KR102554278B1 - Method for predicting structure of protein using twist-based structure updates - Google Patents

Abstract

The present disclosure provides a method for predicting the structure of a protein using a neural network model, performed by a computing device, comprising the steps of obtaining information on each residue of a protein; updating information on each residue by utilizing update information associated with a twist structure; and adjusting the structure of the protein based on the updated information on each residue.

Description

Method for predicting the structure of a protein using twist-based structural update {METHOD FOR PREDICTING STRUCTURE OF PROTEIN USING TWIST-BASED STRUCTURE UPDATES}

The present disclosure relates to a method of predicting the structure of a protein, and more particularly, to a method of predicting the structure of a protein (e.g., a structure before binding or a bound structure of a protein) based on updated information associated with a folding structure. will be.

Prediction of protein structure (eg, pre-binding structure or bound structure of a protein) is one of the very important steps in drug development. Today, as technology for neural network models advances, protein structure prediction using neural network models has been attempted, and remarkable results have been achieved.

On the other hand, in relation to protein structure prediction, the conventional method predicts the protein structure through a method of independently updating the position and orientation information of residues without considering the binding relationship with each other.

Thus, conventional methods have been difficult to model the flexibility of protein structures. For example, in the conventional method, it is difficult to model the motion of a rigid body unit or backbone unit included in a protein structure.

The present disclosure has been derived at least based on the technical background salpin above, but the technical problem or object of the present disclosure is not limited to solving the above salpin problems or disadvantages. That is, the present disclosure may cover various technical issues related to the content to be described below, in addition to the technical issues discussed above.

The present disclosure aims to predict the structure of a protein (eg, a pre-binding structure or a bound structure of a protein) based on updated information associated with a fold structure.

On the other hand, the technical problem to be achieved by the present disclosure is not limited to the above-mentioned technical problem, and may include various technical problems within a range apparent to those skilled in the art from the contents to be described below.

In order to solve the above problems, a method for predicting the structure of a protein using a neural network model performed by a computing device is disclosed. Here, the method includes obtaining information on each residue of a protein; updating information on each residue by utilizing update information associated with a twist structure; and adjusting the structure of the protein based on the updated information on each residue.

In one embodiment, the update information associated with the folding structure may include update information generated by performing a cumulative operation on changes related to the residues of the protein predicted by the neural network model; or at least one of updated information based on the amount of change in the backbone torsion angle of each residue predicted by the neural network model.

In one embodiment, the updating of information on each residue may include obtaining a first change amount for a first residue of the protein predicted by the neural network model; acquiring an n-th change amount for the n-th residue of the protein predicted by the neural network model; and generating update information for the nth residue by performing an accumulation operation on the first through nth changes, where n may be a natural number greater than or equal to 2.

In one embodiment, updating information on each residue may further include updating rigid body motions of some residues of the protein.

In one embodiment, some of the residues of the protein include i-th to j-th residues of the protein, and the step of updating the rigid body motion of some of the residues of the protein comprises i+1-th to j-th residues of the protein. Updating the i+1 th change amount to the j th change amount with default values, wherein i is a natural number and j is a natural number greater than i.

In one embodiment, generating the update information for the n-th residue by performing an accumulation operation on the first to n-th change amounts may include determining a predetermined value for each of the first to n-th residues of the protein. performing scaling on each of the first through n-th changes based on a degree of flexibility; and generating update information for the n-th residue by performing an accumulation operation on the scaled first through n-th changes.

In one embodiment, the degree of flexibility determined in advance for each of the 1st residue to the nth residue of the protein is based on structural information analyzed by nuclear magnetic resonance spectroscopy (NMR), and a temperature factor obtained from an x-ray image. (temperature factor), or may be determined in advance based on the output of another neural network model.

In one embodiment, the updating of information on each residue may include obtaining a first backbone twist angle change amount for a first residue of the protein predicted by the neural network model; generating update information for the first residue based on the amount of change in the twist angle of the first backbone; obtaining an n-th backbone twist angle change for the n-th residue of the protein predicted by the neural network model; and generating update information for the n-th residue based on the amount of change in the n-th backbone twist angle, wherein n may be a natural number greater than or equal to 2.

In one embodiment, the generating of the update information for the first residue based on the change in the first backbone twist angle may include the first backbone twist based on a predetermined degree of flexibility of the first residue of the protein. scaling the amount of angular change; and generating update information for the first residue based on the amount of change in the first scaling backbone twist angle, wherein update information for the nth residue is generated based on the amount of change in the nth backbone twist angle. The generating step may include performing scaling on the change amount of the n-th backbone twist angle based on a degree of flexibility previously determined for the n-th residue of the protein; and generating update information for the n-th residue based on the amount of change in the n-th backbone twist angle at which the scaling is performed.

In one embodiment, adjusting the structure of the protein includes adjusting at least one of a pre-binding structure of the protein and a binding structure of the protein, and the neural network model predicts the pre-binding structure of the protein. neural network model; A neural network model predicting the binding structure of a plurality of proteins; Alternatively, at least one of a neural network model predicting a binding structure between a protein and a ligand may be included.

In addition, a structure of a neural network model for predicting a binding structure of a protein is disclosed to solve the above problems. The structure of the neural network model may include: a first neural network model generating structural information of the protein before binding based on the sequence of the protein; a second neural network model generating pre-combination structural information of the fusion target based on information on the fusion target; and a third neural network model that predicts a structure in which the protein and the binding target are combined based on the pre-binding structural information of the protein and the pre-binding structural information of the binding target, wherein the third neural network model is bent (twist) By utilizing update information associated with the structure, the structure in which the protein and the binding target are bound may be updated.

In one embodiment, the binding target is a ligand, and the third neural network model may update a structure in which the protein and the ligand are bound by utilizing update information associated with a fold structure of the protein.

In one embodiment, the binding target is another protein, the second neural network model is implemented with the same model as the first neural network model, and the third neural network model includes update information associated with the folding structure of the protein and the other A structure in which the protein and the other protein are combined may be updated by using update information associated with the protein's fold structure.

In addition, a method for adjusting the structure of a protein using a neural network model, which is performed by a computing device for solving the above problems, is disclosed. The method includes obtaining information about each residue of a protein; Using at least one of the update information based on the change in the backbone twist angle of each residue or the update information generated by performing an accumulation operation on the predicted change amounts of the protein residues, Updating information about residues; and adjusting the structure of the protein based on the updated information on each residue.

In addition, as an apparatus for solving the above problems, at least one processor; and a memory, wherein the at least one processor is configured to obtain information about each residue of a protein; updating information for each residue using update information associated with the folding structure; And it may be configured to adjust the structure of the protein based on the updated information on each residue.

In addition, as a computer program stored in a computer readable storage medium for solving the above problems, the computer program, when executed by at least one processor, causes the at least one processor to perform operations for predicting the structure of a protein and the operations include: obtaining information about each residue of the protein; updating information on each residue by utilizing update information associated with a folding structure; and adjusting the structure of the protein based on the updated information on each residue.

The present disclosure utilizes update information associated with a protein folding structure to update information on each residue, and predicts a protein structure based on this, to obtain a protein structure (e.g., a protein structure before binding or a linked structure of a protein). ) can improve the accuracy of the prediction.

In addition, the present disclosure may provide a technical solution capable of considering motion of a rigid body in a protein or structural change of a protein backbone unit in relation to protein structure prediction.

On the other hand, the effects of the present disclosure are not limited to the effects mentioned above, and various effects may be included within a range apparent to those skilled in the art from the contents to be described below.

1 is a block diagram of a computing device performing operations according to an embodiment of the present disclosure.
2 is a schematic diagram showing a neural network model according to an embodiment of the present disclosure.
Figure 3 is a schematic diagram for explaining a method of expressing residues of a protein according to an embodiment of the present disclosure.
4 is a flowchart illustrating a method for predicting a protein structure according to an embodiment of the present disclosure.
5 is a flowchart illustrating a method of cumulatively updating residues according to an embodiment of the present disclosure.
6 is a schematic diagram illustrating a method of cumulatively updating residues according to an embodiment of the present disclosure.
7 is a schematic diagram illustrating a method of predicting a protein-protein binding structure using a neural network model according to an embodiment of the present disclosure.
8 is a schematic diagram illustrating a method of predicting a binding structure between a protein and a ligand using a neural network model according to an embodiment of the present disclosure.
9 is a simplified and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Various embodiments are now described with reference to the drawings. In this disclosure, various descriptions are presented to provide an understanding of the disclosure. However, it is apparent that these embodiments may be practiced without these specific details.

The terms “component,” “module,” “system,” and the like, as used in this disclosure, refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or an execution of software. For example, a component may be, but is not limited to, a procedure, processor, object, thread of execution, program, and/or computer running on a processor. For example, both an application running on a computing device and a computing device may be components. One or more components may reside within a processor and/or thread of execution. A component can be localized within a single computer. A component may be distributed between two or more computers. Also, these components can execute from various computer readable media having various data structures stored thereon. Components may be connected, for example, via signals with one or more packets of data (e.g., data and/or signals from one component interacting with another component in a local system, distributed system) to other systems and over a network such as the Internet. data being transmitted) may communicate via local and/or remote processes.

In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or”. That is, unless otherwise specified or clear from the context, “X employs A or B” is intended to mean one of the natural inclusive substitutions. That is, X uses A; X uses B; Or, if X uses both A and B, "X uses either A or B" may apply to either of these cases. Also, the term “and/or” as used in this disclosure should be understood to refer to and include all possible combinations of one or more of the listed related items.

Also, the terms "comprises" and/or "comprising" should be understood to mean that the features and/or components are present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, elements, and/or groups thereof. Also, unless otherwise specified or where the context clearly indicates that a singular form is indicated, the singular in the disclosure and claims should generally be construed to mean "one or more".

In addition, the term “at least one of A or B” should be interpreted as meaning “when only A is included”, “when only B is included” and “when A and B are combined”.

Skilled artisans will further understand that the various illustrative logical blocks, components, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or both. It should be recognized that it can be implemented with combinations of To clearly illustrate the interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented in hardware or as software depends on the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of this disclosure.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure. The general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present disclosure is not limited to the embodiments presented herein. This disclosure is to be interpreted in the widest light consistent with the principles and novel features presented herein.

The configuration of the computing device 100 shown in FIG. 1 is only a simplified example. In one embodiment of the present disclosure, the computing device 100 may include other components for performing a computing environment of the computing device 100, and only some of the disclosed components may constitute the computing device 100.

The computing device 100 may include a processor 110 , a memory 130 , and a network unit 150 .

The processor 110 may include one or more cores, and includes a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU) of a computing device. unit), data analysis, and processors for deep learning. The processor 110 may read a computer program stored in the memory 130 and process data for machine learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 110 may perform an operation for learning a neural network. The processor 110 is used for neural network learning, such as processing input data for learning in deep learning (DL), extracting features from input data, calculating errors, and updating neural network weights using backpropagation. calculations can be performed. At least one of the CPU, GPGPU, and TPU of the processor 110 may process learning of the neural network model. For example, the CPU and GPGPU can process learning of neural network models and data classification using neural network models. In addition, in an embodiment of the present disclosure, the neural network model learning and data classification using the neural network model may be processed by using processors of a plurality of computing devices together. In addition, a computer program executed in a computing device according to an embodiment of the present disclosure may be a CPU, GPGPU or TPU executable program.

According to an embodiment of the present disclosure, the memory 130 may store any type of information generated or determined by the processor 110 and any type of information received by the network unit 150 .

According to an embodiment of the present disclosure, the memory 130 is a flash memory type, a hard disk type, a multimedia card micro type, or a card type memory (eg, SD or XD memory, etc.), RAM (Random Access Memory, RAM), SRAM (Static Random Access Memory), ROM (Read-Only Memory, ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory) -Only Memory), a magnetic memory, a magnetic disk, and an optical disk may include at least one type of storage medium. The computing device 100 may operate in relation to a web storage that performs a storage function of the memory 130 on the Internet. The above description of the memory is only an example, and the present disclosure is not limited thereto.

The network unit 150 according to an embodiment of the present disclosure includes a Public Switched Telephone Network (PSTN), x Digital Subscriber Line (xDSL), Rate Adaptive DSL (RADSL), Multi Rate DSL (MDSL), and VDSL ( Various wired communication systems such as Very High Speed DSL), Universal Asymmetric DSL (UADSL), High Bit Rate DSL (HDSL), and Local Area Network (LAN) may be used.

In addition, the network unit 150 presented in the present disclosure includes Code Division Multi Access (CDMA), Time Division Multi Access (TDMA), Frequency Division Multi Access (FDMA), Orthogonal Frequency Division Multi Access (OFDMA), SC-FDMA ( Single Carrier-FDMA) and other systems.

In the present disclosure, the network unit 150 may be configured regardless of its communication mode, such as wired and wireless, and may be configured with various communication networks such as a personal area network (PAN) and a wide area network (WAN). can In addition, the network may be the known World Wide Web (WWW), or may use a wireless transmission technology used for short-range communication, such as Infrared Data Association (IrDA) or Bluetooth. The techniques described in this disclosure may also be used in other networks mentioned above.

2 is a schematic diagram showing a neural network model according to an embodiment of the present disclosure.

Throughout this disclosure, neural network model and neural network may be used interchangeably. A neural network model may be composed of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network model includes one or more nodes. Nodes (or neurons) constituting neural network models may be interconnected by one or more links.

In the neural network model, one or more nodes connected through links may form a relative relationship of an input node and an output node. The concept of an input node and an output node is relative, and any node in an output node relationship with one node may have an input node relationship with another node, and vice versa. As described above, an input node to output node relationship may be created around a link. More than one output node can be connected to one input node through a link, and vice versa.

In a relationship between an input node and an output node connected through one link, the value of data of the output node may be determined based on data input to the input node. Here, a link interconnecting an input node and an output node may have a weight (at this time, parameters and weights may be used in the same meaning throughout the present disclosure). The weight may be variable, and may be changed by a user or an algorithm in order to perform a desired function of the neural network model. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. An output node value may be determined based on the weight.

As described above, in the neural network model, one or more nodes are interconnected through one or more links to form an input node and output node relationship in the neural network. The characteristics of the neural network model may be determined according to the number of nodes and links in the neural network, the relationship between the nodes and links, and the value of a weight assigned to each link. For example, when there are two neural network models having the same number of nodes and links and different weight values of the links, the two neural network models may be recognized as different from each other.

A neural network model may consist of a set of one or more nodes. A subset of nodes constituting a neural network may constitute a layer. Some of the nodes constituting the neural network model may constitute one layer based on distances from the first input node. For example, a set of nodes having a distance of n from the first input node may constitute n layers. The distance from the first input node may be defined by the minimum number of links that must be passed through to reach the corresponding node from the first input node. However, the definition of such a layer is arbitrary for explanation, and the order of a layer in a neural network may be defined in a method different from that described above. For example, a layer of nodes may be defined by a distance from a final output node.

An initial input node may refer to one or more nodes to which data is directly input without passing through a link in relation to other nodes among nodes in the neural network model. Alternatively, in a relationship between nodes based on a link in a neural network, it may mean nodes that do not have other input nodes connected by a link. Similarly, the final output node may refer to one or more nodes that do not have an output node in relation to other nodes among nodes in the neural network. Also, the hidden node may refer to nodes constituting the neural network model other than the first input node and the last output node.

The neural network model may be trained using at least one of supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Learning of the neural network model may be a process of applying knowledge for the neural network model to perform a specific operation to the neural network model.

The neural network model can be trained in a way that minimizes errors in output. In the learning of the neural network model, the learning data is repeatedly input into the neural network model, the output of the neural network model for the training data and the error of the target are calculated, and the error of the neural network model is transferred from the output layer of the neural network model to the input layer in the direction of reducing the error. This is the process of updating the weight of each node of the neural network model by backpropagating in the same direction. In the case of teacher learning, the learning data in which the correct answer is labeled is used for each learning data (ie, the labeled learning data), and in the case of comparative teacher learning, the correct answer may not be labeled in each learning data. That is, for example, learning data in the case of teacher learning regarding data classification may be data in which each learning data is labeled with a category. Labeled training data is input to a neural network model, and an error may be calculated by comparing an output (category) of the neural network model with a label of the training data. As another example, in the case of comparative history learning for data classification, an error may be calculated by comparing input learning data with a neural network model output. The calculated error is back-propagated in the neural network model in the reverse direction (ie, from the output layer to the input layer), and the connection weight of each node in each layer of the neural network model may be updated according to the back-propagation. The amount of change in the connection weight of each updated node may be determined according to a learning rate. The computation of the neural network model on the input data and the backpropagation of errors may constitute an epoch. A learning rate may be applied differently according to the number of repetitions of epochs of the neural network model. For example, a high learning rate is used in the early stages of neural network model training so that the neural network model can quickly achieve a certain level of performance to increase efficiency, and a low learning rate can be used in the late stage to increase accuracy.

In learning neural network models, generally, training data can be a subset of real data (ie, data to be processed using the trained neural network model), and therefore, errors for training data are reduced, but errors for real data are reduced. There may be increasing epochs. Over-fitting is a phenomenon in which errors for actual data increase due to excessive learning on training data. For example, a phenomenon in which a neural network model that learns a cat by showing a yellow cat does not recognize that it is a cat when it sees a cat other than yellow may be a type of overfitting. Overfitting can act as a cause of increasing the error of machine learning algorithms. Various optimization methods can be used to prevent such overfitting. To prevent overfitting, methods such as increasing training data, regularization, inactivating some nodes in the network during learning, dropout, and batch normalization can be applied. .

According to an embodiment of the present disclosure, a computer readable medium storing a data structure is disclosed.

Data structure can refer to the organization, management, and storage of data that enables efficient access and modification of data. Data structure may refer to the organization of data to solve a specific problem (eg, data retrieval, data storage, data modification in the shortest time). A data structure may be defined as a physical or logical relationship between data elements designed to support a specific data processing function. A logical relationship between data elements may include a connection relationship between user-defined data elements. A physical relationship between data elements may include an actual relationship between data elements physically stored in a computer-readable storage medium (eg, a persistent storage device). The data structure may specifically include a set of data, a relationship between data, and a function or command applicable to the data. Through an effectively designed data structure, a computing device can perform calculations while using minimal resources of the computing device. Specifically, the computing device can increase the efficiency of operation, reading, insertion, deletion, comparison, exchange, and search through an effectively designed data structure.

The data structure can be divided into a linear data structure and a non-linear data structure according to the shape of the data structure. A linear data structure may be a structure in which only one data is connected after one data. Linear data structures may include lists, stacks, queues, and decks. A list may refer to a series of data sets in which order exists internally. The list may include a linked list. A linked list may be a data structure in which data are connected in such a way that each data is connected in a single line with a pointer. In a linked list, a pointer can contain information about connection to the next or previous data. A linked list can be expressed as a singly linked list, a doubly linked list, or a circular linked list depending on the form. A stack can be a data enumeration structure that allows limited access to data. A stack can be a linear data structure in which data can be processed (eg, inserted or deleted) at only one end of the data structure. The data stored in the stack may be a LIFO-Last in First Out (Last in First Out) data structure. A queue is a data listing structure that allows limited access to data, and unlike a stack, it can be a data structure (FIFO-First in First Out) in which data stored later comes out later. A deck can be a data structure that can handle data from either end of the data structure.

The nonlinear data structure may be a structure in which a plurality of data are connected after one data. The non-linear data structure may include a graph data structure. A graph data structure can be defined as a vertex and an edge, and an edge can include a line connecting two different vertices. The graph data structure may include a tree data structure. The tree data structure may be a data structure in which one path connects two different vertices among a plurality of vertices included in the tree. That is, it may be a data structure that does not form a loop in a graph data structure.

The data structure may include a neural network. And the data structure including the neural network may be stored in a computer readable medium. The data structure including the neural network includes preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation function associated with each node or layer of the neural network, and A loss function for learning may be included. A data structure including a neural network may include any of the components described above. That is, the data structure including the neural network includes preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data acquired from the neural network, activation function associated with each node or layer of the neural network, and neural network. It may be configured to include all or any combination thereof, such as a loss function for learning of . In addition to the foregoing configurations, the data structure comprising the neural network may include any other information that determines the characteristics of the neural network. In addition, the data structure may include all types of data used or generated in the computational process of the neural network, but is not limited to the above. A computer readable medium may include a computer readable recording medium and/or a computer readable transmission medium. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network includes one or more nodes.

The data structure may include data input to the neural network. A data structure including data input to the neural network may be stored in a computer readable medium. Data input to the neural network may include training data input during a neural network learning process and/or input data input to a neural network that has been trained. Data input to the neural network may include pre-processed data and/or data subject to pre-processing. Pre-processing may include a data processing process for inputting data to a neural network. Accordingly, the data structure may include data subject to pre-processing and data generated by pre-processing. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure may include the weights of the neural network. (In this disclosure, weights and parameters may be used in the same meaning.) Also, a data structure including weights of a neural network may be stored in a computer readable medium. A neural network may include a plurality of weights. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. A data value output from an output node may be determined based on the weight. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

As a non-limiting example, the weights may include weights that are varied during neural network training and/or weights for which neural network training has been completed. The variable weight in the neural network learning process may include a weight at the start of an epoch and/or a variable weight during an epoch. The weights for which neural network learning is completed may include weights for which epochs are completed. Accordingly, the data structure including the weights of the neural network may include a data structure including weights that are variable during the neural network learning process and/or weights for which neural network learning is completed. Therefore, it is assumed that the above-described weights and/or combinations of weights are included in the data structure including the weights of the neural network. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure including the weights of the neural network may be stored in a computer readable storage medium (eg, a memory or a hard disk) after going through a serialization process. Serialization can be the process of converting a data structure into a form that can be stored on the same or another computing device and later reconstructed and used. The computing device may transmit and receive data through a network by serializing the data structure. The data structure including the weights of the serialized neural network may be reconstructed on the same computing device or another computing device through deserialization. The data structure including the weights of the neural network is not limited to serialization. Furthermore, the data structure including the weights of the neural network is a data structure for increasing the efficiency of operation while minimizing the resource of the computing device (for example, B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree). The foregoing is only an example, and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of the neural network. Also, the data structure including the hyperparameters of the neural network may be stored in a computer readable medium. A hyperparameter may be a variable variable by a user. Hyperparameters include, for example, learning rate, cost function, number of epoch iterations, weight initialization (eg, setting the range of weight values to be targeted for weight initialization), It may include the number of Hidden Units (eg, the number of hidden layers and the number of nodes of the hidden layer). The foregoing data structure is only an example, and the present disclosure is not limited thereto.

Prior to a description of an embodiment of the present disclosure, a description of terms used throughout the present disclosure will be given with reference to FIG. 3 .

The term “residue” as used in this disclosure refers to a residue of an amino acid constituting the structure of a protein. Here, the residue of an amino acid may mean, for example, a portion other than a portion removed when an amino acid forms a peptide bond (eg, a group other than an atomic group removed when forming a peptide bond).

In addition, with regard to the characteristics of the residue, each residue may include structural information. For example, each residue may include [location information, direction information]. In this case, the position information and direction information of each residue may be associated with geometric information within the protein structure of each residue. Meanwhile, the [location information, direction information] of each residue may preferably be expressed in the form of a vector including coordinate information and direction information, but is not limited thereto and may be expressed in various ways. According to an embodiment of the present disclosure, characteristics of a residue may be expressed by a structure representation method of a residue gas method.

을 포함할 수 있다. 여기서,

은 각각의 잔기의 구조 정보를 나타낼 수 있다. 예를 들어,

은 각각의 잔기의 위치 정보, 방향 정보, 그 외의 기하학적인 정보 등을 나타낼 수 있다. 또한, 상기 단백질과 연관된 시퀀스(sequence)가 프로세서(110)에 의해 입력되는 순서를 기초로 잔기의 순서를 결정할 수 있다. 또한, 잔기의 순서는 단백질 전체가 아닌 단백질의 일부분을 기초로 결정될 수 있다. 그러나, 단백질의 순서를 결정하는 것은 이에 한정되지 않고, 통상의 기술자의 필요에 따라 다양하게 결정될 수 있다. (참고로, 본 개시에서는 각각의 잔기를 위치와 방향을 포함하는 특성을 고려하여 도 3과 같이 화살표 도형으로 표현된다.)As mentioned above, residues can be classified by the sequential order of residues connected to each other. For example, assuming that the number of residues included in a protein is N, the N residues are ordered from the C-terminus to the N-terminus of the protein structure or in the opposite direction. A residue (T), identified as

can include here,

may represent structural information of each residue. for example,

may represent position information, direction information, and other geometric information of each residue. In addition, the order of residues may be determined based on the order in which sequences associated with the protein are input by the processor 110 . Also, the sequence of residues may be determined based on parts of the protein rather than the entire protein. However, determining the order of proteins is not limited thereto, and may be determined in various ways according to the needs of a person skilled in the art. (For reference, in the present disclosure, each residue is represented by an arrow shape as shown in FIG. 3 in consideration of characteristics including position and direction.)

은 신경망 모델을 포함하는 다양한 예측 모델을 사용하여 예측되는 잔기에 대한 변화량으로, 다시 말해 n 번째 잔기(

)에 대한 변화량을 의미할 수 있다. 여기서 상기 변화량은 구조 정보와 관련된 변화량을 포함할 수 있으며, 예를 들어, 위치 정보의 변화량, 방향 정보의 변화량, 그 외의 기하학적인 변화량 등을 포함할 수 있다. 본 개시의 일 실시예에서, 각각의 잔기(

)에 대한 변화량은, 각각 제 1 변화량 내지 제 n 변화량(

)으로 표현될 수 있다. 한편, 상기 변화량은 단백질의 잔기들에 관한 정보에 기초하여 신경망 모델을 통해 예측될 수 있다.Throughout this disclosure,

is the amount of change for a residue predicted using various prediction models including neural network models, that is, the n-th residue (

) can mean the amount of change for Here, the change amount may include a change amount related to structural information, and may include, for example, a change amount of location information, a change amount of direction information, and other geometric changes. In one embodiment of the present disclosure, each residue (

The amount of change for ) is the first change amount to the nth change amount (

) can be expressed as Meanwhile, the amount of change may be predicted through a neural network model based on information about protein residues.

'에 기초하여 표현될 수 있다. 여기서 연산자 '

'는 아다마르 곱(Hadamard product) 연산자이거나 또는 유클리드 변환과 연관된 연산자일 수 있다. 또한, 상기 업데이트는 이러한 연산자 이외에도 다양한 연산자에 기초하여 구현될 수도 있다. 한편, 업데이트된 n 번째 잔기는

으로 표현될 수 있다.예컨대,

을 기초로 업데이트된 n 번째 잔기(

)는

일 수 있다. (즉,

) 일 실시예에서,

은 [

]와 같이 업데이트 된 것일 수 있다. 그러나, 이는 예시일 뿐, 본 개시의 일 실시예에 따른 업데이트의 방법은 이에 한정되지 않고, 추후, 이와 관련된 다른 일 실시예들이 개시된다. Updates throughout this disclosure include operator '

It can be expressed based on '. where operator '

' can be a Hadamard product operator or an operator associated with Euclidean transformation. In addition, the update may be implemented based on various operators other than these operators. On the other hand, the updated n-th residue is

It can be expressed as. For example,

The nth residue updated based on (

)Is

can be (in other words,

) In one embodiment,

silver [

]. However, this is only an example, and the update method according to an embodiment of the present disclosure is not limited thereto, and other embodiments related thereto will be disclosed later.

의 잔기 사슬에서

가 꺾이는 잔기에 해당하는 경우,

지점에서의 꺾임에 따라

의 잔기들은 함께 움직이는 강체 운동이 강제 되는데), 상기 "꺾임 구조와 연관된 업데이트"는 이러한 꺾임 구조와 연관된 구조 정보 또는 구조 정보의 변화를 표현할 수 있는 업데이트 방식을 의미할 수 있다. 또한, "꺾임 구조와 연관된 업데이트"는, 단백질에 포함된 꺾임 구조를 모사할 수 있는 다양한 업데이트 방식들이 포함될 수도 있다. 특히, 본 개시의 실시예들은, 도 3의 310과 같은 "그래프 형태의 표현 방식(예를 들어, 도 3의 310과 같은 ball & stick 그래프 형태의 표현 방식)"이 아닌, "잔기 가스(residue gas)의 표현 방식(예를 들어, 도 3의 300과 같은 개별 잔기의 구조 정보를 고려하는 표현 방식)"과 연관하여 상기 "꺾임 구조와 연관된 업데이트"를 구현할 수 있다. 일 실시예에서, 상기 "꺾임 구조와 연관된 업데이트"는, 누적 연산 방식을 활용하여, 상기 잔기 가스의 표현법에서 구현될 수 있다. 다른 실시예에서, 상기 "꺾임 구조와 연관된 업데이트"는, 신경망 모델(즉, 상기 잔기 가스의 표현 방식에 기초하는 신경망 모델)의 출력이 "각각의 잔기에서의 백본 비틀림 각도의 변화량"이 되도록 함으로써, 상기 잔기 가스의 표현법에서 구현될 수 있다.Throughout the present disclosure, “update associated with a folding structure” may refer to an update method capable of reflecting a folding structure itself or a change based on a folding structure during an update process. For example, when amino acid residues form a bend structure, the residue at the bend site undergoes a relatively large structural change, and the residues connected to the bend site undergo rigid body motion (e.g.,

residues in the chain of

If corresponds to the folding residue,

according to the bend at the point

Residues of are forced to move together), and the "update associated with the bending structure" may mean structural information associated with such a bending structure or an update method capable of expressing a change in structural information. In addition, "update associated with a folding structure" may include various update methods capable of mimicking a folding structure included in a protein. In particular, the embodiments of the present disclosure are not "representation method in the form of a graph (eg, expression method in the form of a ball & stick graph such as 310 in FIG. 3)" such as 310 in FIG. gas) (for example, an expression method that considers structural information of individual residues, such as 300 in FIG. In one embodiment, the "update associated with the folding structure" may be implemented in the residual gas representation by utilizing an accumulation operation scheme. In another embodiment, the "update associated with the folding structure" is performed such that the output of the neural network model (ie, the neural network model based on the representation scheme of the residual gas) is "the amount of change in the twist angle of the backbone at each residue". , can be implemented in the expression of the residual gas.

From now on, with reference to steps S400 to S420 of FIG. 4 , a general process for predicting the structure of a protein will be described.

Referring to Figure 4, the processor 110 may perform steps S400 to S420 to predict the structure of the protein. Specifically, the processor 110 includes "acquiring information on each residue of a protein (S400)" and "updating information on each residue by utilizing update information associated with a fold structure (S410). )" and "adjusting the structure of the protein based on the updated information on each residue (S420)" may be performed.

The step S400 is a step in which the processor 110 obtains information on each residue of the protein. In one embodiment, the processor 110 may obtain structural information for each residue of the protein. For example, the processor 110 may obtain position information and direction information of each residue of the protein. Additionally, the processor 110 may obtain other information related to the structural characteristics of each residue of the protein, in addition to these information.

의 잔기 사슬에서

가 꺾이는 잔기에 해당하는 경우,

지점에서의 꺾임에 따라

의 잔기들은 함께 움직이는 강제 운동을 하게 되는데), 상기 "꺾임 구조와 연관된 업데이트"는 이러한 꺾임 구조와 연관된 구조 정보 또는 구조 정보의 변화를 표현할 수 있는 업데이트 방식을 의미할 수 있다. 또한, "꺾임 구조와 연관된 업데이트"는, 단백질에 포함된 꺾임 구조를 모사할 수 있는 다양한 업데이트 방식들이 포함될 수도 있다.The step S410 is a step in which the processor 110 updates information on each residue by utilizing update information associated with the folding structure. Here, "update related to the folding structure", as mentioned above, may mean an update method capable of reflecting the folding structure itself or a change based on the folding structure during the update process. For example, when amino acid residues form a bend structure, the residue at the bend site undergoes a relatively large structural change, and the residues connected to the bend site undergo rigid body motion (e.g.,

residues in the chain of

If corresponds to the folding residue,

according to the bend at the point

Residues of are forced to move together), and the “update related to the bending structure” may mean structural information associated with such a bending structure or an update method capable of expressing a change in structural information. In addition, "update associated with a folding structure" may include various update methods capable of mimicking a folding structure included in a protein.

In one embodiment, the "update information associated with the fold structure" is "update information generated by performing an accumulation operation on changes related to the residues of the protein predicted by the neural network model" (Example-A related) may be included. In addition, in another embodiment, the "update information associated with the folding structure" is "update information based on the amount of change in backbone twist angle of each residue predicted by the neural network model" (related to Example-B) can include On the other hand, in another embodiment, the "update information associated with the bending structure" may include both "update information generated by performing the accumulation operation" and "update information based on the change amount of the backbone twist angle". Also, through this, an ensemble-type update method may be implemented.

)에 대한 제 1 변화량(

), 상기 단백질의 2번째 잔기(

)에 대한 제 2 변화량(

), 상기 단백질의 n-1번째 잔기(

)에 대한 제 n-1 변화량(

), 상기 단백질의 n번째 잔기(

)에 대한 제 n 변화량(

) 을 획득하는 단계(S410-A1)", ②"상기 제 1 변화량(

) 내지 상기 제 n 변화량(

)에 대하여 누적 연산(

)을 수행하여 상기 n 번째 잔기에 대한 업데이트 정보(

=

)를 생성하는 단계(S410-A2)", ③"누적 연산에 기초하여 생성된 각각의 잔기에 대한 업데이트 정보를 활용하여, 상기 각각의 잔기에 대한 정보를 업데이트하는 단계(S410-A3)" 등을 수행할 수 있다. 이러한 경우, 강체 운동으로 변화하는 잔기들에 대해서는 변화량들이 디폴트(default) 값(예컨대, 변화량이 없음을 나타내는 0, identity map, 영 벡터 등)으로 산출될 수 있으므로, 이러한 강체 운동을 표현하기 위해 업데이트 정보, 예측 정보, 결과 정보 등이 만족해야 하는 제약 조건이 간단해질 수 있다.Regarding the above 'Example-A', the processor 110 ① "predicted by the neural network model, the first residue of the protein (

) to the first amount of change (

), the second residue of the protein (

) to the second amount of change (

), the n-1 residue of the protein (

) The n-1th change amount for (

), the nth residue of the protein (

) The nth change amount for (

) Obtaining (S410-A1) ", ②" the first change amount (

) to the nth change amount (

) for the cumulative operation (

) to update information for the n-th residue (

=

) (S410-A2)”, ③ “Utilizing update information on each residue generated based on the accumulation operation, updating information on each residue (S410-A3)”, etc. In this case, since changes can be calculated with default values (e.g., 0 indicating no change, identity map, zero vector, etc.) for residues that change with rigid body motion, such a rigid body Constraints that update information, prediction information, result information, etc. must satisfy in order to express an exercise can be simplified.

) 내지 n번째 잔기(

) 각각에 대하여 미리 결정된 유연성 정도에 기초하여, 상기 제 1 변화량(

) 내지 상기 n 변화량(

) 각각에 대하여 스케일링(scaling)을 수행하는 단계(S410-A2-1)" 및 ②"상기 스케일링이 수행된 제 1 변화량(

) 내지 제 n 변화량(

)에 대하여 누적연산(

)을 수행하여 상기 n번째 잔기에 대한 업데이트 정보(

=

)를 생성하는 단계(S410-A2-2)"를 하위단계로 포함하여 수행할 수 있다. 이때, 미리 결정된 유연성 정도는, 핵자기 공명 분광법(NMR, nuclear magnetic resonance spectroscopy)에 의해 분석된 구조 정보, 엑스레이(x-ray) 이미지에서 획득된 온도인자(temperature factor), 또는 다른 신경망 모델(예컨대, 결합간 유연성이 측정되는 모델)의 출력에 기초하여 미리 결정될 수 있다.In addition, in relation to the S410-A2, the processor 110, ① "the first residue of the protein (

) to the nth residue (

) Based on a degree of flexibility predetermined for each, the first change amount (

) to the n change amount (

) Performing scaling for each (S410-A2-1)" and ②"The first change amount for which the scaling is performed (S410-A2-1)"

) to the nth change amount (

) for cumulative operation (

) to update information on the n-th residue (

=

) generating step (S410-A2-2)" as a sub-step. At this time, the predetermined degree of flexibility is determined by structural information analyzed by nuclear magnetic resonance spectroscopy (NMR). , a temperature factor obtained from an x-ray image, or an output of another neural network model (eg, a model in which flexibility between couplings is measured).

) 내지 'j'번째 잔기(

)(상기 i는 자연수이고, 상기 j는 상기 i보다 큰 자연수일 수 있음)를 포함하는 경우, "상기 단백질의 'i+1'번째 잔기(

) 내지 'j'번째 잔기(

)(즉,

)에 대한 제 i+1 변화량(

) 내지 제 j 변화량(

)(즉,

)이 각각 디폴트(default) 값(예컨대, 변화량이 없음을 나타내는 0, identity map, 영 벡터 등)으로 산출되는 비교적 간단한 연산에 기초하여, 상기 단백질의 일부 잔기들의 강체 운동과 관련된 업데이트가 수행될 수 있다. 달리 말해, 상기 제 i+1 변화량(

) 내지 제 j 변화량(

)(즉,

)이 각각 디폴트 값으로 산출되고, 이에 따라, 누적 연산 기반의 i번째 잔기에 대한 업데이트 정보(

=

) 내지 j번째 잔기에 대한 업데이트 정보(

=

)가 모두 동일한 값으로 산출되게 되는, 비교적 간단한 연산을 통해, 상기 단백질의 'i'번째 잔기(

) 내지 'j'번째 잔기(

)의 강체 운동과 관련된 업데이트가 수행될 수 있다. 한편, i번째 잔기(

)에 대한 제 i 변화량(

)이 디폴트 값이 아닌 값으로 산출되는 경우에는, 업데이트 과정 중에 i번째 잔기(

)에서 강체 단위의 꺾임이 구현될 수 있다. 다시 말해, i번째 잔기(

) 내지 'j'번째 잔기(

)(즉,

)가 강체를 형성하되 i번째 잔기(

)에서 강체의 꺾임이 구현되는 구조로 업데이트될 수 있다. 한편, 이와 같은 방식에서는, 강체 운동과 연관되는 잔기들의 변화량들이 단순히 디폴트 값으로 산출되는 비교적 간단한 연산에 의해 강체 운동이 업데이트될 수 있으므로, 강체 운동을 구현하기 위한 신경망 모델의 연산량이 현저하게 감소될 수 있다.In addition, in relation to the step S410-A3, the update operation can be simplified even when some residues of the protein perform rigid motion (lump motion). For example, some residues of the protein undergoing rigid motion are the 'i'th residue of the protein (

) to 'j'th residue (

) (wherein i is a natural number and j may be a natural number greater than i), "the 'i+1'th residue of the protein (

) to 'j'th residue (

)(in other words,

) the i + 1 th change (

) to the jth change amount (

)(in other words,

Based on a relatively simple operation in which ) is calculated as a default value (eg, 0 indicating no change, identity map, zero vector, etc.), updates related to rigid motion of some residues of the protein can be performed. there is. In other words, the i + 1 change amount (

) to the jth change amount (

)(in other words,

) is calculated as a default value, and accordingly, the update information for the i-th residue based on the cumulative operation (

=

) to update information for the j-th residue (

=

) is calculated as the same value, through a relatively simple operation, the 'i'th residue of the protein (

) to 'j'th residue (

) can be performed. On the other hand, the i-th residue (

) the i th change amount for (

) is calculated as a non-default value, during the update process, the i-th residue (

), bending in units of rigid bodies can be implemented. In other words, the ith residue (

) to 'j'th residue (

)(in other words,

) forms a rigid body, but the ith residue (

) can be updated to a structure in which bending of a rigid body is implemented. On the other hand, in this method, since the rigid body motion can be updated by a relatively simple operation in which changes in residues associated with the rigid body motion are simply calculated as default values, the amount of computation of the neural network model for implementing the rigid body motion can be significantly reduced. can

)에 대한 제 1 백본 비틀림 각도 변화량(

)을 획득하는 단계(S410-B1)", ②"상기 제 1 백본 비틀림 각도 변화량(

)에 기초하여 상기 1번째 잔기(

)에 대한 업데이트 정보(

)를 생성하는 단계(S410-B2)", ③"상기 신경망 모델에 의해 예측된 상기 단백질의 n번째 잔기(

)에 대한 제 n 백본 비틀림 각도 변화량(

)을 획득하는 단계(S410-B3)" 및 ④"상기 제 n 백본 비틀림 각도 변화량(

)에 기초하여 상기 n번째 잔기(

)에 대한 업데이트 정보(

)를 생성하는 단계(S410-B4)"를 수행할 수 있다.(이때, 상기 n은 2이상의 자연수로 가정한다.) 한편, 백본 비틀림 각도 변화량에 기초하여 잔기에 대한 구조 업데이트 정보를 생성하는 것은, 삼각함수에 기초하여 연산 될 수 있다.Next, in relation to the 'Example-B', the processor 110, as a sub-step of the step S410, ① "the first residue of the protein predicted by the neural network model (

) to the first backbone twist angle change amount (

) Obtaining (S410-B1) ", ②" The first backbone twist angle change amount (

) Based on the first residue (

) for update information (

) Generating (S410-B2) ", ③" The n-th residue of the protein predicted by the neural network model (

) The amount of change in the nth backbone twist angle (

) Obtaining (S410-B3) "and ④" the n-th backbone twist angle change (

) Based on the nth residue (

) for update information (

) may be performed (S410-B4). (At this time, it is assumed that n is a natural number of 2 or more.) On the other hand, generating structure update information for residues based on the amount of change in backbone twist angle may be performed. , can be calculated based on trigonometric functions.

)에 대하여 스케일링(

)을 수행하는 단계(S410-B2-1)" 및 ②"상기 스케일링이 수행된 제 1 백본 비틀림 각도 변화량(

)에 기초하여 상기 1번째 잔기에 대한 업데이트 정보(

)를 생성하는 단계(S410-B2-2)"를 수행할 수 있다. In addition, in connection with the step S410-B2, the processor 110, ① based on the degree of flexibility predetermined for the first residue of the protein, the amount of change in the first backbone twist angle (

) for scaling (

) Performing (S410-B2-1)" and ②"The amount of change in the first backbone twist angle at which the scaling is performed (

) Based on the update information for the first residue (

).

)에 대하여 스케일링(

)을 수행하는 단계(S410-B4-1)" 및 ②"상기 스케일링이 수행된 제 n 백본 비틀림 각도 변화량(

)에 기초하여 상기 n번째 잔기에 대한 업데이트 정보(

)를 생성하는 단계(S410-B4-2)"를 수행할 수 있다.In addition, in connection with the steps S410-B4, the processor 110 may: (1) change the n-th backbone twist angle based on a predetermined degree of flexibility for the n-th residue of the protein (

) for scaling (

) Performing (S410-B4-1)" and ②"The amount of change in the twist angle of the nth backbone where the scaling is performed (

) Based on the update information for the nth residue (

).

The step S420 is a step of adjusting the structure of the protein based on the updated information on each residue.

In relation to the step S420, the processor 110 may perform "step of adjusting at least one of the pre-binding structure of the protein and the binding structure of the protein using the neural network model (S420-1)". there is. In addition, the neural network model is a "neural network model predicting the pre-binding structure of a protein (ie, a neural network model predicting a single protein structure)", a "neural network model predicting the binding structure of a plurality of proteins", or a "protein and ligand It may include a "neural network model that predicts the binding structure of". (The neural network model predicting the combined structures will be described in detail later with reference to FIGS. 7 and 8.)

Now, with reference to FIGS. 5 and 6, an example of how the processor 110 performs an update for each residue in a method utilizing the accumulation operation (of Embodiment-A) will be described.

,

,

)에 대한 업데이트 정보(

)를 생성한다고 가정할 때, 프로세서(110)는 상기 1번째 잔기 내지 3번째 잔기(

,

,

)를 기초로 신경망 모델을 사용하여 제 1 변화량 내지 제 3 변화량(

)을 생성할 수 있다. 이어서, 프로세서(110)는 상기 제 1 변화량 내지 제 3 변화량(

)을 기초로 ①상기 제 1 변화량(

)이 고려된 1 번째 잔기(

)에 대한 업데이트 정보(

), ②상기 제 1 변화량(

) 및 제 2 변화량(

)이 누적적으로 고려된 2 번째 잔기(

)에 대한 업데이트 정보(

), 및 ③상기 제 1 변화량(

), 제 2 변화량(

) 및 제 3 변화량(

)이 누적적으로 고려된 3 번째 잔기(

)에 대한 업데이트 정보(

)를 생성할 수 있다. 이어서 프로세서(110)는, 상기 1 번째 잔기에 대한 업데이트 정보(

)를 활용하여 상기 1번째 잔기의 구조 정보를 업데이트하고(

), 상기 2 번째 잔기에 대한 업데이트 정보(

)를 활용하여 상기 2번째 잔기의 구조 정보를 업데이트하고(

), 상기 3 번째 잔기에 대한 업데이트 정보(

)를 활용하여 상기 3번째 잔기의 구조 정보를 업데이트(

)할 수 있다.First, the processor 110 uses the cumulative update method to determine the first to third residues (

,

,

) for update information (

), the processor 110 determines the first to third residues (

,

,

) Based on the neural network model, the first change amount to the third change amount (

) can be created. Subsequently, the processor 110 performs the first to third changes (

) Based on ① the first change amount (

) is considered the first residue (

) for update information (

), ② the first change amount (

) and the second amount of change (

) is the cumulatively considered second residue (

) for update information (

), and ③ the first change amount (

), the second amount of change (

) and the third amount of change (

) is the cumulatively considered third residue (

) for update information (

) can be created. Subsequently, the processor 110 provides update information for the first residue (

) to update the structural information of the first residue (

), update information for the second residue (

) to update the structural information of the second residue (

), update information for the third residue (

) to update the structural information of the third residue (

)can do.

)은, 상기 1 번째 내지 3 번째 잔기에 대한 업데이트 정보(

)에 모두 고려되며, 업데이트된 상기 1 번째 내지 3 번째 잔기(

) 모두에 영향을 줄 수 있다. (S500 단계 참조), 또한, 상기 2 번째 잔기에 대한 변화량(

)은, 상기 2 번째 내지 3 번째 잔기에 대한 업데이트 정보(

)에 고려되며, 업데이트된 상기 2 번째 내지 3 번째 잔기(

)에 영향을 줄 수 있다. (S510 단계 참조) 또한, 상기 3 번째 잔기에 대한 변화량(

)은, 상기 3 번째 잔기에 대한 업데이트 정보(

)에 고려되며, 상기 업데이트된 3 번째 잔기(

)에 영향을 줄 수 있다.According to this cumulative update method, at this time, the amount of change for the first residue (

) is the update information for the first to third residues (

), and the updated 1st to 3rd residues (

) can affect both. (Refer to step S500), and also, the amount of change for the second residue (

) is the update information for the second to third residues (

), and the updated second to third residues (

) can affect (Refer to step S510) In addition, the amount of change for the third residue (

) is the update information for the third residue (

), and the updated third residue (

) can affect

,

,

)가 강체 운동을 하는 경우, 프로세서(110)는, 상기 제 2 변화량(

) 및 상기 제 3 변화량(

)을 디폴트 값으로 설정하여 상기 강체 운동을 쉽게 표현하거나 강제할 수 있다. 따라서, 이러한 누적 업데이트 방식은, "강체 운동을 하는 잔기들에 동일한 업데이트를 강제하는 방식"과 비교하여, 강제 운동을 강제하기 위한 네트워크의 추가 제약 조건을 단순화시킬 수 있고, 학습을 상대적으로 단순화시킬 수 있으며, 업데이트와 관련된 연산의 효율성을 현저하게 향상시킬 수 있다. 추가로, 프로세서(110)는, 상기 1번째 잔기(

)에 대한 상기 제 1 변화량(

)을 조정하여 상기 강체 운동과 연관된 꺾임을 쉽게 조정하거나 업데이트할 수도 있다.On the other hand, if this cumulative update method is used, the rigid body motion of the residues can be easily expressed and enforced. For example, as shown in FIG. 6, the first to third residues (

,

,

) is a rigid body motion, the processor 110, the second change amount (

) and the third change amount (

) as a default value, the rigid body motion can be easily expressed or forced. Therefore, this cumulative update method can simplify the additional constraints of the network for forcing the forced motion compared to the "method of forcing the same update to residues with rigid motion", and will relatively simplify learning. and can significantly improve the efficiency of calculations related to update. In addition, the processor 110, the first residue (

) to the first change amount (

) to easily adjust or update the bending associated with the rigid body motion.

,

,

)에 대한 업데이트 정보(

)는 각각 ((

), (

), (

))에 기초하여 생성될 수 있다.Additionally, as mentioned above, the processor 110 may additionally consider scaling information S based on a degree of flexibility in the process of generating each update information. In this case, the first to third residues (

,

,

) for update information (

) are respectively ((

), (

), (

))).

Hereinafter, a method for predicting a binding structure in which a protein or a ligand is bound to a protein using a neural network model is disclosed, according to an embodiment of the present disclosure, with reference to FIGS. 7 and 8 . (That is, methods for predicting protein-protein binding structures or protein-ligand binding structures are disclosed respectively.)

First, as an example of protein-protein binding structure prediction, referring to FIG. 7 , the processor 110 generates a first neural network model (or protein structure) based on the first protein sequence (700, or protein structure) to be bound. 710) can be used to predict the structure of the first unbound protein (720, unbound protein structure). In this case, the predicted structure 720 of the first unbound protein may include position information and direction information of residues included in the protein. In addition, when the processor 110 predicts the structure 720 of the first unbound protein using the first neural network model 710, the first neural network model output from the hidden layer (hidden layer) of the first unbound protein A hidden vector 730 (hidden vector) may be obtained. Subsequently, the processor 110 may predict the structure 721 (unbound protein structure) of the second unbound protein using the second neural network model 711 based on the second protein sequence 701 to be bound. . In this case, the predicted structure 721 of the second unbound protein may include position information and direction information of residues. In addition, when the processor 110 predicts the structure 721 of the second unbound protein using the second neural network model 711, the second hidden vector 731 output from the hidden layer of the second neural network model ) can be obtained. Subsequently, the processor displays "structure 720 of the first unbound protein, the first hidden vector 730 extracted from the first neural network model 710" and "structure 721 of the second unbound protein, A coupling structure may be predicted using the third neural network model 740 based on the second hidden vector 731 extracted from the second neural network model 711 . In one embodiment, the first neural network model 710 and the second neural network model 711 may be the same neural network model that each performs the task of predicting a protein structure from a protein sequence, but is not limited thereto. Different models may be used. In addition, the third neural network model 740 uses the method of Example-A or Example-B according to the above-mentioned embodiment of the present disclosure to convert the protein residues included in the binding structure 750 one or more times. can be updated By using Example-A or Example-B, it can be effective in that it can reflect everything from changes in side chains included in protein residues to changes in the backbone.

As an additional embodiment, the third neural network model 740 is a protein residue included in the binding structure 750 using the method of Example-A or Example-B according to an embodiment of the present disclosure mentioned above. can be updated by repeating them multiple times. For example, the processor 110 generates update information for each residue using the third neural network model 740 again based on the method of embodiment-A or embodiment-B, and based on the update information The protein structure (ie, each residue) included in the binding structure 750 may be repeatedly adjusted. That is, the processor 110 may fine-tune the coupling structure 750 by repeating the method of the previous example a plurality of times.

Next, as an example of protein-ligand binding structure prediction, referring to FIG. 8 , the processor 110 uses a first neural network model 810 based on the sequence 800 of a protein to be bound. Thus, the structure of the unbound protein (820, unbound protein structure) can be predicted. In this case, the predicted structure 820 of the unbound protein may include position information and direction information of residues included in the protein. In addition, when the processor 110 predicts the structure 820 of the unbound protein using the first neural network model 810, the processor 110 obtains the hidden vector 830 output from the hidden layer of the first neural network model 810. can do. Subsequently, the processor 110 may predict an unbound ligand structure (821) by using the second neural network model 811 based on the ligand graph 801 to be bound. In this case, the predicted structure of unbound ligand' 821 may include position information and binding information of atoms included in the ligand. In addition, when the processor 110 predicts the structure '821 of the unbound ligand using the second neural network model 811, the hidden vector' 831 output from the hidden layer may be obtained. Subsequently, the processor generates [the structure of the unbound protein 820, the hidden vector 830 extracted from the first neural network model 810] and the [structure of the unbound ligand 821], the second neural network model 811 The protein-ligand binding structure 850 may be predicted using the third neural network model 840 based on the hidden vector 831 extracted from ). At this time, the method of Example-A or Example-B can be effective in that it can reflect everything from changes in side chains included in protein residues to changes in backbones.

As an additional embodiment, the third neural network model 840 is a protein residue included in the binding structure 850 using the methods of Examples-A to Example-B according to an embodiment of the present disclosure mentioned above. can be updated by repeating them multiple times. For example, the processor 110 generates update information for each residue using the third neural network model 840 again based on the method of Embodiment-A or Embodiment-B, and based on the update information, the processor 110 The protein structure (ie, each residue) included in linkage structure 850 can be iteratively tailored. That is, the processor 110 may precisely adjust the coupling structure 850 by repeating the method of the previous example a plurality of times.

9 is a simplified and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Although the present disclosure has been described above as being generally embodied by a computing device, those skilled in the art will understand that the present disclosure may be combined with computer executable instructions and/or other program modules that may be executed on one or more computers and/or with hardware. It will be appreciated that it can be implemented as a combination of software.

Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, it will be appreciated by those skilled in the art that the methods of the present disclosure may be used in single-processor or multiprocessor computer systems, minicomputers, mainframe computers as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, and the like. It will be appreciated that other computer system configurations may be implemented, including (each of which may be operative in connection with one or more associated devices).

The described embodiments of the present disclosure may be practiced in a distributed computing environment where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

Computers typically include a variety of computer readable media. Computer readable media can be any medium that can be accessed by a computer, including volatile and nonvolatile media, transitory and non-transitory media, removable and non-transitory media. Includes removable media. By way of example, and not limitation, computer readable media may include computer readable storage media and computer readable transmission media. Computer readable storage media are volatile and nonvolatile media, transitory and non-transitory, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. includes media Computer readable storage media may include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage device, magnetic cassette, magnetic tape, magnetic disk storage device or other magnetic storage device. device, or any other medium that can be accessed by a computer and used to store desired information.

A computer readable transmission medium typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism. Including all information delivery media. The term modulated data signal means a signal that has one or more of its characteristics set or changed so as to encode information within the signal. By way of example, and not limitation, computer readable transmission media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above are also intended to be included within the scope of computer readable transmission media.

An exemplary environment 1100 implementing various aspects of the present disclosure is shown including a computer 1102, which includes a processing unit 1104, a system memory 1106, and a system bus 1108. do. System bus 1108 couples system components, including but not limited to system memory 1106 , to processing unit 1104 . Processing unit 1104 may be any of a variety of commercially available processors. Dual processor and other multiprocessor architectures may also be used as the processing unit 1104.

System bus 1108 may be any of several types of bus structures that may additionally be interconnected to a memory bus, a peripheral bus, and a local bus using any of a variety of commercial bus architectures. System memory 1106 includes read only memory (ROM) 1110 and random access memory (RAM) 1112 . A basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, or EEPROM, and is a basic set of information that helps transfer information between components within computer 1102, such as during startup. contains routines. RAM 1112 may include high-speed RAM, such as static RAM for caching data.

Computer 1102 includes an internal hard disk drive (HDD) 1114 (e.g., EIDE, SATA), which internal hard disk drive 1114 may be configured for external use in a suitable chassis (not shown). , a magnetic floppy disk drive (FDD) 1116 (e.g., for reading from or writing to removable diskette 1118), and an optical disk drive 1120 (e.g., a CD-ROM disk ( 1122) or for reading from or writing to other high-capacity optical media such as DVD). The hard disk drive 1114, magnetic disk drive 1116, and optical disk drive 1120 are connected to the system bus 1108 by a hard disk drive interface 1124, magnetic disk drive interface 1126, and optical drive interface 1128, respectively. ) can be connected to The interface 1124 for external drive implementation includes at least one or both of USB (Universal Serial Bus) and IEEE 1394 interface technologies.

These drives and their associated computer readable media provide non-volatile storage of data, data structures, computer executable instructions, and the like. In the case of computer 1102, drives and media correspond to storing any data in a suitable digital format. Although the description of computer readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art can use zip drives, magnetic cassettes, flash memory cards, and cartridges. It will be appreciated that other tangible computer readable media such as , , and the like may also be used in the exemplary operating environment and that any such media may include computer executable instructions for performing the methods of the present disclosure. .

A number of program modules may be stored on the drive and RAM 1112, including an operating system 1130, one or more application programs 1132, other program modules 1134, and program data 1136. All or portions of the operating system, applications, modules and/or data may be cached in RAM 1112. It will be appreciated that the present disclosure may be implemented in a variety of commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1102 through one or more wired/wireless input devices, such as a keyboard 1138 and a pointing device such as a mouse 1140. Other input devices (not shown) may include a microphone, IR remote control, joystick, game pad, stylus pen, touch screen, and the like. Although these and other input devices are often connected to the processing unit 1104 through an input device interface 1142 that is connected to the system bus 1108, a parallel port, IEEE 1394 serial port, game port, USB port, IR interface, may be connected by other interfaces such as the like.

A monitor 1144 or other type of display device is also connected to the system bus 1108 through an interface such as a video adapter 1146. In addition to the monitor 1144, computers typically include other peripheral output devices (not shown) such as speakers, printers, and the like.

Computer 1102 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 1148 via wired and/or wireless communications. Remote computer(s) 1148 may be a workstation, computing device computer, router, personal computer, handheld computer, microprocessor-based entertainment device, peer device, or other common network node, and generally includes It includes many or all of the components described for, but for simplicity, only memory storage device 1150 is shown. The logical connections shown include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, such as a wide area network (WAN) 1154 . Such LAN and WAN networking environments are common in offices and corporations and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to worldwide computer networks, such as the Internet.

When used in a LAN networking environment, computer 1102 connects to local network 1152 through wired and/or wireless communication network interfaces or adapters 1156. Adapter 1156 may facilitate wired or wireless communications to LAN 1152 , which includes a wireless access point installed therein to communicate with wireless adapter 1156 . When used in a WAN networking environment, computer 1102 may include a modem 1158, be connected to a communicating computing device on WAN 1154, or establish communications over WAN 1154, such as over the Internet. have other means. A modem 1158, which may be internal or external and a wired or wireless device, is connected to the system bus 1108 through a serial port interface 1142. In a networked environment, program modules described for computer 1102, or portions thereof, may be stored on remote memory/storage device 1150. It will be appreciated that the network connections shown are exemplary and other means of establishing a communication link between computers may be used.

Computer 1102 is any wireless device or entity that is deployed and operating in wireless communication, eg, printers, scanners, desktop and/or portable computers, portable data assistants (PDAs), communication satellites, wireless detectable tags associated with It operates to communicate with arbitrary equipment or places and telephones. This includes at least Wi-Fi and Bluetooth wireless technologies. Thus, the communication may be a predefined structure as in conventional networks or simply an ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) makes it possible to connect to the Internet without wires. Wi-Fi is a wireless technology, such as a cell phone, that allows such devices, eg, computers, to transmit and receive data both indoors and outdoors, i.e. anywhere within coverage of a base station. Wi-Fi networks use a radio technology called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in the unlicensed 2.4 and 5 GHz radio bands, for example, at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or in products that include both bands (dual band) .

Those skilled in the art will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, instructions, information, signals, bits, symbols and chips that may be referenced in the above description are voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields s or particles, or any combination thereof.

Those skilled in the art will understand that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein are electronic hardware, (for convenience) , may be implemented by various forms of program or design code (referred to herein as software) or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and the design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Various embodiments presented herein may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques. The term article of manufacture includes a computer program, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash memory devices (eg, EEPROM, cards, sticks, key drives, etc.), but are not limited thereto. Additionally, various storage media presented herein include one or more devices and/or other machine-readable media for storing information.

It is to be understood that the specific order or hierarchy of steps in the processes presented is an example of exemplary approaches. Based upon design priorities, it is to be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of this disclosure. The accompanying method claims present elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure, and the general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present disclosure is not to be limited to the embodiments presented herein, but is to be interpreted in the widest scope consistent with the principles and novel features presented herein.

Claims (16)

A method of predicting the structure of a protein using a neural network model, performed by a computing device, comprising:
Acquiring information on each residue of the protein;
acquiring a first change amount for a 1 th residue of the protein predicted by the neural network model to an n th change amount for an n th residue of the protein;
generating update information for the n-th residue by performing an accumulation operation on the first through n-th changes; and
Adjusting the structure of the protein based on the update information
including,
wherein n is a natural number greater than or equal to 2;
method.
According to claim 1,
The method,
Updating the rigid body motion of some residues of the protein.
Including more,
method.
According to claim 2,
Some of the residues of the protein include i-th to j-th residues of the protein,
Updating the rigid body motion of some residues of the protein,
Updating the i + 1 th change amount to the j th change amount for the i + 1 th residue to the j th residue of the protein to a default value,
wherein i is a natural number, and j is a natural number greater than i,
method.
According to claim 1,
Generating update information for the n-th residue by performing an accumulation operation on the first change amount to the n-th change amount,
performing scaling on each of the first to nth changes based on a degree of flexibility predetermined for each of the first to nth residues of the protein; and
Generating update information for the n-th residue by performing an accumulation operation on the scaled first to n-th change amounts
including,
method.
According to claim 4,
The degree of flexibility predetermined for each of the 1st to nth residues of the protein,
Predetermined based on structural information analyzed by nuclear magnetic resonance spectroscopy (NMR), a temperature factor obtained from an x-ray image, or the output of another neural network model,
method.
A method of predicting the structure of a protein using a neural network model, performed by a computing device, comprising:
obtaining information about each residue of the protein;
obtaining first through n-th backbone twist angle changes for the first to n-th residues of the protein predicted by the neural network model;
generating update information for the first residue to the n-th residue of the protein, based on the first through n-th backbone twist angle changes; and
Adjusting the structure of the protein based on the update information
including,
wherein n is a natural number greater than or equal to 2;
method.
According to claim 6,
Generating update information for the 1st residue to the nth residue of the protein based on the first to nth backbone twist angle change amount to the nth backbone twist angle change amount,
scaling each of the first through n-th backbone twist angle changes based on a degree of flexibility predetermined for each of the 1-th residue to the n-th residue of the protein; and
Generating update information for the first residue to the nth residue of the protein based on the first to nth backbone twist angle change amount to the nth backbone twist angle change amount for which the scaling is performed.
including,
method.
According to claim 1,
Adjusting the structure of the protein,
Adjusting at least one of the pre-binding structure of the protein or the binding structure of the protein
including,
The neural network model,
A neural network model that predicts the pre-binding structure of a protein;
A neural network model predicting the binding structure of a plurality of proteins; or
A neural network model that predicts the binding structure of proteins and ligands
including at least one of
method.
As a structure of a neural network model for predicting the binding structure of a protein,
a first neural network model generating pre-binding structural information of the protein based on the sequence of the protein;
a second neural network model generating pre-combination structural information of the fusion target based on information on the fusion target; and
A third neural network model predicting a structure in which the protein and the binding target are combined based on the pre-binding structural information of the protein and the pre-binding structural information of the binding target
including,
The third neural network model,
obtaining a first change amount for the first residue of the protein to an n-th change amount for the n-th residue of the protein; and
generating update information for the n-th residue by performing an accumulation operation on the first through n-th changes;
based on
Updating a structure in which the protein and the binding target are bound;
wherein n is a natural number greater than or equal to 2;
Structure of the neural network model.
According to claim 9,
The binding target is a ligand,
The third neural network model updates the structure in which the protein and the ligand are bound.
Structure of the neural network model.
According to claim 9,
The binding target is another protein,
The second neural network model is implemented with the same model as the first neural network model,
The third neural network model uses update information associated with the protein and update information associated with the other protein to update a structure in which the protein and the other protein are combined,
Structure of the neural network model.
As a device,
at least one processor; and
Memory;
including,
The at least one processor,
obtaining information about each residue in the protein;
obtaining a first change amount for a 1-th residue of the protein predicted by a neural network model to an n-th change amount for an n-th residue of the protein;
generating update information for the n-th residue by performing an accumulation operation on the first through n-th changes; and
configured to adjust the structure of the protein based on the update information;
wherein n is a natural number greater than or equal to 2;
Device.
A computer program stored on a computer-readable storage medium, which, when executed by at least one processor, causes the at least one processor to perform operations for predicting the structure of a protein, the operations comprising:
obtaining information about each residue of the protein;
obtaining a first change amount for a 1-th residue of the protein predicted by a neural network model to an n-th change amount for an n-th residue of the protein;
generating update information for the n-th residue by performing an accumulation operation on the first through n-th changes; and
Adjusting the structure of the protein based on the update information
including,
wherein n is a natural number greater than or equal to 2;
A computer program stored on a computer readable storage medium.
As a structure of a neural network model for predicting the binding structure of a protein,
a first neural network model generating pre-binding structural information of the protein based on the sequence of the protein;
a second neural network model generating pre-combination structural information of the fusion target based on information on the fusion target; and
A third neural network model predicting a structure in which the protein and the binding target are combined based on the pre-binding structural information of the protein and the pre-binding structural information of the binding target
including,
The third neural network model,
obtaining first through n-th backbone twist angle changes for the first to n-th residues of the protein; and
Generating update information for the 1st residue to the nth residue of the protein based on the 1st backbone twist angle change amount to the n-th backbone twist angle change amount
Based on, updating the structure in which the protein and the binding target are bound,
wherein n is a natural number greater than or equal to 2;
Structure of the neural network model.
As a device,
at least one processor; and
Memory;
including,
The at least one processor,
obtaining information about each residue in the protein;
obtaining first through n-th backbone twist angle changes for the first to n-th residues of the protein predicted by the neural network model;
generating update information for the first residue to the n-th residue of the protein based on the first through n-th backbone twist angle changes; and
configured to adjust the structure of the protein based on the update information;
wherein n is a natural number greater than or equal to 2;
Device.
A computer program stored on a computer-readable storage medium, which, when executed by at least one processor, causes the at least one processor to perform operations for predicting the structure of a protein, the operations comprising:
obtaining information about each residue of the protein;
obtaining first through n-th backbone twist angle changes for the first to n-th residues of the protein predicted by the neural network model;
generating update information for the first to nth residues of the protein, based on the first through nth backbone twist angle changes; and
Adjusting the structure of the protein based on the update information
including,
wherein n is a natural number greater than or equal to 2;
A computer program stored on a computer readable storage medium.

Images