KR102577105B1 - Apparatus and method for building a pipelines to explore adverse drug reaction - Google Patents

Abstract

The present invention relates to a method and device for constructing a pipeline for detecting drug side effects. A method of building a pipeline for detecting drug side effects according to an embodiment of the present invention is to connect to a social channel based on social network services (SNS) and use at least one word related to the target drug based on the first term set. collecting social data; Preprocessing the collected social data; Extracting data related to side effects from the preprocessed social data based on a second term set and performing exploratory data analysis; Analyzing drug side effect patterns for the target drug according to the analysis results and classifying them according to preset categories; and constructing or learning a drug side effect detection prediction model using the classification result, wherein the first term set includes data sets composed of at least one term each representing at least one drug, and the second term set includes data sets consisting of at least one term representing each of at least one drug. The term set may include a data set comprised of terms representing drug side effects.

Description

Pipeline building method and device for detecting drug side effects {APPARATUS AND METHOD FOR BUILDING A PIPELINES TO EXPLORE ADVERSE DRUG REACTION}

The present invention relates to a method and device for building a pipeline for detecting side effects of drugs, and more specifically, a method and device for building a pipeline for detecting side effects of drugs that allows exploring and updating side effects of drugs through social data analysis. It's about.

Recently, the frequency and quantity of drug use is rapidly increasing worldwide due to aging. As a result, drug side effects are also rapidly increasing, threatening patient safety. Accordingly, early detection by analyzing new or serious side effects of commercially available drugs is emerging as an important issue.

In addition to the publicly known side effects of the target drug, new side effects may occur, and additional side effects may occur due to long-term use.

However, it is cumbersome and inconvenient for drug users who take the target drug in their daily lives to input the drug and drug use review data and individually collect or monitor the symptoms caused by taking the drug, so there is a limit to obtaining actual information about the drug. There is.

Meanwhile, as the use of social networking has recently increased rapidly in the medical industry, many people share their emotions and experiences on social network services (SNS). As an act of sharing valuable information, they respond by posting reviews of specific drugs they have taken or by leaving comments.

In this way, information posted or registered through social networking services corresponds to direct reviews of use by individuals, so it will be meaningful as it includes data related to new side effects or indications in addition to the side effects already known for the specific drug.

Therefore, there is a need to develop technology that can more accurately predict drug side effects caused by the target drug by collecting social data related to the target drug based on social network services and analyzing the pattern of drug side effects.

Korean Patent Publication No. 10-2015-0049937 (Publication date: May 8, 2015)

The present invention was proposed to solve the problems described above. By collecting social data related to the target drug based on a social network service and analyzing the pattern of drug side effects, it is possible to more accurately predict drug side effects caused by the target drug. To provide a pipeline construction method and device for detecting drug side effects.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood by those skilled in the art from the description below.

The method of building a pipeline for detecting drug side effects according to an embodiment of the present invention to solve the above-described problem involves accessing a social channel based on social network services (SNS) to create a first term set. collecting at least one social data related to the target drug based on; Preprocessing the collected social data; Extracting data related to side effects from the preprocessed social data based on a second term set and performing exploratory data analysis; Analyzing drug side effect patterns for the target drug according to the analysis results and classifying them according to preset categories; and constructing or learning a drug side effect detection prediction model using the classification result, wherein the first term set includes data sets composed of at least one term each representing at least one drug, and the second term set includes data sets consisting of at least one term representing each of at least one drug. The term set may include a data set comprised of terms representing drug side effects.

Meanwhile, a pipeline construction device for detecting drug side effects according to an embodiment of the present invention includes a communication module; a storage module that stores at least one piece of information or data for building a pipeline for detecting the drug side effects; Access a social channel based on Social Network Services (SNS) to collect and pre-process at least one social data related to the target drug based on a first term set, and pre-process the data based on a second term set. an analysis module that extracts data related to side effects from social data, performs exploratory data analysis, analyzes drug side effect patterns for the target drug according to the analysis results, and classifies them according to preset categories; A learning module that builds or learns a drug side effect detection prediction model using the classification results; And collect and preprocess at least one social data related to the target drug based on the first term set, and perform exploratory data analysis by extracting data related to side effects from the preprocessed social data based on the second term set. And, according to the analysis results, it analyzes the drug side effect pattern for the target drug, classifies it according to preset categories, and then uses the classification result to build or learn a drug side effect detection prediction model. , the first term set may include data sets composed of at least one term representing each of at least one drug, and the second term set may include a data set composed of terms representing drug side effects.

Other specific details of the invention are included in the detailed description and drawings.

According to the present invention, by collecting social data related to a target drug based on a social network service and analyzing the pattern of drug side effects, it is possible to more accurately predict drug side effects caused by the target drug.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

Figure 1 is a block diagram showing the configuration of a pipeline construction device for detecting drug side effects according to an embodiment of the present invention.
Figure 2 is a flow chart showing a pipeline construction method for detecting drug side effects according to an embodiment of the present invention.
Figure 3 is a diagram schematically showing a pipeline for detecting drug side effects according to an embodiment of the present invention.
Figure 4 is a diagram showing the creation procedure of a second term set used to build a pipeline for detecting drug side effects according to an embodiment of the present invention.
Figure 5a is a diagram showing a first embodiment of analyzing drug side effect patterns when building a pipeline for drug side effect detection according to an embodiment of the present invention.
Figure 5b is a diagram showing a second embodiment of analyzing drug side effect patterns when building a pipeline for drug side effect detection according to an embodiment of the present invention.
Figure 5c is a diagram showing a third embodiment of analyzing drug side effect patterns when building a pipeline for drug side effect detection according to an embodiment of the present invention.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to provide a general understanding of the technical field to which the present invention pertains. It is provided to fully inform the skilled person of the scope of the present invention, and the present invention is only defined by the scope of the claims.

The terminology used herein is for describing embodiments and is not intended to limit the invention. As used herein, singular forms also include plural forms, unless specifically stated otherwise in the context. As used in the specification, “comprises” and/or “comprising” does not exclude the presence or addition of one or more other elements in addition to the mentioned elements. Like reference numerals refer to like elements throughout the specification, and “and/or” includes each and every combination of one or more of the referenced elements. Although “first”, “second”, etc. are used to describe various components, these components are of course not limited by these terms. These terms are merely used to distinguish one component from another. Therefore, it goes without saying that the first component mentioned below may also be a second component within the technical spirit of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless clearly specifically defined.

Spatially relative terms such as “below”, “beneath”, “lower”, “above”, “upper”, etc. are used as a single term as shown in the drawing. It can be used to easily describe the correlation between a component and other components. Spatially relative terms should be understood as terms that include different directions of components during use or operation in addition to the directions shown in the drawings. For example, if a component shown in a drawing is flipped over, a component described as “below” or “beneath” another component will be placed “above” the other component. You can. Accordingly, the illustrative term “down” may include both downward and upward directions. Components can also be oriented in other directions, so spatially relative terms can be interpreted according to orientation.

As used in the specification, the term “unit” or “module” refers to a hardware component such as software, FPGA, or ASIC, and the “unit” or “module” performs certain roles. However, “part” or “module” is not limited to software or hardware. A “unit” or “module” may be configured to reside on an addressable storage medium and may be configured to run on one or more processors. Thus, as an example, a “part” or “module” refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, Includes procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functionality provided within components and “parts” or “modules” can be combined into smaller components and “parts” or “modules” or into additional components and “parts” or “modules”. Could be further separated.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless clearly specifically defined.

Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

Figure 1 is a block diagram showing the configuration of a pipeline construction device for detecting drug side effects according to an embodiment of the present invention.

Referring to Figure 1, the pipeline construction device (hereinafter referred to as 'construction device') 100 for predicting drug side effects includes a communication module 110, a storage module 130, an analysis module 150, and a learning module 170. ) and a control module 190.

The communication module 110 transmits and receives various information or data needed to build a pipeline for detecting drug side effects with an external device. Specifically, social data can be collected by connecting to a social network service (portal site, messenger service, etc.) through this communication module 110. In other words, this communication module 110 is used to communicate with other terminals, servers, devices, etc., and transmits and receives wireless signals in a communication network based on wireless Internet technologies.

Wireless Internet technologies include, for example, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband), and WiMAX (Worldwide). Interoperability for Microwave Access), HSDPA (High Speed Downlink Packet Access), HSUPA (High Speed Uplink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), etc., and meal monitoring devices (100 ) transmits and receives data according to at least one wireless Internet technology, including Internet technologies not listed above.

For short range communication, Bluetooth™ RFID (Radio Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra Wideband), ZigBee, NFC (Near Field Communication), Wi-Fi Short-distance communication can be supported using at least one of (Wireless-Fidelity), Wi-Fi Direct, and Wireless USB (Wireless Universal Serial Bus) technologies. The device 100 for constructing such wireless area networks It can support wireless communication between and external devices.At this time, the short-range wireless communication network may be a short-range wireless personal area network.

The storage module 130 stores at least one set of terms as drug-related information and at least one process required to build a pipeline for detecting drug side effects.

The analysis module 150 collects social data about the target drug based on a social network service, builds at least one term set, and analyzes and classifies drug side effect patterns for the target drug based on this. Here, social data may be a portal site or messenger service based on a social network service.

To this end, the analysis module 150 may be configured to include a collection unit 151, a preprocessing unit 153, a term generation unit 155, a pattern analysis unit 157, and a classification unit 159.

Specifically, the collection unit 151 connects to a social network and collects at least one social data related to the target drug, and the preprocessing unit 153 preprocesses (refines) the collected at least one social data to generate a term. Part 155 builds at least one term set for the target drug. Thereafter, the pattern analysis unit 157 extracts side effect-related data from the collected social data based on at least one constructed term set to analyze the side effect pattern, and the classification unit 159 preliminarily based on the analysis results. Classify side effect information according to established categories. At this time, the preset categories may be divided into known side effects, unknown side effects, and indications, but this is not limited, and may be changed or additional categories may be added.

Here, at least one term set is a first term set, which is a data set for terms representing (referring to) the target drug (e.g., ingredient, product name, brand, nickname, abbreviation, etc.) and a standardized term set related to drug side effects. It may include a second term set, which is a data set of terms indicating drug side effects related to the target drug constructed based on a public database. At this time, the first term set and the second term set may separately include data sets for each of a plurality of drugs. However, when the pipeline is first built, input is made by an administrator or worker, the pipeline is run, and the first term set and the second term set can be updated, respectively.

Meanwhile, when the analysis module 150 accesses a social network and collects at least one social data related to a target drug, it may use a pre-established first term set. That is, the analysis module 150 collects social data using keywords included in the data set for the target drug in the first term set.

The learning module 170 performs drug side effect labeling on the data classified by the analysis module 150 and then uses the labeled data as learning data to build and/or learn a drug side effect detection prediction model.

At this time, the learning module 170 can build a drug side effect detection prediction model based on the Recurrent Neural Network (RNN) learning method. As a structure constituting the Recurrent Neural Network (RNN), the LSTM (Long short term memory network) ) method can be used.

Meanwhile, the learning module 170 uses the drug side effect detection prediction model for learning based on the data classified as known side effects among the classified data, and when the data classified as unknown side effects accumulates more than a preset number of times, it becomes known. It can be reclassified as a side effect and used to learn a prediction model for detecting drug side effects.

The control module 190 connects to a social network to collect at least one social data related to the target drug, and when the collected at least one social data is preprocessed (refined), the term generator 155 generates the target drug. Control to build at least one term set for Thereafter, the control module 190 extracts data related to side effects from the collected social data based on at least one constructed term set, analyzes side effect patterns, and provides side effect information according to preset categories based on the analysis results. Control to classify.

Figure 2 is a flow chart showing a pipeline construction method for detecting drug side effects according to an embodiment of the present invention.

Referring to FIG. 2, in order to build a pipeline for detecting drug side effects, the analysis module 150 connects to a social network and uses a data set related to the target drug among the first term sets to create at least one social network related to the target drug. Data is collected and preprocessed (S201), and data related to side effects are extracted from the preprocessed social data based on the second term set to analyze side effect patterns (S203).

Thereafter, labeling of drug side effects is performed based on the data extracted according to the analysis results (S205). At this time, labeling may be performed automatically by the analysis module 150 or manually by a manager or worker. You can.

Afterwards, the learning module 170 uses the labeled data as learning data to learn a drug side effect prediction module (S207).

Figure 3 is a diagram schematically showing a pipeline for detecting drug side effects according to an embodiment of the present invention. This pipeline predicts (detects) side effects for target drugs continuously or periodically and updates the results. Here, the first term set and the second term set will be explained assuming that pre-constructed (generated) ones are used.

Referring to FIG. 3, the construction device 100 accesses a social channel based on a social network service (S301) and provides at least one social data for the target drug based on the first term set for the target drug. Collect (S303).

At least one social data collected in this way is unstructured and unexpected, and must be preprocessed (refined) in order to use it. In this case, the unstructured data can be refined through natural language processing or text mining (S305) .

Here, text mining is a mining process for unstructured data, extracting statistically meaningful concepts or characteristics from data, and deriving high-quality information such as patterns or trends between them.

Meanwhile, at least one preprocessed social data is analyzed based on exploratory data analysis based on the second term set (S307), and drug side effect patterns are analyzed using the analysis results (S309). Here, exploratory data analysis may be analyzing valid data among big data according to user settings, condensing the analyzed valid data according to preset setting options, and visualizing it.

Here, exploratory data analysis can be done based on correlation analysis or word embedding (word2vec) model. Specifically, correlation analysis is an analysis that finds interrelationships or dependencies between data that exist within data, enabling simple but clear interpretation of results. Through this, it can be useful in discovering unexpected knowledge from data by creating easy-to-understand rules for data. In addition, word embedding is a reasoning-based technique that vectorizes the meaning of words to take into account similarities between words (keywords). It quantifies the text and discovers knowledge from a new perspective. For this purpose, a vector that contains the meaning of the word as much as possible is created and the similarity or relatedness of the word pair is checked and interpreted.

According to the analysis results in step S309, the drug side effect pattern is classified into one category. For example, the target drug may be classified into any one of known side effects, unknown side effects, and indications.

Afterwards, labeling is performed based on data classified as known side effects among the classified data, and a prediction model for detecting drug side effects is learned using the labeled data as learning data.

The drug side effect detection prediction model learned in this way is used to detect side effects of social data collected later, and as its performance improves each time social data is collected and analyzed, more accurate prediction and detection becomes possible.

Figure 4 is a diagram showing the creation procedure of a second term set used to build a pipeline for detecting drug side effects according to an embodiment of the present invention.

Referring to FIG. 4, the second term set may be created using a list of drug side effects created based on WHO-ART and SIDER, which are standardized and public databases, and a dictionary of pre-generated consumer terms. Here, WHO-ART represents an international classification system for drug side effect terminology, and is already the most widely used in ADR (Adverse Drug Reaction) reporting in Korea. In addition, SIDER is a drug side effect database that provides Drug-ADR relationships and includes information on commercially available drugs and drug side effects for those drugs. Additionally, the consumer terminology dictionary may be defined by an administrator or worker. That is, the second term set is created using standardized sets such as WHO-ART and SIDER.

Specifically, a list of drug side effects is created based on WHO-ART (S401), the list of drug side effects is supplemented based on SIDER (S403), and a list of drug side effects is created based on consumer terms (S405). By mutual mapping (S407), a second term set is generated as a result (S409).

However, the execution order of steps S401 to S405 is not specified, and can be performed independently (individually), and updates can be individually performed automatically or manually as needed.

However, WHO-ART and SIDER are only examples of standardized and publicly available databases, and are not limited thereto, and other drug systems, drug data, etc. may be used.

Meanwhile, various methods may be applied to exploratory data analysis for the construction device 100 to analyze drug side effect patterns from at least one social data collected based on a social network service. Examples of each method will be described using FIGS. 5A to 5C.

Figure 5a is a diagram showing a first embodiment of analyzing drug side effect patterns when building a pipeline for drug side effect detection according to an embodiment of the present invention.

According to FIG. 5A, the analysis module 150 of the construction device 100 extracts words (keywords) that appear in at least one collected social data, analyzes the frequency of each of the extracted words, and produces (a) ) Create a frequency table as shown in .

Additionally, by using the terms to create a word cloud as shown in (b), it is possible to visually check which words are used and how frequently they are used.

Figure 5b is a diagram showing a second embodiment of analyzing drug side effect patterns when building a pipeline for drug side effect detection according to an embodiment of the present invention.

According to FIG. 5B, the analysis module 150 of the construction device 100 extracts words (keywords) that appear in at least one collected social data, and creates a table of the top n related words among the extracted words. It also creates a visualization graph so that you can visually check it by analyzing the relationships between the words.

Figure 5c is a diagram showing a third embodiment of analyzing drug side effect patterns when building a pipeline for drug side effect detection according to an embodiment of the present invention.

According to FIG. 5C, the analysis module 150 of the construction device 100 extracts words (keywords) that appear in at least one social data collected through word embedding, embeds the extracted words, and then matches them with the side effect dictionary. By doing this, you can identify the side effect pattern for each organ system (SOC) of the human body and check the drug side effect pattern by identifying the top words in order of the cosine distance of a specific side effect.

Here, word2vec can be used as one of the word embedding methods.

The construction device 100 according to an embodiment of the present invention may be a server that builds a pipeline for detecting drug side effects, and the construction device 100 supports multiple applications (i.e., applications) desired by an administrator or worker. Computers that can be installed and run, UMPC (Ultra Mobile PC), workstations, net-books, PDAs (Personal Digital Assistants), portable computers, web tablets, wireless phones ), mobile phone, smart phone, Pad, smart watch, wearable terminal, e-book, PMP (portable multimedia player), It may be a portable game console, navigation device, black box or digital camera, or other mobile communication terminal. Accordingly, the construction device 100 may need to install a separate program or application to build the pipeline. However, this is only one example, and it is also possible to build a pipeline by accessing a web page.

The steps of the method or algorithm described in connection with embodiments of the present invention may be implemented directly in hardware, implemented as a software module executed by hardware, or a combination thereof. The software module may be RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), Flash Memory, hard disk, removable disk, CD-ROM, or It may reside on any type of computer-readable recording medium well known in the art to which the present invention pertains.

Above, embodiments of the present invention have been described with reference to the attached drawings, but those skilled in the art will understand that the present invention can be implemented in other specific forms without changing its technical idea or essential features. You will be able to understand it. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive.

100: Pipeline construction device 110: Communication module
130: storage module 150: analysis module
170: Learning module 190: Control module
151: collection unit 153: preprocessing unit
155: term generation unit 157: pattern analysis unit
159: Classification department

Claims (5)

In a method of building a pipeline for detecting drug side effects, performed by a device,
An analysis module accessing a social channel based on a social network service (Social Network Services, SNS) to collect at least one social data related to a target drug based on a first term set;
Preprocessing the collected social data by the analysis module;
performing exploratory data analysis by the analysis module extracting data related to side effects from the preprocessed social data based on a second term set;
The analysis module analyzes drug side effect patterns for the target drug according to the analysis results and classifies them according to preset categories; and
A learning module includes a step of building or learning a drug side effect detection prediction model using the classification results,
The first term set includes data sets composed of at least one term representing each of at least one drug, and the second term set includes a data set composed of terms representing drug side effects,
The preset categories are set by the analysis module by dividing them into known side effects, unknown side effects, and indications for the target drug,
The drug side effect detection prediction model is learned by using the data classified as known side effects by the learning module as learning data, and among the data classified as unknown side effects, data accumulated more than a preset number of times is used to study the known side effects. Characterized in that it is reclassified and later used as learning data to learn the drug side effect detection prediction model,
How to build a pipeline for drug side effect detection.
delete According to paragraph 1,
The exploratory data analysis is,
This is achieved by extracting keywords appearing in the at least one social data collected by the analysis module and creating a visualized graph using the extracted keywords,
The visualized graph is characterized in that it is generated using any one of a frequency analysis technique, a correlation analysis technique, and a word embedding technique.
How to build a pipeline for drug side effect detection.
According to paragraph 1,
The second set of terms is,
Characterized in that it is generated by mapping the list of drug side effects for the target drug obtained based on SIDER by the analysis module and the pre-generated dictionary of consumer terms to WHO-ART,
How to build a pipeline for drug side effect detection.
In a pipeline construction device for detecting drug side effects,
communication module;
a storage module that stores at least one piece of information or data for building a pipeline for detecting the drug side effects;
Access a social channel based on Social Network Services (SNS) to collect and pre-process at least one social data related to the target drug based on a first term set, and pre-process the data based on a second term set. an analysis module that extracts data related to side effects from the social data, performs exploratory data analysis, analyzes drug side effect patterns for the target drug according to the analysis results, and classifies them according to preset categories;
A learning module that builds or learns a drug side effect detection prediction model using the classification results; and
Collect and preprocess at least one social data related to the target drug based on the first term set, extract data related to side effects from the preprocessed social data based on the second term set, and perform exploratory data analysis; , Analyzing the drug side effect pattern for the target drug according to the analysis results, classifying it according to preset categories, and then using the classification result to control a control module to build or learn a drug side effect detection prediction model,
The first term set includes data sets composed of at least one term representing each of at least one drug, and the second term set includes a data set composed of terms representing drug side effects,
The preset categories are divided into known side effects, unknown side effects, and indications for the target drug by the analysis module, and
The drug side effect detection prediction model is learned by using the data classified as known side effects by the learning module as learning data, and among the data classified as unknown side effects, data accumulated more than a preset number of times is used to study the known side effects. Characterized in that it is reclassified and later used as learning data to learn the drug side effect detection prediction model,
Pipeline building device for drug side effect detection.

Images