When describing with reference to the drawings, identical or corresponding components are given the same reference numerals and redundant descriptions thereof will be omitted.

[0054] The present invention relates to a method for selecting genes related to biological characteristics and a system for predicting related targets.

[0055] In particular, the present invention relates to a method for selecting genes related to biological characteristics through a machine learning model and a system for predicting targets based on the determined genes.

[0056] Here, a biological characteristic refers to a characteristic associated with life activities within multiple cells and / or tissues. In an embodiment, a biological characteristic in the present invention may be a concept having both upper and lower characteristics.

[0057] For example, a superordinate characteristic may correspond to a superordinate condition that includes multiple subordinate conditions, and a subordinate characteristic may refer to subordinate conditions that may fall into the category of a common superordinate condition.

[0058] In an example, the primary disease may be cancer, and the secondary disease may refer to various types of cancer, such as lung cancer, stomach cancer, and colon cancer. As another example, if the primary disease is diabetes, the secondary disease may refer to specific types of diabetes (e.g., stroke, myocardial infarction, etc.).

[0059] In the following description, the biological characteristics to be predicted in the present invention will be described as cancer and specific cancers.

[0060] Therefore, since the target, which is the target of prediction, may mean the presence or absence of a specific type of cancer among the types of cancer, drug response, progression stage, possibility of treatment, etc., the present invention first determines key genes related to the specific cancer corresponding to the target, and then performs a task of predicting the occurrence, progression stage, possibility, status, degree of genetic mutation, number of mutations or / and aggressiveness, etc. of the specific cancer based on the determined key genes.

[0061] First, the present invention collects user bio-data related to genes as training data to determine genes related to detailed cancer, and analyzes the collected training data to determine highly relevant genes based on gene importance to the target.

[0062] In an embodiment, biodata may include single cell sequencing profiles optimized for machine learning analysis of biological characteristics for which prior research on the microenvironment is lacking, such as cancer.

[0063] Here, the single-cell gene expression profile includes gene expression information of genes extracted from each of multiple single cells separated from the tissue, and the gene expression information may include at least one of a gene name, a gene expression amount value, and a gene expression pattern.

[0064] More specifically, the biodata may be single-cell gene expression profiles for tumor biopsy samples taken from tissues of multiple subtypes of cancer.

[0065] That is, an embodiment of the present invention analyzes common characteristics between genes through analysis of tissues of various cancers having common characteristics, thereby learning about the tumor microenvironment in which cancer develops, and performs an analysis to determine genes highly related to a specific targeted cancer within the learned tumor microenvironment, thereby determining genes highly related to the targeted specific cancer with high accuracy.

[0066] And the present invention can provide a service that makes predictions about specific biological characteristics with high accuracy by performing predictions related to a target using genetic information reconstructed based on genes determined in this manner.

[0067] In an embodiment, the target may include a biological characteristic to be predicted, such as the presence or absence of a sub-disease, drug response, progression stage, likelihood, biological value, etc.

[0068] [Exemplary system providing a method for screening and predicting genes related to biological traits]

[0069] Hereinafter, the present invention will be described in detail with reference to the attached drawings an exemplary system for implementing a method for selecting and predicting genes related to biological characteristics, which generates a gene module with a plurality of genes having similar biological mechanisms and / or gene expression patterns based on bio data including single-cell gene expression profiles, determines the plurality of genes using a first machine learning model, and performs target prediction using a second machine learning model.

[0070] FIG. 1 illustrates an example of a block diagram of a computing system that implements a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0071] Referring to FIG. 1, a computing system (1000) implementing a method for selecting and predicting genes related to biological characteristics of the present invention includes a user computing device (110), a server computing system (130), and a training computing system (150), and each device and system are communicatively connected through a network (170).

[0072] The method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention may 1) be implemented and provided locally by a user computing device (110), 2) be implemented and provided in the form of a web service by a server computing system (130) communicating with the user computing device (110), or 3) be implemented and provided by the user computing device (110) and the server computing system (130) in connection with each other.

[0073] At this time, in the embodiment, the user computing device (110) and / or the server computing system (130) can train the machine learning model (120 and / or 140) through interaction with a training computing system (150) communicatively connected via a network (170). The training computing system (150) may be separate from the server computing system (130) or may be part of the server computing system (130).

[0074] And at this time, the artificial intelligence model can be 1) trained directly locally by the user computing device (110), 2) trained by the server computing system (130) and the user computing device (110) interacting with each other through a network (170), and 3) trained by a separate training computing system (150) using various training techniques and learning techniques. And the artificial intelligence model trained by the training computing system (150) can be implemented in a manner of being provided / updated by transmitting the model to the user computing device (110) and / or the server computing system (130) through a network (170).

[0075] In some embodiments, the training computing system (150) may be part of a server computing system (130) or part of a user computing device (110).

[0076] -User Computing Device (110)

[0077] The user computing device (110) may include any type of computing device, such as a smart phone, a mobile phone, a digital broadcasting device, a personal digital assistant (PDA), a portable multimedia player (PMP), a desktop, a wearable device, an embedded computing device, and / or a tablet PC.

[0078] Additionally, in the embodiment, the user computing device (110) may further include a predetermined server computing device that provides an environment for implementing a method for selecting and predicting genes related to biological characteristics.

[0079] Additionally, in the embodiment, the user computing device (110) may further include a predetermined equipment (e.g., Illumina TruSight Oncology 500, FoundationOne CDx, and / or Thermo Fisher Ion Torrent Genexus System, etc.) that provides an environment for implementing a method for screening and predicting genes related to biological characteristics.

[0080] Such a user computing device (110) includes at least one processor (111) and memory (112).

[0081] Here, the processor (111) of the user computing device (110) may be composed of at least one or a plurality of processors electrically connected among a central processing unit (CPU), a graphics processing unit (GPU), application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors and / or other electrical units for performing functions.

[0082] According to an embodiment, such a processor (111) may be configured based on a field programmable gate array (FPGA) implementation and / or an application specific integrated circuit (ASIC), which is a hardware technology for implementing a predetermined digital circuit.

[0083] Here, a field programmable gate array (FPGA) can mean a flexible digital circuit that can be programmed according to user needs.

[0084] In an embodiment, the field programmable gate array implementation may include a register that temporarily stores data and controls the flow and timing of signals to maintain intermediate results of operations or state information to support synchronized operation of the FPGA; programmable logic that programs operations within the FPGA to perform specific functions or operations as logic circuits that can be configured according to user needs; and an input interface that receives signals from an external device or sensor as a path for receiving data from outside the FPGA and transmits them to the internal circuit.

[0085] By combining the above components, field programmable gate array implementations can provide flexible and diverse forms of digital circuits.

[0086] Meanwhile, an application-specific integrated circuit (ASIC) can refer to a custom integrated circuit that is specifically designed to perform a specific purpose or function.

[0087] As an example, the application-specific integrated circuit may include a register, which is a small memory device that temporarily stores and manages data and stores intermediate results of calculations or status information to support rapid processing of the operation of the ASIC; a microprocessor, which is a central processing unit that performs control and operations within the ASIC and, if necessary, performs various operations or generates control signals to coordinate the operation of the entire system; and an input block, which is an interface that receives data from the outside and receives data to be processed by the ASIC and transmits it to the inside, and receives various input data through a connection with a sensor or an external device.

[0088] By combining the above components, application-specific integrated circuits can perform specific tasks in an optimized manner.

[0089] Returning again, the memory (112) of the user computing device (110) may include one or more non-transitory / transitory computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, and combinations thereof, and may include web storage of a server that performs a memory storage function on the Internet. This memory (112) may store data (113) and instructions (114) necessary for the at least one processor (111) to perform functional operations such as training an artificial intelligence model or implementing a method for selecting and predicting genes related to biological characteristics through an artificial intelligence model.

[0090] In one embodiment, the user computing device (110) can perform various deep learning for selecting and predicting biological characteristics-related genes in conjunction with a deep learning neural network.

[0091] Here, the deep learning neural network according to the embodiment may include a convolutional neural network (CNN), an R-CNN (Regions with CNN features), a Fast R-CNN, a Faster R-CNN, a Mask R-CNN, etc., and may include any deep learning neural network that includes an algorithm capable of performing the embodiment described below, and the embodiment of the present invention does not limit or restrict such deep learning neural network itself.

[0092] At this time, depending on the embodiment, the deep learning neural network may be installed directly in the server computing system (130) or may operate as a device separate from the server computing system (130) to perform deep learning for the biological characteristic-related gene selection and prediction method.

[0093] Additionally, in one embodiment, the user computing device (110) may store at least one machine learning model (120).

[0094] For example, the user computing device (110) may be a variety of machine learning models, such as a plurality of neural networks (e.g., a deep neural network) that perform a method of implementing a method of selecting and predicting genes related to biological characteristics based on structured / quantitative data, or other types of machine learning models including nonlinear models and / or linear models, and may be configured as a combination thereof.

[0095] For example, the machine learning model may store linear regression, decision tree, random forest, gradient boosting, pre-trained language model, or / and deep learning model. The neural network may include at least one of feed-forward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, or / and other types of neural networks.

[0096] In an embodiment, the user computing device (110) may include a first machine learning model (120) that is a tree-based model and a second machine learning model that is an artificial neural network model.

[0097] In one embodiment, a user computing device (110) may receive at least one machine learning model (120) from a server computing system (130) via a network (170), store the model in a memory (112), and then execute the stored machine learning model (120) by a processor (111) to perform a method for selecting and predicting genes related to biological characteristics.

[0098] In another embodiment, a user computing device (110) may perform an operation through a machine learning model (140) including at least one machine learning model (140) in conjunction with a server computing system (130) and communicate data related thereto to the outside, thereby providing a method for selecting and predicting genes related to biological characteristics to the user.

[0099] For example, a user computing device (110) can perform a method for selecting and predicting biological characteristics-related genes in a manner that a server computing system (130) provides output for a user's input using a machine learning model (140) via the web.

[0100] Additionally, the artificial intelligence model may be implemented in such a way that at least some of the machine learning models (120 and / or 140) are executed on a user computing device (110) and the rest are executed on a server computing system (130).

[0101] Additionally, the user computing device (110) may include at least one input component (121) that detects user input.

[0102] For example, the user input component (121) may include a touch sensor (e.g., a touch screen and / or touch pad, etc.) that detects the touch of a user's input medium (e.g., a finger or a stylus), an image sensor that detects the user's motion input, a microphone, button, mouse, and / or keyboard that detects the user's voice input, etc.

[0103] Here, the image sensor may include an image processing module. Specifically, the image sensor may process still images or moving images obtained by an image sensor device (e.g., CMOS or CCD).

[0104] Additionally, the image sensor can process still images or moving images acquired through the image sensor device using an image recognition process (e.g., OCR, etc.) and / or an image processing module to extract necessary information and transmit the extracted information to the processor.

[0105] Additionally, the input component (121) may receive input for an external controller (e.g., mouse, keyboard, etc.) based on the interface module, and in this case may include an external output device (e.g., speaker).

[0106] At this time, the interface module may be configured to include at least one of a wired / wireless headset port, an external charger port, a wired / wireless data port, a memory card port, a port for connecting a device equipped with an identification module, an audio I / O (Input / Output) port, a video I / O (Input / Output) port, an earphone port, a power amplifier, an RF circuit, a transceiver, and other communication circuits.

[0107] Additionally, the external output device may include a display system that outputs various information related to a method for selecting and predicting genes related to biological characteristics as graphic images.

[0108] Such a display system can be implemented by including at least one of a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), an organic light-emitting diode (OLED), a flexible display, a 3D display, and an e-ink display.

[0109] Meanwhile, the user computing device (110) including the above-described components may further perform at least some of the functional operations performed by the server computing system (130) described below.

[0110] -Server Computing System (130)

[0111] The server computing system (130) can perform a series of processes to provide a method for selecting and predicting genes related to biological characteristics.

[0112] In detail, in an embodiment, the server computing system (130) can provide the biological characteristic-related gene selection and prediction method by exchanging data necessary to enable the biological characteristic-related gene selection and prediction method process to be driven by an external device, such as a user computing device (110), with the external device.

[0113] In more detail, in an embodiment, a server computing system (130) may provide an environment in which an application can operate on a user computing device (110).

[0114] To this end, the server computing system (130) may include application programs, data and / or commands for the application to operate, and may transmit and receive various data based thereon with the external device.

[0115] Additionally, the server computing system (130) includes at least one processor (131) and memory (132).

[0116] Here, the processor (131) may be composed of at least one or a plurality of processors electrically connected among a central processing unit (CPU), a graphics processing unit (GPU), application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors and / or other electrical units for performing functions.

[0117] According to an embodiment, such a processor (131) may be configured based on a Field Programmable Gate Array (FPGA) implementation, which is a hardware technology for implementing a predetermined digital circuit, and / or an Application Specific Integrated Circuit (ASIC). A detailed description thereof is omitted in accordance with the description of the FPGA and ASIC described above.

[0118] And the memory (132) may include one or more non-transitory / transitory computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. This memory (132) may store data (133) and instructions (134) necessary for the processor (131) to perform functional operations such as training an artificial intelligence model or executing selection and prediction of genes related to biological characteristics through an artificial intelligence model.

[0119] In one embodiment, the server computing system (130) may be implemented to include at least one computing device. For example, the server computing system (130) may be implemented to operate multiple computing devices according to a sequential computing architecture, a parallel computing architecture, or a combination thereof. Additionally, the server computing system (130) may include multiple computing devices connected via a network (170).

[0120] Additionally, the server computing system (130) may store at least one machine learning model (140). For example, the server computing system (130) may include a neural network and / or other multi-layer nonlinear models as the machine learning model (140). Exemplary neural networks may include feedforward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks. In an embodiment, the server computing system (130) may include a first machine learning model, which is a tree-based model, and a second machine learning model, which is an artificial neural network model, as the machine learning model (140).

[0121] In an embodiment, the server computing system (130) may further include a data store computing system (hereinafter, “data store”), which is a storage for continuously storing and managing raw data that serves as the basis for a method for selecting and predicting genes related to biological characteristics.

[0122] These data stores can include various forms of data storage, ranging from file systems to cloud storage. For example, a data store can include at least one of the following: a relational database that uses a structured query language (SQL) to define and manipulate data; a NoSQL database designed for flexibility and scalability to handle unstructured and semi-structured data; a data warehouse, which is a system used for reporting and data analysis, centralizes large amounts of data from multiple sources and is optimized for querying and analysis; a data warehouse that stores large amounts of raw data in its native formats of structured, semi-structured, and unstructured data; and a database on a local storage device or Network Attached Storage (NAS) that stores data in files in a format typically accessible by a computer operating system.

[0123] -Training Computing System (150)

[0124] The training computing system (150) includes at least one processor (151) and memory (152).

[0125] Here, the processor (151) may be composed of at least one or a plurality of processors electrically connected among a central processing unit (CPU), a graphics processing unit (GPU), application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors and / or other electrical units for performing functions.

[0126] According to an embodiment, such a processor (151) may be configured based on a Field Programmable Gate Array (FPGA) implementation, which is a hardware technology for implementing a predetermined digital circuit, and / or an Application Specific Integrated Circuit (ASIC). A detailed description thereof is omitted in accordance with the description of the FPGA and ASIC described above.

[0127] And the memory (152) may include one or more non-transitory / transitory computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. This memory (152) may store data (153) and instructions (154) necessary for the processor (151) to perform learning of an artificial intelligence model, etc.

[0128] For example, the training computing system (150) may include a model trainer (160) that trains a machine learning model (120 and / or 140) stored in the user computing device (110) and / or server computing system (130) using various training or learning techniques, such as backpropagation of errors (according to the framework illustrated in FIG. 3).

[0129] For example, such a model trainer (160) may perform updates for one or more parameters of a machine learning model (120 and / or 140) for a biological trait-related gene selection and prediction method in a backpropagation manner based on a defined loss function.

[0130] In some implementations, performing backward propagation of errors may include performing truncated backpropagation through time. The model trainer (160) may perform a number of generalization techniques (e.g., weight reduction, dropout, and / or knowledge distillation) to improve the generalization ability of the trained machine learning model (120 and / or 140).

[0131] Additionally, the model trainer (160) may train the machine learning model (120 and / or 140) based on a series of training data (161). Here, the training data (161) may include data in different formats, such as, for example, images, audio samples, and / or text. Examples of image types that may be used may include video frames, LiDAR point clouds, X-ray images, computed tomography scans, hyperspectral images, and / or various other forms of images.

[0132] In the present invention, training data (161) may include biodata in which genetic information is stored. In particular, training data (161) may include gene expression values ​​and may be data including single-cell gene expression information.

[0133] In an embodiment, the training data (161) may be data on single-cell gene expression profiles for 181 tumor biopsy samples collected from multiple cancer tissues, for example, 13 different cancer types.

[0134] In an embodiment, biodata may include a single cell sequencing profile optimized for machine learning analysis of biological characteristics, such as cancer, for which prior research on the tumor microenvironment is lacking. This training data (161) may be provided by a user computing device (110) and / or a server computing system (130). When a training computing device trains a machine learning model (120 and / or 140) on specific data of the user computing device (110), the machine learning model (120 and / or 140) may be characterized as a personalized model.

[0135] And the model trainer (160) includes computer logic utilized to provide the desired function.

[0136] Additionally, the model trainer (160) may be implemented as hardware, firmware, and / or software that controls a general-purpose processor. In one implementation, the model trainer (160) includes a program file stored in a storage device, which may be loaded into a memory (152) and executed by one or more processors (151). In another implementation, the model trainer (160) includes one or more sets of computer-executable data (153) and instructions (154) stored in a tangible computer-readable storage medium, such as a RAM hard disk or an optical or magnetic medium.

[0137] The network (170) includes, but is not limited to, a 3rd Generation Partnership Project (3GPP) network, a Long Term Evolution (LTE) network, a World Interoperability for Microwave Access (WIMAX) network, the Internet, a Local Area Network (LAN), a Wireless Local Area Network (Wireless LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), a Bluetooth network, a satellite broadcasting network, an analog broadcasting network, and / or a Digital Multimedia Broadcasting (DMB) network.

[0138] In general, communication over a network (170) may be performed using any type of wired and / or wireless connection, using various communication protocols (e.g., TCP / IP, HTTP, SMTP, and / or FTP), encodings or formats (e.g., HTML and / or XML), and / or protection schemes (e.g., VPN, Secure HTTP, and / or SSL).

[0139] FIG. 2 illustrates an example of a block diagram of a computing device that implements a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0140] As shown in FIG. 2, the computing device (100) included in the user computing device (110), the server computing system (130), and the training computing system (150) includes a plurality of applications. Each application may include a machine learning library and one or more machine learning models. For example, the applications may include an image processing (e.g., detection, classification, and / or segmentation) application, a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, and / or a chat-bot application.

[0141] In particular, the application may include an application that determines a gene associated with a biological characteristic to be predicted based on input bio-data and / or an application that predicts a biological characteristic based on input bio-data.

[0142] In an embodiment, a computing device (100) may include a model trainer (160) for training an artificial intelligence model, and may store and operate the trained artificial intelligence model to provide output data according to predetermined input data (e.g., a single cell gene expression profile, etc.).

[0143] Each application of the computing device (100) may communicate with a number of other components of the computing device (100), such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In one embodiment, each application may communicate with each device component using an API (e.g., a public API). In one embodiment, the API used by each application may be specific to that application.

[0144] FIG. 3 illustrates an example block diagram of another aspect of a computing device implementing a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0145] Referring to FIG. 3, a computing device (200) includes a plurality of applications. Each application may communicate with a central intelligence layer. For example, the applications may include an image processing application, a text messaging application, an email application, a dictation application, a virtual keyboard application, and / or a browser application. In one embodiment, each application may communicate with the central intelligence layer (and the models stored therein) using an API (e.g., a common API across all applications).

[0146] The central intelligence layer may include multiple machine learning models. For example, as illustrated in FIG. 3, at least some of the machine learning models may be provided to each application and managed by the central intelligence layer. In other implementations, two or more applications may share a single machine learning model. For example, in some implementations, the central intelligence layer may provide a single model to all applications. In some implementations, the central intelligence layer may be included within the operating system of the computing device (200) or implemented differently.

[0147] The central intelligence layer may communicate with a central device data layer. The central device data layer may be a centralized data repository for the computing device (200). As illustrated in FIG. 3, the central device data layer may communicate with a number of other components of the computing device (200), such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, the central device data layer may communicate with each device component using an API (e.g., a private API).

[0148] The techniques described herein may refer to servers, databases, software applications, and other computer-based systems, as well as actions taken and information transmitted to or from such systems. It will be appreciated that the inherent flexibility of computer-based systems allows for a wide range of possible configurations, combinations, and division of labor and functionality between and among components. For example, the processes described herein may be implemented using a single device or component, or multiple devices or components operating in combination. Databases and applications may be implemented on a single system or in a distributed system across multiple systems. Distributed components may operate sequentially or in parallel.

[0149] [Method for screening and predicting genes related to biological characteristics]

[0150] Hereinafter, a process of determining a gene highly related to a target based on a collected single-cell gene expression profile by a computing system (1000) according to one embodiment of the present invention will be described in detail.

[0151] Specifically, a method for selecting and predicting genes related to biological characteristics performed by a computing system (1000) according to one embodiment of the present invention can perform high-dimensional weighted gene co-expression network analysis based on a single-cell gene expression profile and determine genes using a first machine learning model.

[0152] In addition, a method for predicting genes related to biological characteristics performed by a computing system (1000) according to one embodiment of the present invention can perform a task of accurately predicting a target biological characteristic by learning a second machine learning model based on determined genes, and can re-perform the task by re-determining important genes based on the results of the second machine learning model.

[0153] Hereinafter, a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention will be described in more detail with reference to the attached drawings.

[0154] FIG. 4 illustrates a flowchart for explaining a method for providing a method for selecting and predicting genes related to biological characteristics according to one embodiment of the present invention.

[0155] Referring to FIG. 4, a method for selecting and predicting genes related to biological characteristics according to an embodiment of the present invention may include a step (S101) of obtaining a single-cell gene expression profile including a plurality of single-cell gene expression patterns of cancer tissue, a step (S103) of generating gene correlation information by clustering a plurality of genes related by co-expression characteristics using a network analysis of the obtained single-cell gene expression profile, a step (S105) of determining gene importance for each gene for the first cancer through a first machine learning model when targeting the first cancer, a step (S107) of generating second learning data for a task of the target based on the important genes determined for the target, a step (S109) of performing the task of the target through the second learning data including information on the genes determined for the target, and a step (S111) of re-determining at least some of the genes determined according to the performed task and re-performing a task of predicting the target based on information on the re-determined genes.

[0156] In this case, determining a gene may mean determining a single gene and / or a group of genes that are grouped together with similar expression patterns.

[0157] Specifically, a computing system (1000) according to one embodiment of the present invention can obtain a single-cell gene expression profile including gene expression patterns of multiple single cells of various types of cancer tissues. (S101)

[0158] Specifically, the computing system (1000) can first perform an analysis of the tumor microenvironment through single-cell gene expression profiles for tissues of various types of cancer, as well as for specific cancers, the target cancer to be predicted. In other words, by performing an analysis of the tumor microenvironment for multiple cancers before genetic determination for a specific cancer, the accuracy of genetic determination for a specific cancer can be improved.

[0159] In the example, if the data used in the machine learning model is data based on tissue samples of 181 types of cancer, the target may mean the presence or absence of colon cancer, the stage of colon cancer progression, and / or a numerical value related to colon cancer.

[0160] FIG. 5 is a diagram of a method for generating a single cell gene expression profile according to one embodiment of the present invention.

[0161] Referring to FIG. 5, the computing system (1000) can obtain a single cell gene expression profile including a plurality of single cell gene expression patterns.

[0162] In an embodiment, a single-cell gene expression profile may include gene expression information such as gene expression amount values ​​and gene expression patterns at the single-cell level.

[0163] Specifically, a single-cell gene expression profile may include information about gene expression patterns representing expression levels for each gene for each single cell included in multiple cancer tissues.

[0164] Additionally, single-cell gene expression profiles may further include information on gene expression patterns (bulk RNA-seq), which represents gene-by-gene expression levels in a tissue sample unit of a specific cancer.

[0165] In a detailed embodiment, the computing system (1000) can classify various types of cancer tissue collected using bioanalysis equipment and a nucleic acid extraction system into single cells through cell sorting, and perform barcoding to distinguish each cell. When the target is a specific cancer, this is to first analyze the tumor microenvironment in which cancer develops through sample tissues of various types of cancer, including the specific cancer.

[0166] Here, barcoding refers to the act of classifying cells using genetic markers to identify them. In one embodiment, the computing system (1000) can retrieve barcoded data from an external genetic database and perform analysis on the tumor microenvironment.

[0167] The computing system (1000) can identify the results of cell sorting through barcoding and perform sequencing based on these results.

[0168] Here, sequencing refers to the process of analyzing the genetic information of an organism.

[0169] In an embodiment, sequencing may be a process of analyzing the base sequence of a gene to convert gene expression values ​​into data and / or generate a single-cell gene expression profile.

[0170] In other words, the computing system (1000) can generate and / or obtain a single-cell gene expression profile by converting genetic information into data through barcoding and sequencing processes or by loading data from an external genetic database.

[0171] The single-cell gene expression profile obtained in the example may have differences in the activated gene expression level values ​​depending on the type of cell, developmental stage, etc. within the tissue connected to the target.

[0172] FIG. 6 is a diagram of clustering genes and / or single cells according to one embodiment of the present invention and projecting them onto a low-dimensional space.

[0173] Referring to FIG. 6, a computing system (1000) that has obtained a direct or analyzed single-cell gene expression profile in this manner can cluster single cells and / or genes based on the generated single-cell gene expression profile and project them into a low-dimensional space to generate and / or output visualization data.

[0174] Here, clustering refers to a technique for grouping single cells and / or genes by characteristics such as similar functions, expression patterns, and pathways.

[0175] In an embodiment, the computer system (1000) can cluster cells and / or genes according to similar biological mechanisms and / or functions using the K-means or Louvain algorithm.

[0176] In an embodiment, the computing system (1000) can cluster single cells based on the function, pathway, and biological characteristic values ​​of each cell, and group them into groups of cells with high similarity, such as immune cells and cancer cells.

[0177] Additionally, in an embodiment, the computing system (1000) can cluster genes that have similar biological mechanisms or perform similar functions.

[0178] In another embodiment, the computing system (1000) acquires a scRNA-seq dataset of 181 tumor biopsy samples from 13 different cancer types, and clusters the scRNA-seq data of tissue samples by cancer type based on biological characteristics such as gene expression levels, expression patterns, and functions.

[0179] At this time, the computing system (1000) can visualize and confirm single cells and / or genes grouped in the same group by projecting them into a low-dimensional space using a dimensionality reduction technique such as t-SNE or UMAP. However, since existing clustering techniques lack prior research and may be difficult to apply to complex tumor microenvironments, in the embodiment, the computing system (1000) can apply a new clustering technique without using an existing clustering base.

[0180] Through this, the heterogeneity of each single cell and cell-to-cell interactions within the tumor microenvironment can be precisely analyzed based on the single-cell gene expression profile itself, thereby identifying gene expression mutations and increasing the accuracy of predicting cancer progression and treatment responsiveness.

[0181] Specifically, when a single-cell gene expression profile is obtained, the computing system (1000) can generate gene correlation information by clustering multiple genes related to co-expression characteristics using network analysis of the obtained single-cell gene expression profile. (S103)

[0182] Here, network analysis may be an analytical technique that analyzes relationships by forming gene modules that group genes with identical or similar co-expression and / or biological functions based on single-cell gene expression profiles for various cancer tissue samples.

[0183] In detail, the computing system (1000) can perform network analysis of a single cell gene expression profile to detect co-expressed genes that perform biological functions such as differentiating into cells with specific functions or interacting in the same pathway, and group the detected genes to determine them as gene modules.

[0184] That is, a genetic module may mean a group in which multiple cells and / or genes are grouped by similarities such as similar biological mechanisms and / or functions.

[0185] At this time, the gene module may include gene module location information. Specifically, the gene module location information is information indicating which gene module a single cell and / or gene belongs to.

[0186] For example, in the case of an oncogenic gene, the module location information of the oncogenic gene can be used to indicate the function of the gene and provide location information as a cancer-related gene, and location information can be provided by indicating a single cell and / or gene with a number or letter.

[0187] In an embodiment, the computing system (1000) can further perform network analysis by utilizing genetic module location information.

[0188]

[0189] *Figure 7 is a diagram illustrating a network analysis of a single cell gene expression profile according to one embodiment of the present invention.

[0190] Referring to Fig. 7, in the embodiment, genetic correlation information generated through network analysis may be composed of nodes, which are objects, and edges, which represent relationships between objects.

[0191] In detail, a node represents each object in network analysis, and one gene can be represented by one node.

[0192] Additionally, an edge refers to a connection line between nodes and can express the relationship between nodes.

[0193] In an embodiment, the computing system (1000) can represent the relevance and expression level of two genes by the thickness and length of an edge, and multiple edges can be connected to one node.

[0194] In a detailed embodiment, the computing system (1000) can perform network analysis by reflecting weights according to the degree of connection centrality index.

[0195] Specifically, in an embodiment, the computing system (1000) may reflect the weight as the sum of nodes directly connected to one node based on the number of other nodes connected to each node in the network.

[0196] In an embodiment, the computing system (1000) may increase the centrality as the connection frequency increases, thereby increasing the assigned weight value, thereby constructing a network representing genetic relationships.

[0197] That is, in the embodiment, the computing system (1000) can represent the correlation between nodes by the thickness and length of the edge, and can identify a hub gene within a genetic network formed through the correlation.

[0198] In an embodiment, the computing system (1000) can determine that the biological mechanisms or roles between the genes being compared are similar when the weight is large, the thickness of the edge is thick, the length of the edge is short, and the degree of co-expression is strong in the results of the configured network analysis.

[0199] Additionally, in the embodiment, the computing system (1000) may have a lower correlation between genes and a lower degree of co-expression as the weights in the configured network analysis results are smaller.

[0200] Meanwhile, co-expression refers to the correlation of multiple genes in their expression patterns under the same environmental conditions. In some embodiments, co-expression may include instances where gene expression patterns are similar depending on biological pathways and / or functions.

[0201] Additionally, in embodiments, co-expression may include cases where two genes have similar gene expression patterns even if they do not have similar biological functions and / or pathways, and where one gene is expressed when the other gene is expressed.

[0202] In an embodiment, the computing system (1000) can perform network analysis of single-cell gene expression profiles for various cancers to generate gene correlation information that clusters co-expression for various biological functions, pathways, unknown cell types, etc. in the tumor microenvironment.

[0203] Specifically, the computing system (1000) can calculate a correlation coefficient for inter-gene correlation using the gene expression profile expressed in each single cell.

[0204] Specifically, the computing system (1000) can calculate a correlation coefficient based on the similarity of expression patterns between genes using a correlation coefficient (typically the Pearson correlation coefficient). A higher correlation coefficient indicates a higher likelihood that two genes play a role in the same biological pathway or function, and thus can serve as a criterion for determining co-expressed genes.

[0205] The computing system (1000) can then generate a weighted network based on the correlation coefficients between each gene pair. Here, the weights reflect the similarity of expression patterns between genes, and a higher weight may indicate a closer relationship between genes.

[0206] Next, the computing system (1000) can use a clustering algorithm to group genes with similar expression patterns into a single module. Each gene module may be a set of genes belonging to similar functions or pathways within the tumor microenvironment, and co-expressed genes may be clustered.

[0207] Additionally, the computing system (1000) can detect genes with high network centrality within each gene module as hub genes.

[0208] Here, hub genes are likely to play important roles in biological pathways and may be key genes that act specifically in disease states.

[0209] For example, the computing system (1000) can use the above network analysis process as a high-dimensional weighted gene co-expression network analysis, and can generate gene correlation information through the high-dimensional weighted gene co-expression network analysis.

[0210] Here, high-dimensional weighted gene co-expression network analysis (hdWGCNA) refers to a network analysis that weights genes with high co-expression and / or correlation using single-cell gene expression profiles, which are high-dimensional data.

[0211] Therefore, gene correlation information is information indicating correlations between genes based on single-cell gene expression profiles, and may be gene co-expression information including the results of high-dimensional weighted gene co-expression network analysis.

[0212] Specifically, the computing system (1000) can display correlations by calculating correlation coefficients by including more genetic module information.

[0213] In this embodiment, the correlation coefficient can be calculated using parametric and non-parametric methods, such as the Pearson correlation coefficient, Spearman correlation coefficient, and / or Kendall's rank correlation coefficient.

[0214] In an embodiment, the computing system (1000) may include node centrality and edge weight information in the genetic correlation information.

[0215] Here, the centrality of a node is an indicator of the importance of each gene in the genetic module within the network, and the centrality value can be set higher as it corresponds to an important gene.

[0216] In detail, the centrality of a node can include information such as the number of connected edges (Degree Centrality) and the proportion of paths between other nodes in the network (Betweenness Centrality).

[0217] In detail, a key gene may be a gene with a high correlation with the target in a biological function and / or pathway, and the key gene may include a hub gene.

[0218] Specifically, important genes may have high explanatory power for biological pathways and / or functions of gene modules.

[0219] Network analysis according to an embodiment of the present invention can identify a gene network that interacts in a biological process by analyzing multiple genes co-expressed within a tumor microenvironment, thereby increasing the possibility of understanding the mechanism of a disease and / or discovering new therapeutic targets.

[0220] In addition, the computing system (1000) according to an embodiment of the present invention performs high-dimensional weighted gene co-expression network analysis to identify correlations between multiple genes, thereby identifying multiple important genes and having the effect of identifying the biological mechanisms and / or functions of the genes.

[0221] When generating genetic correlation information, the computing system (1000) can determine the gene importance for each gene for the first cancer using the first machine learning model when targeting the first cancer. (S105)

[0222] A target can refer to a specific, detailed target among the biological characteristics to be predicted. In the example, the target refers to the characteristics of a specific cancer type among the cancer types used as samples. The first cancer will be described as the target below.

[0223] At this time, the characteristics of cancer may include the occurrence of cancer, grade, progression stage, tumor mutation burden (TMB), microsatellite instability (MSI), mutation classification (MUT), cancer grading (GRAD), cancer subtyping (SUB), etc., and therefore, the task may be a task of predicting these characteristics of cancer.

[0224] Additionally, the computing system (1000) may generate first learning data by preprocessing a single-cell gene expression profile based on gene correlation information to determine the gene importance of a target. Furthermore, the first learning data herein may refer to input data in which a single-cell gene expression profile is labeled with information such as cancer type and cancer characteristics, and the single-cell gene expression profile is reconstructed by reflecting the gene correlation information.

[0225] In detail, the computing system (1000) can generate first learning data by adding characteristics of a tissue sample from which a single cell gene expression profile is generated as a label.

[0226] Specifically, if a tissue sample has a first-level cancer and a second-level cancer grade, the first-level cancer and the second-level cancer grade can be added as labels to multiple single-cell gene expression profiles generated through the tissue sample.

[0227] And the computing system (1000) can generate first learning data to represent the expression pattern of genes included in the gene module for each cancer type.

[0228] Specifically, the computing system (1000) can generate first learning data by adding information on the expression levels of genes included in each gene module in each sample (single cell expression information for tissue samples by cancer type).

[0229] In an embodiment, the computing system (1000) may use scRNA-seq data and / or TCGA data for 181 tumor biopsy samples of 13 cancer types with labels as first training data.

[0230] Additionally, in the embodiment, the computing system (1000) may reconstruct data by performing data preprocessing processes such as normalization, filtering, scaling, and / or data augmentation based on the first learning data.

[0231] Returning back, the computing system (1000) can train the first machine learning model with the generated first learning data to calculate the feature importance of each gene for the first cancer as a target.

[0232]

[0233] Here, gene importance can mean a value indicating a high correlation between the expression of the first cancer and the characteristics of the first cancer.

[0234] In an embodiment, the first machine learning model that the computing system (1000) uses to calculate gene importance may be an XGBoost model to which a tree-based boosting technique is applied.

[0235] In an embodiment, the computing system (1000) can train a first machine learning model with first learning data, and output a result using the learning data labeled with the first cancer as the target and the learning data labeled with respect to the remaining cancers as input data.

[0236] At this time, the computing system (1000) can calculate the gene importance for the first cancer target based on criteria such as the amount of data processed by the first machine learning model and the frequency used for classification. In addition, in the embodiment, the computing system (1000) can normalize the gene importance values ​​based on the calculated gene importance values ​​and convert them into a range between 0 and 1 to generate a list of important genes.

[0237] Here, the list of important genes refers to multiple genes determined based on the gene importance derived from the first machine learning model, and multiple genes may be included in the list.

[0238] The computing system (1000) can determine genes based on the results of the generated important gene list, based on a gene importance greater than a specific numerical value and / or a preset number of genes.

[0239] In an embodiment, the computing system (1000) may generate a list of important genes by setting a feature importance of the XGBoost model, which is a gene importance of 0.0001 or higher, as a criterion, and determining genes with an importance value of 0.0001 or higher as important genes. Here, the feature may mean each gene and / or gene module containing gene expression information.

[0240] In another embodiment, the computing system (1000) may set n, which is a number suitable for input into a second machine learning model set according to a target, as a criterion for determining important genes, and may determine the top n genes with high gene importance values ​​from a list listed according to gene importance.

[0241] For example, the computing system (1000) can determine genes corresponding to 10% of the list of important genes of the target and / or the purpose of the target.

[0242] In this embodiment, the computing system (1000) can select multiple genes with the highest explanatory power by splitting the nodes of the tree based on gene expression levels. Subsequently, the computing system (1000) can calculate gene importance based on the frequency used in node splitting and the contribution to prediction performance for multiple genes with high explanatory power. Finally, the computing system (1000) can derive a gene importance score using one or more measurements such as gain, frequency, and cover.

[0243] Here, gain refers to an indicator that measures how much a specific gene improves the prediction performance when splitting a node.

[0244] Additionally, frequency can be measured as the number of times a gene is used to split a node in a tree, and the higher the frequency, the higher the importance.

[0245] Cover also refers to a metric that measures how many data points a gene covers when splitting in the tree.

[0246] The computing system (1000) uses the first machine learning model to determine genes, thereby utilizing lower-dimensional data rather than high-dimensional genetic data, thereby reducing the cost and time required for the preprocessing process.

[0247] Once the target's gene importance is determined through the first machine learning model, the computing system (1000) can generate second learning data for the target's task based on the determined important genes for the target. (S107)

[0248] The computing system (1000) can generate second learning data for learning a second machine learning model that performs a task based on important genes for a target determined through the first machine learning model.

[0249] Additionally, the computing system (1000) can preprocess and / or convert genetic information data of the determined gene before transmitting it to the second machine learning model.

[0250] In an embodiment, if genes A, B, and C are determined as important genes based on the gene importance of the target, the computing system (1000) may determine at least some of the important genes as genes for the task, preprocess and / or convert them, and utilize them in the second machine learning model.

[0251] For example, if the target is colon cancer, the important genes may be genes related to tumor suppressor genes, cancer cell proliferation-related genes, or immune-related genes, respectively. Here, if the target task is to determine the progression stage of colon cancer, the computing system (1000) may determine a tumor suppressor gene among the important genes as the important gene for the target task.

[0252] Additionally, the computing system (1000) may determine to include all of the determined plurality of important genes in the second learning data, and may also vary the weights of each important gene.

[0253] Through this step, the computing system (1000) has the effect of easily processing noise in high-dimensional genetic data by determining genes based on gene importance, thereby increasing the prediction accuracy of the machine learning model.

[0254] After determining the genes to be included in the second learning data for the target task, the computing system (1000) can perform the target task using the second learning data including information about the genes determined for the target. (S109)

[0255] At this time, the computing system (1000) can generate data including a single-cell gene expression profile for a target among multiple targets, a list of important genes determined by a first machine learning model, and / or a label according to a task for the target as learning data (hereinafter, second learning data).

[0256] In an embodiment, the computing system (1000) can train a second machine learning model using the generated second learning data.

[0257] At this time, in the embodiment, the computing system (1000) can perform task prediction using data including graph data consisting of nodes and edges using a second machine learning model including a multi-layer perceptron model (hereinafter referred to as MLP) and a graph neural network (hereinafter referred to as GNN).

[0258] Additionally, in the embodiment, when the computing system (1000) performs a prediction on a task using a second machine learning model, it can assign weights to nodes that are important genes and edges that are relationships between genes.

[0259] Additionally, in the embodiment, the computing system (1000) may further include a step of determining a second important gene based on the result of applying weights to each node and each edge in the second machine learning model.

[0260] The computing system (1000) can train the second learning data on the second machine learning model and perform a task to derive the first result.

[0261] FIG. 8 is a table of targets and target tasks for a second machine learning model according to one embodiment of the present invention.

[0262] Referring to FIG. 8, in an embodiment, the computing system (1000) can perform TMB, which is a task of the first target, based on second learning data.

[0263] Here, TMB refers to the number of mutations in tumor cells, and can be an indicator of the number of mutations that have occurred in cancer cells.

[0264] In detail, the higher the TMB value, the more mutations there are in the tumor, making it easier for the immune system to recognize it, and TMB can be used to predict immune response or treatment potential.

[0265] In more detail, the computing system (1000) can predict the TMB of colon cancer by building a TMB prediction model by training a second machine learning model with second learning data consisting of data on important genes of the target as labels to predict the TMB value of colon cancer.

[0266] Accordingly, the computing system (1000) can predict the TMB value using a second machine learning model learned only based on the expression patterns of determined important genes, and can derive the result that a patient with a high TMB value is likely to respond to immunotherapy without data on unnecessary genes.

[0267] Additionally, in the embodiment, the computing system (1000) can perform MSI, which is a task of the first target, based on the second learning data.

[0268] Here, MSI stands for microsatellite instability, which refers to a mutation that occurs in microsatellites and causes a change in the number of repeats, and can be an indicator for evaluating the degree of response to immunotherapy.

[0269] At this time, microsatellites are regions in which the same bases are repeated in the gene, and if there is microsatellite instability, cancer cell gene mutations may increase, which may cause a strong immune response.

[0270] In detail, the computing system (1000) learns second learning data with MSI values ​​set as labels to build an MSI classification model, and receives important gene expression information as input data to predict the possibility of gastric cancer occurrence according to MSI.

[0271] In more detail, the computing system (1000) trains a second machine learning model with second learning data including data on important genes of a target whose MSI value is a label, and determines whether gastric cancer has developed by determining whether it is MSI-H, MSI-L, or MSS based on the MSI value through the trained second machine learning model.

[0272] At this time, the computing system (1000) learns a second machine learning model through second learning data on the expression patterns of the determined important genes to predict the MSI value, thereby classifying a patient with a high MSI value as MSI-H, and thus can derive a result indicating that there is a high possibility of a response to immunotherapy.

[0273] Additionally, in the embodiment, the computing system (1000) can perform cancer subtyping, which is a target task, based on the second learning data.

[0274] Here, cancer subtyping (SUB) can be the process of classifying cancer into subtypes based on genetic mutations, expression patterns, biological characteristics, etc.

[0275] In particular, since cancer is divided into several subtypes, more precise diagnosis and treatment strategies can be applied to each patient by considering the genetic and biological characteristics of the cancer.

[0276] In detail, the computing system (1000) learns second learning data in which the gynecological cancer subtype classification value is set as a label, builds a gynecological cancer subtype classification model, and receives important gene expression information as input data to classify which subtype of cancer it is.

[0277] Specifically, the computing system (1000) can classify subtypes of gynecological cancer into ovarian cancer (OV), endometrial cancer (UCEC), cervical cancer (CESC), etc.

[0278] Additionally, the computing system (1000) can perform a target task of predicting cancer grade based on the second learning data.

[0279] Here, cancer grade prediction (GRAD) is a task that predicts the malignancy and degree of cell differentiation of cancer, and can evaluate the invasiveness, growth rate, and metastatic potential of cancer based on the pathological cancer grade.

[0280] In detail, the computing system (1000) can predict the grade of prostate cancer by predicting the Gleason Score using a gene expression pattern.

[0281] In more detail, the computing system (1000) can evaluate the degree of cell differentiation of prostate cancer and measure the score of invasiveness by adding up the grades of major patterns from 1 to 5.

[0282] Additionally, the computing system (1000) can train a Gleason score model by generating and / or recombining second learning data.

[0283] At this time, the computing system (1000) can input a new patient's sample based on the expression pattern of genes, predict the Gleason score, and determine the grade of prostate cancer based on the predicted result.

[0284] Additionally, in the embodiment, the computing system (1000) can perform mutation classification, which is a target task, based on second learning data.

[0285] Here, mutation classification (MUT) can classify and / or predict the presence or absence of genetic mutation types such as copy number variations, deletions, and single nucleotide polymorphisms.

[0286] In detail, the computing system (1000) can train a mutation classification model by generating and / or recombining second learning data, and classify mutations such as TP53 and KRAS to predict disease progression and treatment responsiveness.

[0287] In addition, the computing system (1000) can input a new patient's sample based on the expression pattern of genes, determine whether there is a mutation in the TP53 gene, and / or predict whether there is gastric cancer, and use this to establish a treatment strategy.

[0288] In an embodiment, the computing system (1000) can determine TP53, KRAS, and BRCA1 as important genes for lung cancer using a second machine learning model and perform tasks on the genes.

[0289] Specifically, when the computing system (1000) builds a mutation classification model using the second machine learning model, it can predict that there is no mutation in the BRCA1 gene from the prediction result.

[0290] And the computing system (1000) can re-determine some of the important genes determined according to the performed task and perform the task of predicting the target using data on the second important gene determined again. (S111)

[0291] At this time, the computing system (1000) can re-perform the task by re-determining some of the important genes based on the prediction performance of the second machine learning model.

[0292] At this time, the computing system (1000) may further include a step of determining some of the important genes as second important genes through weights given to the important genes when the second machine learning model performs the task.

[0293] In an embodiment, the computing system (1000) may further include a step of determining a second important gene based on weights in the second machine learning model.

[0294] Additionally, in the embodiment, the computing system (1000) can extract expression information for the determined second important gene to generate third learning data.

[0295] Here, the third learning data may mean data containing labels according to the task for the second important genes and / or targets determined by the second machine learning model.

[0296] In an embodiment, the computing system (1000) may further include a step of re-performing the task by training a second machine learning model with third learning data.

[0297] Additionally, in an embodiment, the computing system (1000) may re-output the results of the important genes and / or tasks based on the important genes and prediction results used in the first result, if the performance of the re-executed second result is lower than that of the first result.

[0298] That is, the computing system (1000) can determine and / or predict important genes for the target task as the first result and / or the second result.

[0299] Through this, the computing system (1000) has the effect of improving the prediction performance results of the machine learning model by retraining the model with the secondarily determined genes, and identifying common important genes that can explain various biological phenomena.

[0300] In addition, the computing system (1000) effectively reflects complex cell interactions and cell heterogeneity within a tumor microenvironment by analyzing single-cell gene expression patterns, thereby having the effect of understanding cell-to-cell interactions and biological mechanisms.

[0301] Specifically, Fig. 9 shows the superior performance of the downstream task of the second machine learning model using the hdWGCNA-XGBoost gene set of the present invention compared to the traditional Bulk-RNA-based gene set.

[0302] Specifically, the second machine learning model applying the hdWGCNA-XGBoost gene set of the present invention recorded the highest AUROC in 17 (approximately 85%) of the 20 pan-Cancer downstream tasks. For example, it can be confirmed that the performance improved by up to 0.11 points or more compared to the existing reference, such as LUAD TMB prediction - AUROC 0.791, STAD MSI prediction - AUROC 0.990, and PAAD-KRAS mutation prediction - AUROC 0.904.

[0303] Furthermore, referring to Figure 10, as confirmed in the Upset plot, DPM1 was consistently selected across all 13 tasks, demonstrating its function as a pan-cancer prognostic biomarker. BAD, FKBP4, DUSP4, and LASP1 were also repeatedly selected as important across more than nine tasks, confirming that the invention system effectively captures common molecular mechanisms across cancer types.

[0304] The present invention achieves this performance using a refined gene set of 1,857 genes, effectively reducing the dimensionality by approximately 90% compared to the full set (>20k genes) and improving prediction accuracy. This reduces model training time and significantly reduces computational resources for clinical applications.

[0305] In addition, since the present invention uses a gene network built based on scRNA-seq, key gene features can be utilized as guides even when combined with heterogeneous data such as pathological images and proteomics, thereby providing high expandability as a multimodal precision medicine platform.

[0306]

[0307] The embodiments of the present invention described above may be implemented in the form of program commands that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program commands, data files, data structures, etc., either singly or in combination. The program commands recorded on the computer-readable recording medium may be specially designed and configured for the present invention or may be known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specifically configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. Hardware devices may be changed into one or more software modules to perform processing according to the present invention, and vice versa.

[0308] The specific implementations described in the present invention are exemplary embodiments and do not limit the scope of the present invention in any way. For the sake of brevity, descriptions of conventional electronic components, control systems, software, and other functional aspects of the systems may be omitted. In addition, the lines connecting or connecting members between components illustrated in the drawings are merely representative of functional connections and / or physical or circuit connections, and may be replaced or represented as various additional functional connections, physical connections, or circuit connections in an actual device. In addition, unless specifically mentioned as “essential,” “important,” etc., a component may not be absolutely necessary for the application of the present invention.

[0309] Although the detailed description of the present invention has been described with reference to preferred embodiments of the present invention, it will be understood by those skilled in the art or having ordinary knowledge in the art that various modifications and changes can be made to the present invention without departing from the spirit and technical scope of the present invention as set forth in the claims below. Accordingly, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification, but should be defined by the claims.

[0310] The present invention is a data processing technology that selects genes related to a task target using a machine learning model using RNA sequencing data and performs a task based on the selected genes, and therefore has industrial applicability.

Claims

1. A method for determining multiple genes related to specific biological characteristics based on biodata by a computing system including a memory and a processor, A step of obtaining a single cell gene expression profile of a tissue sample for a higher biological characteristic including the target, which is a specific biological characteristic; A step of generating gene correlation information by clustering multiple genes that co-express the single cell gene expression profile obtained above; A step of determining the gene importance for the target through a first machine learning model according to the generated gene correlation information; and A step of determining a plurality of genes as important genes for the task of the target according to the gene importance of the determined target Methods for screening and predicting genes related to biological traits.

2. In paragraph 1, The step of obtaining the single cell gene expression profile is as follows: A step of obtaining a single cell gene expression profile for a tissue sample for a biological characteristic having a specific biological characteristic of the target as a superordinate concept. Methods for screening and predicting genes related to biological traits.

3. In paragraph 2, The step of obtaining the single cell gene expression profile is as follows: A step of obtaining a single cell gene expression profile of a tissue sample for the first type of cancer as the target and a plurality of tissue samples for each of the remaining cancer types. Methods for screening and predicting genes related to biological traits.

4. In paragraph 1, The step of generating the above genetic correlation information is: A step of performing network analysis on the single cell gene expression profile and calculating the correlation between genes as a weight, A step of clustering multiple co-expressed genes according to the weights between the genes calculated above. Methods for screening and predicting genes related to biological traits.

5. In paragraph 4, The step of generating the above genetic correlation information is: It includes a step of matching genes to each node and assigning weights between genes to edges between nodes. Methods for screening and predicting genes related to biological traits.

6. In paragraph 5, The step of generating the above genetic correlation information is: A step of calculating a correlation coefficient for inter-gene correlation based on the single-cell gene expression profile, A step of creating a weighted network based on the correlation coefficients between genes calculated above, A step of grouping nodes in the above-generated network based on a clustering algorithm and classifying co-expressed genes into gene modules. Methods for screening and predicting genes related to biological traits.

7. In paragraph 6, The step of generating the above genetic correlation information is: A step of performing high-dimensional weighted gene co-expression network analysis (hdWGCNA) based on the single-cell gene expression profile is included. Methods for screening and predicting genes related to biological traits.

8. In paragraph 1, The step of determining the gene importance for the above target through the first machine learning model is: A step of generating first learning data by adding information about the tissue sample as a label based on the gene correlation information to the single cell gene expression profile. Methods for screening and predicting genes related to biological traits.

9. In paragraph 8, The step of generating the above first learning data is: A step of adding labels to the type and characteristics of the tissue sample from which the single cell gene expression profile is generated, A step of reflecting the gene expression information of the single cell gene expression profile into the network of the gene correlation information. Methods for screening and predicting genes related to biological traits.

10. In paragraph 8, The step of determining the above gene importance through the first machine learning model is: A step of calculating gene importance by converting the correlation to the target into a numerical value based on the frequency of genes used to predict the target through a tree-based first machine learning model with the first learning data is included. Methods for screening and predicting genes related to biological traits.

11. In paragraph 1, The step of determining the above important genes is: A step of determining a gene whose gene importance is greater than a preset value as the important gene, At least one step of determining the n genes set above based on the gene importance as the important genes is included. Methods for screening and predicting genes related to biological traits.

12. In paragraph 1, A step of generating second learning data by reconstructing single-cell gene expression profiles for the above-mentioned important genes, Further comprising a step of learning a second machine learning model based on the second learning data and performing a task for the target. Methods for screening and predicting genes related to biological traits.

13. In paragraph 12, The step of generating second learning data by reconstructing the single cell gene expression profile is as follows: A step of extracting gene expression information for the above important genes and generating second learning data is included. Methods for screening and predicting genes related to biological traits.

14. In paragraph 12, A step of performing a task for the target through the second machine learning model and then determining some of the important genes as second important genes through the weights given to the important genes by the second machine learning model when performing the task is further included. Methods for screening and predicting genes related to biological traits.

15. In paragraph 14, A step of extracting gene expression information for the second important gene determined above to generate third learning data, A step of re-performing a task by training the second machine learning model with the third learning data generated above. Methods for screening and predicting genes related to biological traits.

16. At least one memory; and At least one processor for performing selection and prediction of biological trait-related genes by reading out at least one application stored in the memory; wherein the instructions of the processor are, A step of obtaining a single cell gene expression profile of a tissue sample for a higher biological characteristic including the target, which is a specific biological characteristic; A step of generating gene correlation information by clustering multiple genes that co-express the single cell gene expression profile obtained above, A step of determining the gene importance for the target based on the gene correlation information generated above through a first machine learning model, Includes a command for performing a step of determining a plurality of genes as important genes for the task of the target according to the gene importance of the determined target. A system for screening and predicting genes related to biological traits.

17. A plurality of neurons comprising an array including at least one register, at least one programmable logic, and at least one input interface; a plurality of synaptic circuits storing synaptic weights that adjust connection strengths between the plurality of neurons; and at least one routing network that controls data flow between the plurality of neurons; Each of the plurality of neurons further includes a field programmable gate array (FPGA) implementation for a predetermined artificial neural network that is connected to at least one other neuron through the routing network to set a transmission path of the weight. A system for screening and predicting genes related to biological traits.

18. A plurality of neurons organized in an array including at least one register, at least one microprocessor, and at least one input; and a plurality of synaptic circuits storing synaptic weights that adjust connection strengths between the plurality of neurons; Each of the plurality of neurons further includes an application specific integrated circuit (ASIC) for a given artificial neural network that is connected to at least one other neuron through one of the plurality of synaptic circuits. A system for screening and predicting genes related to biological traits.

Images