Microorganism identification method using learned deep learning and system thereof

Abstract

A method for identifying microorganisms performed by a processor according to one embodiment may include: a step of generating training data including multiple mass spectrum data and reference strain information mapped to each mass spectrum data by classifying multiple strain data included in a strain database by strain; a step of preprocessing mass spectrum data from the generated training data; a step of generating peak characteristic data related to peak values ​​of the mass spectrum data from the preprocessed mass spectrum data; a step of training a deep learning model by updating parameters of the deep learning model using the peak characteristic data; and a step of outputting an identification result for an input strain using the trained deep learning model.

Description

Technology Field

[0001] Below, technology regarding a method and apparatus for identifying microorganisms using a trained deep learning model is provided. Background Technology

[0002] The method of identifying microorganisms using a trained deep learning model is based on the fact that the mass spectrum patterns of individual microorganisms differ by species, and thus, the species can be identified by applying a statistical algorithm capable of distinguishing mass spectrum patterns. As results capable of differentiation at the species level accumulate, about 2,600 species can be distinguished, and 2,000 to 20,000 Da of data can be used to distinguish species at the species level.

[0003] Similarity-based algorithms used for species-level identification cannot distinguish strain-level differences such as methicillin-resistant Staphylococcus aureus infection (MRSA) and methicillin-sensitive Staphylococcus aureus infection (MSSA). Although MRSA and MSSA can be distinguished based on the presence or absence of specific peak markers and surrogate markers, such as the phenol-soluble modulin (PSM) family of amphiphilic alpha-peptide toxins, the sensitivity is low. Prior art literature

[0004] Japanese Patent Publication No. 2015-507182 (Date of publication: March 5, 2015) means of solving the problem

[0005] A strain identification method performed by a processor according to one embodiment may include: a step of generating training data including multiple mass spectrum data and reference strain information mapped to each mass spectrum data by classifying multiple strain data included in a strain database by strain; a step of preprocessing mass spectrum data from the generated training data; a step of generating peak characteristic data related to peak values ​​of the mass spectrum data from the preprocessed mass spectrum data; a step of training a machine learning model by updating the parameters of the machine learning model using the peak characteristic data; and a step of outputting an identification result for an input strain using the trained machine learning model.

[0006] According to one aspect, the method further includes the step of mapping reference strain information corresponding to the mass spectrum data to the generated peak characteristic data as true value data, and the step of training the machine learning model may include the step of updating the parameters of the machine learning model based on the peak characteristic data and the true value data mapped to the peak characteristic data.

[0007] The step of preprocessing the mass spectrum data of the strain identification method according to one embodiment may include a step of determining whether the mass spectrum data meets a predetermined criterion, a step of adding the mass spectrum data to training data if it is determined that the mass spectrum data meets the criterion, and a step of excluding the mass spectrum data from the training of a machine learning model if the mass spectrum data does not meet the criterion.

[0008] The above criteria may include whether the signal-to-noise ratio (SNR) of the mass spectrum data is greater than or equal to a threshold ratio.

[0009] Additionally, the step of preprocessing the mass spectrum data may include removing noise from the mass spectrum data, removing the baseline of the noise-removed mass spectrum data, normalizing the signal intensity of the mass spectrum data, detecting peaks in the mass spectrum data, and aligning the detected peaks in the mass spectrum data according to a preset standard.

[0010] The step of detecting the peaks may include comparing the spectral intensity with surrounding points at each point corresponding to an individual mass-to-charge ratio in the mass spectrum data, and detecting peak points from the mass spectrum data based on the result of the comparison of spectral intensities.

[0011] The step of detecting the peak point may include detecting the peak point as a point that is greater than a threshold ratio compared to the spectral intensity of an adjacent point among the points corresponding to a plurality of mass-charge ratios in the mass spectrum data.

[0012] The step of aligning the peaks may include comparing the mass-charge ratio value of the detected peak with a preset representative peak point, and if the difference between the mass-charge ratio value of the detected peak and the representative peak point is less than or equal to a threshold value, aligning the mass-charge ratio value of the detected peak in the mass spectrum data to the representative peak point.

[0013] The step of aligning the above peaks may include the step of determining a representative peak value for the peak point of the aligned peaks.

[0014] According to one embodiment, the step of generating the peak characteristic data may include the step of calculating a characteristic matrix containing element values ​​indicating the spectral intensity for each of a plurality of mass-charge ratio points from the preprocessed mass spectrum data.

[0015] The step of calculating the characteristic matrix according to one side may include the step of determining the element value corresponding to the mass peak point in the characteristic matrix as the peak value of the mass peak point.

[0016] The step of calculating the above characteristic matrix may include the step of excluding the spectral intensity corresponding to a point from the characteristic matrix for a point among the plurality of mass-charge ratio points where no mass peak is detected.

[0017] According to one embodiment, the step of calculating the characteristic matrix may include the step of excluding the spectral intensity corresponding to a point from the characteristic matrix for a point among the plurality of mass-charge ratio points where no mass peak is detected.

[0018] According to another embodiment, the step of calculating the characteristic matrix may include setting the element value corresponding to the point where no mass peak is identified among the plurality of mass-charge ratio points as the spectral intensity at that point.

[0019] According to one aspect, the step of training the machine learning model may include the step of updating the parameters of the temporary machine learning model based on the temporary result value output from the training data and the objective function using the temporary machine learning model.

[0020] Additionally, the machine learning model comprises a plurality of decision trees, each including one or more nodes, and the step of training the machine learning model may include the step of inputting the generated feature data to each of the plurality of decision trees, and the step of training the machine learning model by independently updating the parameters of the machine learning model for the plurality of decision trees.

[0021] The above plurality of strain data may include strain data regarding methicillin-sensitive Staphylococcus aureus infection (MSSA) strains and methicillin-resistant Staphylococcus aureus infection (MRSA) strains.

[0022] In addition, the step of generating the training data may include, for each of the plurality of strain data, a step of mapping a ground truth value indicating the class to which the strain corresponding to the strain data belongs among two classes to a reference strain information for the strain data, and the step of training the machine learning model may include a step of training the machine learning model based on the reference strain information indicating one of the two classes for each strain data.

[0023] Additionally, the step of generating the training data may include, for each of the plurality of strain data, a step of mapping a true value indicating the class to which the strain corresponding to the strain data belongs among three or more multi-classes to a reference strain information for the strain data, and the step of training the machine learning model may include a step of training the machine learning model based on the reference strain information indicating one of the multi-classes for each strain data.

[0024] According to one aspect, the step of mapping the reference strain information may include: mapping a first true value indicating MRSA as the reference strain information to MRSA strain data that carries at least one of the mobile genetic elements of Staphylococcal Cassette Chromosome mec (SCCmec) type II, type III, and type IVA among the plurality of strain data; mapping a second true value indicating MSSA as the reference strain information to MSSA strain data among the plurality of strain data; and mapping a value between the first true value and the second true value as the reference strain information to MRSA strain data that carries the mobile genetic element of SCCmec type IV among the plurality of strain data. Brief explanation of the drawing

[0025] Figure 1 is a flowchart illustrating a strain identification method according to one embodiment. Figure 2 is a diagram showing the total mass spectra of MRSA and MSSA in one embodiment. Figure 3 is a diagram comparing the mass spectra of MRSA and MSSA in a specific mass range according to one embodiment. FIG. 4 is a diagram showing a processed mass spectrum according to one embodiment. FIG. 5 is a diagram illustrating the adjustment of a mass-charge ratio value having a peak according to one embodiment. FIG. 6 is a table showing the characteristic matrix for a plurality of strains according to one embodiment. Figure 7 is a graph showing the results of multidimensional scaling analysis (MDS) according to one embodiment. FIG. 8 is a table showing the results of a two-sample test according to one embodiment. Figure 9 is a graph illustrating a logistic regression model according to one embodiment. FIG. 10 is a graph illustrating a decision tree model according to one embodiment. FIG. 11 is a graph illustrating a decision tree of a random forest model according to one embodiment. FIG. 12 is a graph illustrating a random forest model including a plurality of decision trees according to one embodiment. FIG. 13 is a table and graph showing the results of a random forest model according to one embodiment. FIG. 14 is a graph illustrating the characteristics of MRSA having multiple types according to one embodiment. FIG. 15 is a graph illustrating the characteristics of MRSA having multiple types according to one embodiment. FIG. 16 is a flowchart illustrating an apparatus for determining the strain type of a strain according to one embodiment. Specific details for implementing the invention

[0026] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed forms, and the scope of this specification includes modifications, equivalents, or substitutions that fall within the technical concept.

[0027] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0028] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0029] The singular expression includes the plural expression unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to specify the existence of the described features, numbers, steps, actions, components, parts, or combinations thereof, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0030] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this specification. Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. Identical reference numerals in each drawing indicate identical components.

[0031] Figure 1 is a flowchart illustrating a strain identification method according to one embodiment.

[0032] In step (110), the processor of the strain identification device can generate training data including multiple mass spectrum data and reference strain information mapped to each mass spectrum data by classifying multiple strain data included in the strain database by strain. The mass spectrum data of the strain may represent a spectrum for one of the mass value and mass-charge ratio value of the microorganism's unique protein as a result of analyzing the microorganism's unique protein by a MALDI-TOF mass spectrometer. The mass spectrum described below may represent the spectral intensity for the mass-charge ratio value. However, it is not limited thereto, and the mass spectrum may indicate the spectral intensity for the mass value.

[0033] Strain data may include mass spectrum data for strains analyzed by the MALDI-TOF mass spectrometer described above. Training data consists of pairs of training inputs and training outputs and can be used to train a machine learning model. Training inputs may include strain data for any strain. A training output mapped to any training input may include a ground truth value for that training input. For example, if the training input is mass spectrum data for any strain, the training output mapped to it may include a ground truth value indicating the class of that strain. Thus, training data may include multiple strain data and strain classes mapped to each strain data.

[0034] For example, the strain data described above may include mass spectrum data regarding methicillin-resistant Staphylococcus aureus infection (MRSA) and methicillin-sensitive Staphylococcus aureus infection (MSSA). For example, if the training data consists of two classes, the training output may be one of a first true value (e.g., 1) indicating MRSA and a second true value (e.g., 0) indicating MSSA. For another example, if the training data consists of multiple classes, the training output may be one of true values ​​indicating subtypes of MRSA and true values ​​indicating MSSA. However, this is merely an example and does not limit the composition of the training output to what is described above.

[0035] In step (120), the processor can preprocess mass spectrum data from the training data generated in step (110). The processor can determine whether the mass spectrum data meets a predetermined criterion. If the processor determines that any mass spectrum data in the strain database meets the criterion, it can add the corresponding mass spectrum data to the training data. According to one embodiment, the criterion may indicate whether the signal-to-noise ratio (SNR) of the mass spectrum data is greater than or equal to a threshold ratio. Based on the SNR, effective intensity, etc., the processor can select and classify data having mass spectra suitable for training from the mass spectrum data as training data. A machine learning model trained by the above-described training data set can exhibit high accuracy.

[0036] The processor can smooth the noise from the mass spectrum data to preprocess the mass spectrum data. The processor can baseline correct the removed mass spectrum data. The processor can normalize the signal intensity of the baseline-removed mass spectrum data. The preprocessing operation of the mass spectrum data by the processor will be described in detail later with reference to Fig. 4.

[0037] After the mass spectrum data is normalized, the processor can detect peaks in the mass spectrum data. The processor can align the peaks detected in the mass spectrum data according to preset criteria. The peak detection and alignment operations by the processor will be described later with reference to FIG. 5.

[0038] In step (130), the processor may generate peak characteristic data related to peak values ​​from the preprocessed mass spectrum data. The peak characteristic data may be a type of peak profile data of the mass spectrum. The peak profile data may represent data indicating the intensity of peaks by mass or data indicating the intensity of peaks by mass-charge ratio. An example of peak profile data is illustrated in FIG. 2 below. According to one embodiment, the processor may generate a characteristic matrix containing element values ​​indicating the spectral intensity for each of a plurality of mass-charge ratio points from the preprocessed mass spectrum data. The characteristic matrix may be an input matrix of a machine learning model to be input into the machine learning model to update the parameters of the machine learning model. The element values ​​of the characteristic matrix may be the mass spectrum intensity indicated by a representative point within a certain mass-charge ratio interval. If a peak point exists within a certain mass-charge ratio interval, the processor may align the mass-charge ratio values ​​of the detected peaks based on the representative point according to the peak alignment of the preprocessing process.

[0039] In step (140), the processor can train the machine learning model by updating the parameters of the machine learning model using peak characteristic data. According to one embodiment, the machine learning model is a model designed to produce strain identification results from input strain data, and may also be referred to as, for example, a strain identification model. The input strain data may include mass spectrum data for strains analyzed by a MALDI-TOF mass spectrometer. The strain identification results may include data indicating strains corresponding to the input strain data. The machine learning model may be a combination of one or more of a neural network model with a neural network structure, a logistic regression model, a decision tree model, and a random forest model. The machine learning model may be a model capable of outputting two-class classification or three or more multi-class classification results for the input peak characteristic data. The parameters may have values ​​optimized to produce strain identification results for the input strain data through the training of the machine learning model described above.

[0040] As described in detail in step (110), the processor can map reference strain information corresponding to the mass spectrum to the peak characteristic data as true value (GT value, Ground Truth value) data, and can update the parameters of the machine learning model using the peak characteristic data and the true value data. Specifically, the processor can input the peak characteristic data into a temporary machine learning model and update the parameters of the temporary machine learning model. As the parameter update is repeated, the output result of the temporary machine learning model can become similar to the true value data. According to one embodiment, the parameters of the machine learning model can be updated according to a training objective function. For example, the processor can input multiple peak characteristic data into a temporary machine learning model (e.g., a random forest model) to update the parameters (e.g., parameters of a split function) so that the value of the training objective function is maximized. However, this is not limited thereto, and the training objective function may be designed differently. In this specification, a model during training may be referred to as a temporary machine learning model.

[0041] According to one embodiment, the processor can perform training by iteratively updating the parameters of the partitioning function in the machine learning model until the value according to the training objective function reaches the target value. The processor can terminate the training and determine the final parameters in response when the value according to the training objective function reaches the target value. Training of the temporary machine learning model can be completed upon determination of the final parameters. The temporary machine learning model that has completed training can be referred to as a machine learning model.

[0042] In step (150), the processor can output an identification result for the input strain using a machine learning model that has completed training. First, the processor can generate peak characteristic data for the input strain by applying the preprocessing operation of step (120) described above to the strain data regarding the input strain and performing the operation of step (130). The processor can output an identification result regarding the strain class of the strain by implementing the machine learning model that has completed training on the strain data regarding the input strain. For example, the processor can output an identification result by inputting the peak characteristic data of the input strain into the machine learning model.

[0043] A strain identification device according to one embodiment can distinguish MRSA and MSSA at the strain level by using data from 2,000 Da to 20,000 Da, which is a measured range used to distinguish bacterial species at the species level, by utilizing a machine learning model described below. Similarity-based algorithms mainly used for species-level identification cannot distinguish MRSA and MSSA, and there is a problem that distinguishing MRSA and MSSA based on the presence or absence of specific peaks and surrogate markers, such as the phenol-soluble modulin (PSM) series of amphiphilic alpha-peptide toxins, has low sensitivity; however, MRSA and MSSA can be distinguished by the identification method through this embodiment.

[0044] Figure 2 is a diagram showing the total mass spectra of MRSA and MSSA in one embodiment.

[0045] According to one embodiment, a MALDI-TOF mass spectrometer can ionize components such as proteins constituting the strain by radiating a pulsed laser onto a specific treated strain, and then apply a voltage to the ionized proteins to measure the time-of-flight (ToF) from the ion source to the detector. The measured time-of-flight can be determined based on the weight of the protein and the amount of ionized charge. Since the time-of-flight is inversely proportional to speed, it can be proportional to the mass-to-charge ratio of the strain protein. Therefore, the mass spectrometer can generate mass spectrum data of the strain using the time-of-flight. The unit of the mass-to-charge ratio may be m / z. The mass spectrum data may be expressed in a peak profile data format. For example, FIG. 2 illustrates mass spectrum data (210) of an MRSA strain and mass spectrum data (220) of an MSSA strain. The x-axis of the mass spectrum data (210, 220) may represent the mass-to-charge ratio, and the y-axis may represent the intensity of the mass spectrum. In the x-axis shown in Fig. 2, the unit of the mass-charge ratio may be Da (Dalton). For reference, the range of mass-charge ratios that can be analyzed by a MALDI-TOF mass spectrometer may be 2,000 Da to 20,000 Da, but is not limited thereto. Only a portion of the range is shown in Fig. 2.

[0046] Figure 3 is a diagram comparing the mass spectra of MRSA and MSSA in a specific mass range according to one embodiment.

[0047] The difference in mass spectra between strains MRSA and MSSA from 2,000 Da to 3,000 Da according to one embodiment can be shown as in Fig. 3. The similarity in mass spectrum patterns between MRSA and MSSA is relatively high compared to the similarity between other strains, making it difficult for existing similarity measurement algorithms to distinguish between MRSA and MSSA. Furthermore, techniques based on the presence or absence of specific peak markers and surrogate markers, such as the phenol-soluble modulin (PSM) family of amphiphilic alpha-peptide toxins, have low sensitivity, making it difficult to distinguish differences at the strain level, such as between MRSA and MSSA.

[0048] In contrast, a strain identification device according to one embodiment can more accurately identify MRSA strains and MSSA strains by preprocessing mass spectrum data and identifying strains from peak characteristic data using a pre-trained machine learning model. Furthermore, the strain identification device can identify MRSA strains by type of Staphylococcal Cassette Chromosome mec (SCCmec) of mobile genetic elements possessed by the strain.

[0049] FIG. 4 is a diagram showing a processed mass spectrum according to one embodiment.

[0050] Preprocessing operations may include smoothing, baseline correction, and intensity normalization.

[0051] For example, a processor can remove noise from mass spectrum data by flattening the mass spectrum data. Removing noise from mass spectrum data involves removing unwanted spikes, trends, and outliers from the data, and may involve removing noise using Savitzky-Golay filters, moving averages, moving medians, linear regression, and quadratic regression.

[0052] The processor can remove the baseline of the flattened mass spectrum data. The processor can adjust the baseline by removing low-frequency signals.

[0053] The processor can normalize the measured intensity of baseline-adjusted mass spectrum data. For example, Total Ion Current (TIC) normalization may represent an operation that adjusts the sum of the total spectral intensities to 1 to facilitate the analysis of mass spectrum data. The processor can change the distribution of the data by adjusting the intensity range of the mass spectrum data. For example, the processor can limit the maximum intensity of the mass spectrum data.

[0054] The processor can generate spectrum data (420) suitable for data analysis by removing noise through preprocessing mass spectrum data (410) obtained through a MALDI-TOF mass spectrometer.

[0055] Figure 6 is a graph illustrating a decision tree model according to one embodiment.

[0056] Decision tree models can be easy to interpret and capture the overall pattern of the data. However, they may lack generality due to a high risk of overfitting and sensitivity to small changes in the dataset. The accuracy of evaluating the data using a decision tree model composed of a training dataset can be 94.4%. However, the decision tree generated when the training dataset is expanded can differ significantly from the initial tree model.

[0057] Figure 7 is a graph illustrating a decision tree of a random forest model according to one embodiment.

[0058] According to one embodiment, the machine learning model can be implemented as a random forest model. The processor can train the machine learning model by updating the parameters of the machine learning model based on the random forest model. The random forest model is an ensemble model of multiple decision trees, and the parameters of each decision tree included in the random forest model can be learned independently by randomness. In the test phase, the processor can finally determine the strain class for the input monarch data by synthesizing the identification results based on the multiple decision trees included in the trained random forest model.

[0059] FIG. 8 illustrates an exemplary decision tree included in a random forest model. The decision tree may be a set of nodes and edges arranged in a hierarchical structure. Nodes are divided into internal nodes (1120, 1130, 1131) and terminal nodes (1140), and the processor can update the parameters of the nodes on a node-by-node basis. The parameters may include, for example, test parameters of all nodes and parameters at terminal nodes.

[0060] According to one embodiment, the processor inputs peak characteristic data to the root node (1110) and can optimize parameters for the terminal node and parameters of the node split function associated with the internal nodes (1120, 1130, 1131). For example, the processor can update the parameters of the tree so that the value of the objective function is maximized according to the temporary output produced by inputting training inputs to the random forest model. The parameters of the tree may include hyperparameters mtry and split rule. In the hyperparameters, mtry may represent the number of variables used in the split function (e.g., the order of the function).

[0061] FIG. 9 is a graph illustrating a random forest model including a plurality of decision trees according to one embodiment.

[0062] A random forest model may include multiple decision trees (1210, 1220, 1230), and the processor may train the multiple decision trees (1210, 1220, 1230) independently of each other. Therefore, the tree structures of the multiple decision trees (1210, 1220, 1230) may differ from each other, and the partitioning functions and parameters may also differ from each other. When the processor inputs peak characteristic data into the multiple decision trees (1210, 1220, 1230) during the testing and judgment process, it may output multiple judgment results in parallel through the multiple decision trees (1210, 1220, 1230).

[0063] The processor may output a final judgment result by considering all of the multiple judgment results. For example, the processor may output a final judgment result by averaging, multiplying, or a majority voting method of the judgment results. The processor may determine the final identification result based on a first identification result by a first decision tree (1210), a second identification result by a second decision tree (1220), and a third identification result by a third decision tree (1230). For example, each identification result may indicate the probability that the input strain belongs to each class. If the random forest model is trained based on training data consisting of two classes, each identification result may indicate the probability that the input strain is a specific strain (e.g., MRSA). For example, if the first identification result is 0.85, the second identification result is 0.9, and the third identification result is 0.98, the processor may determine the average of the identification results, 0.91, as the final identification result.

[0064] However, the determination of the final identification result is not limited to what is described above. For example, if a random forest model is trained based on training data composed of multiple classes, each identification result may indicate the probability that the input strain belongs to each strain class. For example, the identification results may consist of [probability of MRSA type II, probability of MRSA type III, probability of MRSA type IVA, probability of MSSA], and the processor may determine the final identification result using the corresponding values ​​of each identification result. Taking the majority method as an example, if the first identification result = [0.1, 0.1, 0.2, 0.6], the second identification result = [0.1, 0.1, 0.8, 0], and the third identification result = [0.1, 0.1, 0.1, 0.7], the processor may determine that the strain is MSSA because the probability of MSSA was highest in the first and third identification results.

[0065] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0066] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0067] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0068] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

Claims

Claim 1 A method for identifying microorganisms using learned deep learning, comprising: (a) a step in which a processor classifies a plurality of strain data from a strain database by strain and generates training data including reference strain information mapped to each mass spectrum data; (b) a step in which a processor preprocesses mass spectrum data from the generated training data to select data that meets the signal-to-noise ratio (SNR) criteria; (c) a step in which a processor extracts peak characteristic data from the preprocessed mass spectrum data and updates the parameters of a machine learning model using the same; and (d) a step in which a processor calculates a strain identification result based on the peak characteristic data of an input strain using the updated machine learning model.

Images