Synthetic lethality determination device of synthetic lethality relationship, and method and computer program for searching for genes in synthetic lethality relationship by using gaussian restricted boltzmann machine

Abstract

The specification of the present disclosure relates to a device, method, and computer program for searching for genes in a synthetic lethality relationship by using a Gaussian restricted Boltzmann machine. According to any one of the above-described means for solving the problem, a gene in a synthetic lethality relationship with a target gene may be output through an artificial intelligence model trained by receiving a training data set including an mRNA expression value in RNA Seq data, the presence or absence of a mutation in WES data, and a dependency score in CRISPR KO data. In addition, it is possible to increase the efficiency of anticancer treatment by using the genes in a synthetic lethality relationship calculated by using the trained artificial intelligence model. In addition, it is possible to present a treatment route with low drug resistance by using the genes in a synthetic lethality relationship calculated by using the trained artificial intelligence model.

Description

Device for determining composite lethality relationships, method for searching for genes in composite lethality relationships using a Gaussian restricted Boltzmann machine, and computer program

[0001] The specification of the present disclosure relates to an apparatus, method, and computer program for searching for synthetic lethality genes using a Gaussian Restricted Boltzmann Machine.

[0002] Genes serve as guideline systems for producing specific proteins or regulating functions in a subject's cells, and in cancer research, understanding genetic changes can be crucial for cancer diagnosis and treatment.

[0003] Genes include oncogenes, which cause cancer through mutations in genes that regulate normal cell growth and differentiation; tumor suppressor genes, which inhibit cell growth and division and protect cells from transforming into cancer; and DNA repair genes, which correct errors occurring during DNA replication or repair DNA damage caused by external factors.

[0004] Among genes, there may be two genes that have a synthetic lethal relationship. If two genes with a synthetic lethal relationship are simultaneously inactivated, they can kill the cell. Even if only one gene with a synthetic lethal relationship is inactivated, the cell will survive.

[0005] Even if a gene related to cancer is discovered, if the synthetic lethal gene combined with that gene is not known, the cancer cannot be treated with that gene alone.

[0006] Well-known synthetic lethal genes include the PARP gene and the BRCA1 / 2 gene. It has become possible to treat breast cancer or ovarian cancer dramatically by using PARP inhibitors while simultaneously killing BRCA1 / 2, which are target genes associated with breast cancer or ovarian cancer.

[0007] Therefore, for the treatment of cancer, there is a growing need to discover synthetic lethal genes associated with the disease.

[0008] The embodiments disclosed in this specification aim to provide an apparatus and purpose for outputting a gene that has a synthetic lethal relationship with a target gene through an artificial intelligence model trained using a training data set including mRNA expression values ​​in RNA Seq data, the presence or absence of mutations in WES data, and dependency scores in CRISPR KO data as input.

[0009] In addition, the purpose is to provide a device and method that enhance the efficiency of anticancer treatment by utilizing genes with synthetic lethality relationships calculated using a trained artificial intelligence model.

[0010] In addition, the purpose is to provide a device and method that propose a treatment pathway with low drug resistance using genes with synthetic lethality relationships calculated using a trained artificial intelligence model.

[0011] A method according to embodiments of the present disclosure may include: a synthetic lethality determination device inputting information regarding a first gene and a second gene; a synthetic lethality determination device executing a synthetic lethality determination model learned from input data of the first gene and the second gene; a synthetic lethality determination device inputting first input data for the first gene into the synthetic lethality determination model and inputting second input data in which the dependency score of the second gene is changed to NULL; a synthetic lethality determination device calculating a dependency score in the synthetic lethality determination model and outputting second input data having the dependency score; a synthetic lethality determination device outputting whether the first gene and the second gene have a synthetic lethality relationship based on whether they have a dependency score of 0 using output data from the synthetic lethality determination model; and a synthetic lethality determination device generating result data including whether they have a synthetic lethality relationship and transmitting the result data to an external electronic device.

[0012] The above composite lethality judgment model is trained based on a Gaussian-restricted Boltzmann machine and may include a visible layer and a hidden layer.

[0013] The above synthetic lethality determination model is trained with input data including gene expression levels, mutation status, and dependency scores, and parameter values ​​can be optimized by performing a fine-tuning process using data on gene pairs with well-known synthetic lethality relationships.

[0014] The above synthetic lethality determination model may be modified based on parameter values ​​determined by the type of gene, whether there is a mutation in the gene, the type of cancer that developed, and the type of tissue cell.

[0015] The above synthetic lethality determination model can be designed to output a gene with a dependency score of 0 as the second gene when the expression level of the first gene is set to 0 and the mutation status is set to 0.

[0016] A computer program according to an embodiment of the present invention may be stored on a medium to execute any one of the methods according to an embodiment of the present invention using a computer.

[0017] In addition to the above, other methods for implementing the present invention, other systems, and computer-readable recording media for recording a computer program for executing said methods are further provided.

[0018] Other aspects, features, and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

[0019] According to any one of the aforementioned means for solving the problem, a target gene and a gene that has a synthetic lethal relationship can be output through an artificial intelligence model trained using a training data set that includes mRNA expression values ​​from RNA Seq data, the presence or absence of mutations from WES data, and dependency scores from CRISPR KO data as input.

[0020] In addition, the efficiency of anticancer treatment can be enhanced by utilizing genes with synthetic lethal relationships derived from a trained artificial intelligence model.

[0021] In addition, by utilizing genes with synthetic lethality relationships derived from a trained artificial intelligence model, it is possible to suggest treatment pathways with low drug resistance.

[0022] FIG. 1 illustrates a synthetic lethal screening network system including a server and a user terminal according to one embodiment of the present disclosure.

[0023] FIG. 2 is a drawing for explaining the detailed configuration of a synthetic lethality determination device according to embodiments of the present disclosure.

[0024] FIG. 3 is a drawing showing the detailed structure of the composite lethality judgment unit (120).

[0025] FIG. 4 is a diagram showing data transmission and reception between an electronic device (T) and a learning device (300).

[0026] FIG. 5 is an example diagram of the format of input data for training a model according to embodiments of the present disclosure.

[0027] Figure 6 is a diagram illustrating the structure of a composite lethality judgment model based on a Gaussian restricted Boltzmann machine.

[0028] Figure 7 is a diagram of the process for determining the synergistic lethality relationship for BRCA1 / 2 and PARP.

[0029] Figure 8 is a diagram of the process of determining the genes in a synthetic lethal relationship using TP53 as the target gene.

[0030] FIG. 9 is a flowchart of a method for determining synthetic lethality according to embodiments of the present disclosure.

[0031] The structure and operation of the present invention will be described in detail below with reference to embodiments of the present invention illustrated in the attached drawings.

[0032] The present invention is capable of various modifications and may have various embodiments; specific embodiments are illustrated in the drawings and described in detail in the detailed description. The effects and features of the present invention, and the methods for achieving them, will become clear by referring to the embodiments described below in detail together with the drawings. However, the present invention is not limited to the embodiments disclosed below but can be implemented in various forms.

[0033] 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.

[0034] In the following, terms described as “upper” or “upper” may include not only those directly above in contact, but also those above without contact.

[0035] In the following embodiments, terms such as first, second, etc. are used not in a limiting sense, but for the purpose of distinguishing one component from another component.

[0036] In the following embodiments, singular expressions include plural expressions unless the context clearly indicates otherwise.

[0037] In the following embodiments, terms such as "include" or "have" mean that the features or components described in the specification are present, and do not preclude the possibility that one or more other features or components may be added.

[0038] In the drawings, the size 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 depicted arbitrarily for convenience of explanation, so the present invention is not necessarily limited to what is illustrated.

[0039] Additionally, terms such as “…part,” “…area,” etc., as described in this specification may refer to a unit that processes at least one function or operation.

[0040] WGS data is a method of analyzing the entire DNA sequence of an organism, and as data obtained by decoding the base sequences of all chromosomes, it may include human genetic information, single nucleotide variations, structural variations, copy number variations, etc.

[0041] According to embodiments of the present disclosure, a synthetic lethality determination device may select and provide a gene that has a synthetic lethality relationship with a target gene. The synthetic lethality determination device may determine a new gene that has a synthetic lethality relationship with the target gene. The synthetic lethality determination device may generate input data including sequence data (RNA-Seq), whole-exome sequencing (WES), and a correlation value with survival rate, and select and provide a gene that has a synthetic lethality relationship with the target gene. The input data may include expression levels based on sequence data, mutation status based on whole-exome sequencing, and dependency scores based on correlation values ​​with survival rate. The input data is obtained for each gene and may be obtained for multiple cell lines. The synthetic lethality determination device may output output data using a synthetic lethality determination model trained with a training data set of expression levels, mutation status, and dependency scores as input. The synthetic lethality determination model may continuously output dependency scores based on gene expression levels based on a Gaussian-restricted Boltzmann machine. The trained model can undergo a fine-tuning process to change parameter values ​​based on the characteristics of the input data. By setting the parameters to these modified values, the performance of the trained model can be improved.

[0042] FIG. 1 illustrates a synthetic lethal screening network system including a server and a user terminal according to one embodiment of the present disclosure.

[0043] The synthetic lethality screening network system (1) of the present disclosure may include a server (20) and at least one user terminal (11 to 16). The server (20) may provide various online activities through a network. The server (20) may provide online activities to at least one user terminal (11 to 16) simultaneously. Here, the online activities may include providing a synthetic lethality determination service.

[0044] The synthetic lethality determination service searches for and provides genes that have a synthetic lethality relationship with a specified gene, and more specifically, it may search for and provide genes with a synthetic lethality relationship using a Gaussian restricted Boltzmann machine.

[0045] According to one embodiment of the present disclosure, the term server (20) may include a single server, a collection of servers, a cloud server, etc., but is not limited to the above examples. The server (20) may provide various online activities and may include a database that stores data for online activities. As previously described, the server (20) may be a composite lethality determination device.

[0046] According to one embodiment of the present disclosure, a network refers to a connection established (or formed) using any communication method, and may refer to a communication network connected through any communication method that transmits and receives data between terminals or between a terminal and a server.

[0047] The term "all communication methods" may include all communication methods, such as communication through a specified communication standard, a specified frequency band, a specified protocol, or a specified channel. For example, it may include communication methods via Bluetooth, BLE, Wi-Fi, Zigbee, 3G, 4G, 5G, 6G, LTE, and ultrasound, and may include short-range communication, long-range communication, wireless communication, and wired communication. Of course, it is not limited to the above examples.

[0048] According to one embodiment of the present disclosure, a short-range communication method may mean a communication method in which communication is possible only when a device (terminal or server) performing communication is within a predetermined range, and may include, for example, Bluetooth, NFC, etc. A long-range communication method may mean a communication method in which a device performing communication can communicate regardless of distance. For example, a long-range communication method may mean a method in which two devices performing communication through a repeater such as an AP can communicate even when they are more than a predetermined distance apart, and may include communication methods using cellular networks (3G, 4G, 5G, LTE) such as SMS and telephone. Of course, it is not limited to the above examples. The meaning of receiving a service using a network may include the meaning that communication between a server and a terminal can be performed through any communication method.

[0049] Throughout the specification, the term "at least one user terminal (11 to 16)" may include not only a personal computer (11), a tablet (12), a cellular phone (13), a laptop (14), a smartphone (15), and a TV (16), but also various electronic devices such as personal digital assistants (PDA), portable multimedia players (PMP), navigation systems, and MP3 players, and is not limited to the examples above. As previously described, at least one user terminal (11 to 16) may be a composite lethality determination device.

[0050] According to one embodiment of the present disclosure, a server (20) can train a composite lethality determination model based on a Gaussian-restricted Boltzmann machine by inputting input data including gene expression levels, whether mutations have occurred, and dependency scores. The server (20) can output output data by inputting input data to the composite lethality determination model. The composite lethality determination model can be trained by inputting a training data set including gene expression levels, whether mutations have occurred, and dependency scores. The composite lethality determination model can determine parameter values ​​(weight values, bias values) through a fine-tuning process.

[0051] The composite lethality determination model can be trained by receiving factors such as expression levels, mutation occurrences, and dependency scores of expressed genes as input in a training dataset. Based on the results of training with the input data set, the composite lethality determination model can output composite lethality correlation values ​​between genes while predicting dependency scores between genes. Since the composite lethality determination model is trained based on a Gaussian-restricted Boltzmann machine, it can continuously calculate mutation occurrences and dependency scores according to changes in gene expression levels. The composite lethality determination model can be trained using a set of gene expression levels, mutation occurrences, and dependency scores among well-known genes. As a result of training, the composite lethality determination model can output a graph corresponding to the continuous changes of each factor.

[0052] The synthetic lethality judgment model can store parameter values ​​that improve the accuracy of the output by adjusting model parameters (weight values, bias values). The synthetic lethality judgment model can output a dependency score by adjusting parameter values. The synthetic lethality judgment model can adjust model parameters (weight values, bias values) so that the output dependency score matches the actual dependency score. In this case, the model parameters (weight values, bias values) can be optimized while learning with continuous real values. The synthetic lethality judgment model can be retrained by changing the batch size and epochs in relation to the training data set. The synthetic lethality model can be executed while adjusting parameter values ​​considering the characteristics of the input data. The synthetic lethality model can output data with parameter values ​​determined by considering the type of input gene. The synthetic lethality model can output data with parameter values ​​determined by considering whether the input gene has a mutation. The synthetic lethality model can output data with parameter values ​​determined by considering the input tissue cells.

[0053] According to one embodiment of the present disclosure, at least one of the user terminals (11 to 16) can input input data including gene expression levels, whether mutations occurred, and dependency scores to train a composite lethality determination model based on a Gaussian-restricted Boltzmann machine. At least one of the user terminals (11 to 16) can input input data to the composite lethality determination model and output output data. The composite lethality determination model can be trained by inputting a training data set including gene expression levels, whether mutations occurred, and dependency scores. The composite lethality determination model can determine parameter values ​​(weight values, bias values) through a fine-tuning process.

[0054] In addition, according to one embodiment of the present disclosure, a data generation network system (1) can train a composite lethality determination model based on a Gaussian-restricted Boltzmann machine by inputting input data including gene expression levels, whether mutations occur, and dependency scores. The data generation network system (1) can output output data by inputting input data to the composite lethality determination model. The composite lethality determination model can be trained by inputting a training data set including gene expression levels, whether mutations occur, and dependency scores. The composite lethality determination model can determine parameter values ​​(weight values, bias values) through a fine-tuning process.

[0055] This is explained in more detail below.

[0056] FIG. 2 is a drawing for explaining the detailed configuration of a synthetic lethality determination device according to embodiments of the present disclosure.

[0057] As illustrated in FIG. 2, a composite lethality determination device (100) according to some embodiments may include a processor (110), an input / output unit (130), a memory (140), a communication unit (150), and a composite lethality determination unit (120). However, not all components illustrated in FIG. 2 are essential components of the composite lethality determination device (100). The composite lethality determination device (100) may be implemented with more components than those illustrated in FIG. 2, or with fewer components than those illustrated in FIG. 2. The composite lethality determination device (100) may be a user terminal, a server, a data generation service network system, or a separate device.

[0058] According to one embodiment of the present disclosure, the processor (110) typically controls the overall operation of the composite lethality determination device (100). For example, the processor (110) can generally control the components included in the composite lethality determination device (100) by executing a program stored in the composite lethality determination device (100).

[0059] According to one embodiment of the present disclosure, a processor (110) can train a composite lethality determination model based on a Gaussian-restricted Boltzmann machine by inputting input data including gene expression levels, whether mutations have occurred, and dependency scores. The processor (110) can output output data by inputting input data to the composite lethality determination model. The composite lethality determination model can be trained by inputting a training data set including gene expression levels, whether mutations have occurred, and dependency scores. The composite lethality determination model can determine parameter values ​​(weight values, bias values) through a fine-tuning process.

[0060] The processor (110) is configured to control the composite lethality determination device (100) overall. Specifically, the processor (110) controls the overall operation of the composite lethality determination device (100) using various programs stored in the storage medium (150) of the composite lethality determination device (100). For example, the processor (110) may include a CPU, RAM, ROM, and a system bus. Here, the ROM is a configuration in which an instruction set for system booting is stored, and the CPU copies the operating system stored in the composite lethality determination device (100) to the RAM according to the instructions stored in the ROM, and executes the O / S to boot the system. Once the system booting is complete, the CPU can copy various stored applications to the RAM and execute them to perform various operations. Although the composite lethality determination device (100) has been described above as including only one CPU, it may be implemented with multiple CPUs (or DSP, SoC, etc.) during implementation.

[0061] According to one embodiment of the present invention, the processor (110) may be implemented as a digital signal processor (DSP) that processes digital signals, a microprocessor, or a TCON (Time controller). However, it is not limited thereto, and may include or be defined by one or more of a central processing unit (CPU), a Micro Controller Unit (MCU), a micro processing unit (MPU), a controller, an application processor (AP), a communication processor (CP), or an ARM processor. Additionally, the processor (110) may be implemented as a System on Chip (SoC) or Large Scale Integration (LSI) with a built-in processing algorithm, or may be implemented in the form of a Field Programmable Gate Array (FPGA).

[0062] According to one embodiment of the present disclosure, the input / output unit (130) can display an interface generated by the memory (140). According to one embodiment of the present invention, the input / output unit (130) can display a user interface for input user input. The input / output unit (130) can output stored graphic data, visual data, auditory data, and vibration data under the control of the memory (140).

[0063] The input / output unit (130) can be implemented as a display panel of various forms. For example, the display panel can be implemented as a display technology such as LCD (Liquid Crystal Display), OLED (Organic Light Emitting Diodes), AM-OLED (Active-Matrix Organic Light-Emitting Diode), LcoS (Liquid Crystal on Silicon), or DLP (Digital Light Processing). Additionally, the input / output unit (130) may be combined with at least one of the front area, side area, and rear area of ​​the display panel in the form of a flexible display.

[0064] The input / output unit (130) can be implemented as a touch screen with a layer structure. The touch screen can have a function to detect not only the display function but also the touch input location, the touched area, and the touch input pressure, and can also have a function to detect not only real touch but also proximity touch.

[0065] The input / output unit (130) may include a user interface for inputting various information to the synthetic lethality determination device (100).

[0066] According to one embodiment of the present disclosure, the memory (140) may store a program for processing and controlling the processor (110) and / or the synthetic lethality determination unit (120), and may also store data input to or output from the synthetic lethality determination device (100). According to one embodiment of the present disclosure, the memory (140) may store information such as gene expression levels, mutation status, dependency scores, and gene pairs in a synthetic lethality relationship. The memory (140) may include a database that stores the above information.

[0067] According to one embodiment of the present disclosure, the memory (140) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, a magnetic disk, and an optical disk. Additionally, according to one embodiment of the present disclosure, programs stored in the memory (140) may be classified into a plurality of modules according to their functions.

[0068] According to one embodiment of the present disclosure, the communication unit (150) can communicate with an external device of the processor (110). For example, the communication unit (150) can communicate with an external device, such as a payment server, an authentication server, or a user terminal, under the control of the processor (110). Additionally, the communication unit (150) may obtain user information or user input through communication with an external interface.

[0069] FIG. 3 is a drawing showing the detailed structure of the composite lethality judgment unit (120).

[0070] The composite lethality determination unit (120) may include a data input unit (121), a composite lethality calculation unit (122), and a result data generation unit (123).

[0071] The data input unit (121) can acquire genomic data by inputting biological cells, tissue cells, etc. collected from the body into a device such as a qPCR device, NGS equipment, or a microarray scanner. The genomic data may include the expression level and mutation status of each gene. Dependency scores between genes may be obtained from an external device. The data input unit (121) may receive the gene expression level, mutation status, and dependency score. The data input unit (121) may transmit a data set containing the gene expression level, mutation status, and dependency score to a learning device (300) to train a synthetic lethality judgment model. The gene expression level may be an mRNA expression value obtained from RNA-Seq data. The expression level value may be a real value in the range of 0 to 15.0, normalized in the form of log2(TPM+1). The mutation status may be 0 or 1, which causes loss of function of the gene in WES data. The dependency score may be a real value in the range of 0 to 1.0. The format of the input data can be (expression, mutation, dependency score). For example, the BRCA1 gene can be entered in the format ([12.9, 1.0, 3.2, ...], [1, 0, 0, ...], [0.932, 0.324, 0.235, ...]).

[0072] The synthetic lethality calculation unit (122) can input the input gene into a synthetic lethality judgment model and output a gene that has a synthetic lethality relationship with the gene. The synthetic lethality calculation unit (122) can set the gene expression amount between the first and second genes as a first value and the mutation status as a second value based on the results learned from the synthetic lethality judgment model, and output a dependency score in such cases. If the dependency score is a determined third value, the first and second genes may be a pair of genes that have a synthetic lethality relationship. Here, the first value may be 0. The second value may be 0. The third value may be 0.

[0073] The synthetic lethality calculation unit (122) can output pairs of genes through a synthetic lethality judgment model in which the gene expression value is the first value, the mutation status is the second value, and the dependency score is the third value. At this time, the synthetic lethality judgment model can output pairs of genes having a point where the gene expression value changes to the first value, the mutation status changes to the second value, and the dependency score changes to the third value. Pairs of genes having such a point can be inferred to interact.

[0074] The synthetic lethality calculation unit (122) can execute a synthetic lethality judgment model learned from input data of the first gene and the second gene. At this time, the synthetic lethality judgment model may be set with parameter values ​​learned from the data of the first gene and the second gene.

[0075] The synthetic lethality calculation unit (122) can input first input data for the first gene into the synthetic lethality judgment model and input second input data in which the dependency score in the second gene input data is changed to NULL.

[0076] The composite lethality calculation unit (122) can calculate a dependency score for the second input data and output the second input data including the calculated dependency score in a composite lethality judgment model based on a Gaussian restricted Boltzmann machine.

[0077] The result data generation unit (123) can generate result data for pairs of genes inferred based on a synthetic lethality judgment model.

[0078] The result data generation unit (123) can output whether the first gene and the second gene have a synthetic lethal relationship based on whether they have a dependency score of 0 using output data from the synthetic lethal judgment model.

[0079] FIG. 4 is a diagram showing data transmission and reception between an electronic device (T) and a learning device (300).

[0080] According to embodiments of the present disclosure, an electronic device (T) may be a device for inputting a learning data set, a device for requesting a gene with a synthetic lethal relationship for a target gene, and a device for receiving a synthetic lethal determination model and outputting a pair of genes with a synthetic lethal relationship. The electronic device (T) may be a computing device comprising a processor, a communication unit for communication, a memory for storing data, and a storage medium. The electronic device (T) may further include components other than those mentioned.

[0081] The learning device (300) can receive a data set including {expression amount, mutation status, dependency score} and train a synthetic lethality judgment model. The learning device (300) can receive the data set by communicating with a plurality of electronic devices. The data set can be determined for each gene, such as a data set for a first gene. The data set for the first gene may include the gene expression amount value, mutation status, and dependency score of each gene according to the change in the expression amount of the first gene. In this way, the model can be trained with the data set for each gene. At this time, the first gene may be a gene related to the development of cancer.

[0082] The learning device (300) can train a synthetic lethal judgment model through the visible layer, hidden layer, etc. of a Gaussian restricted Boltzmann machine.

[0083] The model learning unit (342) can train a synthetic lethality determination model based on a Gaussian-restricted Boltzmann machine. The model learning unit (342) can train a synthetic lethality determination model by receiving factors such as gene expression levels, whether mutations occur, and dependency scores as a training data set. The synthetic lethality determination model can output dependency scores, expression levels, and mutation status as continuous values ​​based on a Gaussian-restricted Boltzmann machine. For example, for the BRCA1 gene, a training data set can be input in the format ([12.9, 1.0, 3.2, ...], [1, 0, 0, ...], [0.932, 0.324, 0.235, ...]).

[0084] The model learning unit (342) can fine-tune the model using information on well-known synthetic lethality gene pairs. The model learning unit (342) can fine-tune the model to adjust parameter values. More specifically, the model learning unit (342) can output factors at the point where the predicted dependency score becomes 0 by adjusting the parameter values. The model learning unit (342) can output factors (gene expression amount, mutation status value, etc.) at the point where the dependency score becomes 0. The model learning unit (342) can evaluate the learning results of the model through these factor values. For example, if there is a factor at the point where the dependency score becomes 0, the learning model can be evaluated as learned.

[0085] The model learning unit (342) can store an optimized parameter value learned from a learning data set. The model learning unit (342) can store a first parameter value obtained by learning from a learning data set for a first tissue cell in correspondence with the first tissue cell. The learning data set may include a gene expression value, a mutation status, a dependency score value, etc. The gene expression value may be a numerical value proportional to the amount of gene expressed, and may be a continuous real number value within 0 to 16. The mutation status may be a value of 1 if there is a mutation, and 0 if there is no mutation. The mutation value may be set to 0 or 1.

[0086] The model learning unit (342) can apply the first parameter value to the learning model in conjunction with the first tissue cell, the disease of the first tissue cell, etc. For example, if the first tissue cell is a breast cancer cell, the model that outputs the gene with a synthetic lethal relationship for the gene related to the breast cancer cell can output output data based on the parameter value learned from the breast cancer cell. The model that outputs the gene with a synthetic lethal relationship for a patient with a mutation of the first gene in the first tissue cell can output output data based on the parameter value learned from the learning data set with the mutation of the first gene. As described above, the parameter values ​​can be stored according to the tissue cell, gene, whether a mutation has occurred, etc.

[0087] A model trained based on a Gaussian-restricted Boltzmann machine can output expression values, which are parameters of the output data, as continuous real values. In the case of a general model, since the output data is generated based on the training data set used for training, it is not output as a continuous value but as an actual measured value.

[0088] The model learning unit (342) can evaluate the reliability of output data containing an expression amount value of 0 as low for values ​​output from the learned model and determine such output data as incorrect data.

[0089] The model training unit (342) can receive data through the data input unit (341) that is BRCA1 (expression=no correlation & mutation=1) & PARP1 (expression>1.0 & mutation=1) and has a dependency score of 0.5 or higher.

[0090] The data input unit (341) can change the mut of PARP1 to 0 and remove the dependency score in relation to the input data to a null value.

[0091] The model training unit (342) can train the model using the above data set as input. The model training unit (342) can store parameter values ​​when the dependency score is 0 or as close as possible to 0 through iterative training as a result of training.

[0092] The model learning unit (342) can input an input to the learned model that, when the expression of the gene PARP1 is 0, changes the mutation value of PARP1 to 0 and removes the dependency score to change it to a null value. At this time, the parameter value of the learned model can be set to a value that was previously set as a result of learning.

[0093] The model learning unit (342) can train a model by inputting data on gene pairs with a well-known synthetic lethal relationship and can store relationship equations between factors included in the data as a result of training. The model learning unit (342) can train a model by inputting data such as expression values, mutation status, and dependency scores for gene pairs, and can output optimized parameter values ​​through the process of inputting the input data of gene pairs with a well-known synthetic lethal relationship into the trained model. The trained model can output output data according to the input data by changing the optimized parameter values ​​according to the input data. The trained model can input a gene expression value by changing it and output a dependency score value corresponding to it as output data. The trained model can output a dependency score value by inputting the gene and gene expression amount of the input data. The trained model can output genes with a synthetic lethal relationship according to cell type, gene expression status, mutation status, etc. by adjusting the parameter values.

[0094] The learning device (300) may include a communication unit (310) that communicates with an external electronic device, a processor (320) that reads and processes computer-readable instructions, and a memory (330) that stores data for reading instructions.

[0095] A synthetic lethality determination model learned through a learning device (300) can be learned by receiving factors such as expression levels, mutation occurrence, and dependency scores of expressed genes as input data sets. As a result of learning from the input data sets, the synthetic lethality determination model can predict dependency scores between genes and output synthetic lethality correlation values ​​between genes. The synthetic lethality determination model is learned based on a Gaussian-restricted Boltzmann machine, so that mutation occurrence and dependency scores can be continuously calculated according to changes in gene expression levels. The synthetic lethality determination model can be learned using a set of gene expression levels, mutation occurrence, and dependency scores between well-known genes. As a result of learning, the synthetic lethality determination model can output a graph according to the continuous changes of each factor.

[0096] A composite lethality decision model can output a dependency score by adjusting model parameters (weight values, bias values) related to the performance of the learning model. The composite lethality decision model can adjust model parameters (weight values, bias values) so that the output dependency score matches the actual dependency score. In this case, the model parameters (weight values, bias values) can be optimized while learning with continuous real values. The composite lethality decision model can be retrained by changing the batch size and epochs in relation to the training data set.

[0097] FIG. 5 is an example diagram of the format of input data for training a model according to embodiments of the present disclosure.

[0098] A synthetic lethality determination model is trained using a vector containing expression values, mutation status, and dependency scores as input data. The input data can be generated per gene. An electronic device acquires a matrix of gene expression values ​​(M1), a matrix of mutation status (M2), and a matrix of dependency scores (M3) for a cell line, and can convert these matrices into input data of expression values ​​(V1), mutation status (V2), and dependency scores (V3).

[0099] Figure 6 is a diagram illustrating the structure of a composite lethality judgment model based on a Gaussian restricted Boltzmann machine.

[0100] Since the synthetic lethality judgment model is designed based on a Gaussian-restricted Boltzmann machine, it includes a visible layer and a hidden layer; by inputting data for each gene, parameter values ​​(W1, W2, W3, …, Wn) for weights and biases for each gene can be obtained. Output data can be produced by setting these parameter values ​​(W1, W2, W3, …, Wn) for weights and biases, respectively.

[0101] Figure 7 is a diagram of the process for determining the synergistic lethality relationship for BRCA1 / 2 and PARP.

[0102] The input data can be implemented as a training data set and used for model training. The synthetic lethality determination device according to the embodiments of the present disclosure can generate input data (D11) of expression levels, mutation status, and dependency scores obtained from a plurality of cell lines for the gene BRCA1 / 2.

[0103] For the gene PARP, input data (D12) of expression levels, mutation status, and dependency scores obtained from multiple cell lines can be generated.

[0104] A composite lethality determination device according to embodiments of the present disclosure can input input data (D11, D12) and calculate parameter values ​​such as weight values ​​and bias values ​​from a learned model.

[0105] A synthetic lethality determination device according to embodiments of the present disclosure generates input data (D21) in which the expression value and the mutation status value are set to 0 in the input data (D11) of BRCA1 / 2, and sets the dependency score to NULL in the input data (D22) of PARP.

[0106] A synthetic lethality determination device according to embodiments of the present disclosure inputs D21 and D22 into a trained model, outputs output data D32 filled with dependency scores of PARP genes for BRCA1 / 2, and searches for and outputs a vector with a dependency score of 0 among D32.

[0107] At this time, the trained model may be set with parameter values ​​determined as a result of previous training. If there is no vector with a dependency score of 0, the training model can be operated by changing the parameter value from the first parameter to the second parameter.

[0108] Depending on the types of genes and target genes, the trained model can be optimized with different parameter values.

[0109] Figure 8 is a diagram of the process of determining the genes in a synthetic lethal relationship using TP53 as the target gene.

[0110] A composite lethality determination device according to embodiments of the present disclosure can acquire input data from a plurality of cell lines in which TP53 is expressed, as mentioned in FIG. 7, and train a composite lethality determination model based on a Gaussian-restricted Boltzmann machine while inputting such input data. Through the training process, the parameter values ​​(weight values, bias values) of the composite lethality determination model are optimized.

[0111] A synthetic lethality determination device according to embodiments of the present disclosure may input data for Gene A, Gene B, Gene C, etc. into a model by setting the expression level value and mutation status value in the input data of TP53 to 0 in order to search for genes in a synthetic lethality relationship with the target gene TP53, and input data for Gene A, Gene B, Gene C, etc. into the model. At this time, the input data for Gene A, Gene B, Gene C, etc. may be data in which the dependency score is left empty as NULL.

[0112] The trained model infers and outputs a dependency score for each of the other genes, and can select genes with an inferred dependency score of 0 and output information about these genes.

[0113] Genes with data having a dependency score of 0 can be output as being in a synthetic lethal relationship with the target gene TP53.

[0114] FIG. 9 is a flowchart of a method for determining synthetic lethality according to embodiments of the present disclosure.

[0115] In S110, the synthetic lethality determination device (100) can obtain information regarding the first gene and the second gene. The information regarding the first gene and the second gene may be received through a network or input by a user through an input / output device. Input data for the first gene and input data for the second gene may be stored in advance. Input data for the first gene and input data for the second gene include expression level values, mutation status, and dependency score values, and may include values ​​measured in multiple cell lines.

[0116] In S120, the synthetic lethality determination device (100) can execute a synthetic lethality determination model learned from input data of the first gene and the second gene. At this time, the synthetic lethality determination model may be set with parameter values ​​learned from the data of the first gene and the second gene.

[0117] In S130, the synthetic lethality determination device (100) can input first input data for a first gene into a synthetic lethality determination model and input second input data in which the dependency score in the input data of a second gene is changed to NULL.

[0118] In S140, the composite lethality determination device (100) can calculate a dependency score for the second input data in a composite lethality determination model based on a Gaussian restricted Boltzmann machine and output the second input data including the calculated dependency score.

[0119] In S150, the synthetic lethality determination device (100) can output whether the first gene and the second gene have a synthetic lethality relationship based on whether they have a dependency score of 0 using output data from the synthetic lethality determination model.

[0120] Furthermore, although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. It is understood that various modifications can be made by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention.

[0121] Accordingly, the scope of the present invention should not be limited to the embodiments described above, and all scopes equivalent to or equivalently modified from the claims set forth below, as well as the claims set forth below, shall be considered to fall within the scope of the concept of the present invention.

Claims

1. A synthetic lethality determination device inputs information regarding a first gene and a second gene; The above synthetic lethality determination device executes a synthetic lethality determination model learned with input data of the first gene and the second gene; The above synthetic lethality determination device inputs first input data for the first gene into the above synthetic lethality determination model, and inputs second input data in which the dependency score in the second input data of the second gene is changed to NULL; The above composite lethality determination device calculates a dependency score in the above composite lethality determination model and outputs second input data having the dependency score; The above synthetic lethality determination device outputs whether the first gene and the second gene have a synthetic lethality relationship based on whether they have a dependency score of 0 using output data from the above synthetic lethality determination model; and A method for determining a composite lethality, comprising the step of the above composite lethality determination device generating result data including whether there is a composite lethality relationship and transmitting said result data to an external electronic device.

2. In Paragraph 1, The above composite lethality determination model is, A composite lethality determination method that is trained based on a Gaussian-restricted Boltzmann machine and includes a visual layer and a hidden layer.

3. In Paragraph 1, The above composite lethality determination model is, A method for determining synthetic lethality, which is trained on input data including the expression levels, mutation status, and dependency scores of each gene for multiple cell lines, and optimizes parameter values ​​by performing a fine-tuning process using data on gene pairs with well-known synthetic lethality relationships.

4. In Paragraph 3, The above composite lethality determination model is, A method for determining synthetic lethality that can be changed to parameter values ​​determined based on the type of gene, whether there is a mutation in the gene, the type of cancer that developed, and the type of tissue cell.

5. Includes one or more processors, communication units and memory, and The above processor inputs information about the first gene, and Enter information about the first gene and the second gene, and Executing a synthetic lethality judgment model trained with input data of the first gene and the second gene, and In the above synthetic lethality judgment model, first input data for the above first gene is input, and second input data is input in which the dependency score in the second input data of the above second gene is changed to NULL, and In the above composite lethality judgment model, a dependency score is calculated, and second input data having the said dependency score is output, and Using the output data from the above synthetic lethality determination model, output whether the first gene and the second gene have a synthetic lethality relationship based on whether they have a dependency score of 0, and A composite lethality determination device that generates result data including whether there is a composite lethality relationship and transmits said result data to an external electronic device.

6. In Paragraph 5, The above composite lethality determination model is, A composite lethality determination device that is trained based on a Gaussian-restricted Boltzmann machine and includes a visible layer and a hidden layer.

7. In Paragraph 5, The above composite lethality determination model is, A composite lethality determination device that is trained with input data including gene expression levels, mutation status, and dependency scores, and optimizes parameter values ​​by performing a fine-tuning process through data on gene pairs with well-known composite lethality relationships.

8. In Paragraph 7, The above composite lethality determination model is, A synthetic lethality determination device that can be changed to parameter values ​​determined based on the type of gene, whether there is a mutation in the gene, the type of cancer that developed, and the type of tissue cell.

9. A computer program stored on a computer-readable storage medium to execute the method of any one of paragraphs 1 through 4 using a computer.

Images