Estimation device, estimation method, and program

Abstract

To provide an estimation device, an estimation method and a program which can easily estimate applications of species, without requesting a user to have expert knowledge related to bioinformatics.SOLUTION: An estimation device includes a shaping part for acquiring text information indicating a base sequence of a nucleic acid of species as an estimation object of application, and shaping the text information to predetermined numbers of lines and columns, a conversion part for associating luminance values with each character of the shaped text information, and converting the text information into image information, and an estimation part for inputting the image information into a learned model, and thereby estimating the application of the species, in which the learned model is a machine learning model which has luminance values of the predetermined numbers of lines and columns, and is learned to output the application of the species when image information indicating the base sequence of the nucleic acid of the species is input.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an estimation device, an estimation method, and a program. [Background technology]

[0002] Conventionally, techniques for estimating the uses of industrial microorganisms such as koji mold have been known. For example, Non-Patent Document 1 discloses a technique for identifying the phylogenetic relationships between multiple koji mold strains and their uses by performing comparative genome analysis on these strains. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Watarai N, Yamamoto N, Sawada K, Yamada T (2019) Evolution of Aspergillus oryzae before and after domestication inferred by large-scale comparative genomic analysis. DNA Res 26: 465-472 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the above-mentioned conventional techniques require practitioners to have advanced specialized knowledge and know-how in bioinformatics, such as comparative genome analysis, and are not easily implemented by non-experts, such as starter koji makers who routinely develop applications for new strains. Furthermore, existing strain characterization methods require processes such as cultivation, koji production, enzyme activity measurement, toxicity testing, metabolite measurement, brewing tests, and sensory evaluation, which are time-consuming and labor-intensive for starter koji makers.

[0005] The present invention has been made in consideration of the above circumstances, and one of its objects is to provide an estimation device, an estimation method, and a program that enable users to easily estimate the uses of biological species without requiring specialized knowledge of bioinformatics. [Means for solving the problem]

[0006] An estimation device according to one aspect of the present invention includes a formatting unit that acquires text information representing the base sequence of nucleic acid of a biological species whose use is to be estimated and formats the text information into a predetermined number of rows and columns; a conversion unit that converts the text information into image information by associating a brightness value with each character of the formatted text information; and an estimation unit that estimates the use of the biological species by inputting the image information into a trained model, wherein the trained model is a machine learning model that has been trained to output the use of the biological species when image information representing the base sequence of nucleic acid of the biological species and having brightness values with the predetermined number of rows and columns is input.

[0007] The text information may include adenine, thymine, guanine, cytosine, and a gap representing a space as each character, and the conversion unit may convert the text information into image information by associating brightness values with the adenine, thymine, guanine, cytosine, and gaps included in the text information.

[0008] The conversion unit may convert the text information into grayscale image information by associating a brightness value with each character of the formatted text information.

[0009] The biological species may be koji mold, and the trained model may be trained to output the uses of the koji mold.

[0010] The trained model may be trained to output the position of the koji mold on a phylogenetic tree as the use of the koji mold.

[0011] In one aspect of the present invention, an estimation method involves a computer acquiring text information representing the base sequence of nucleic acid of a biological species whose use is to be estimated, shaping the text information into a predetermined number of rows and columns, associating a brightness value with each character of the shaped text information, converting the text information into image information, and inputting the image information into a trained model, thereby estimating the use of the biological species, wherein the trained model is a machine learning model trained to output the use of the biological species when image information having brightness values of the predetermined number of rows and columns and representing the base sequence of nucleic acid of the biological species is input.

[0012] A program according to one aspect of the present invention causes a computer to acquire text information representing the base sequence of nucleic acid of a biological species whose use is to be estimated, format the text information into a predetermined number of rows and columns, associate a brightness value with each character of the formatted text information, convert the text information into image information, and input the image information into a trained model to estimate the use of the biological species, wherein the trained model is a machine learning model that has been trained to output the use of the biological species when image information representing the base sequence of nucleic acid of the biological species and having brightness values of the predetermined number of rows and columns is input. [Effects of the Invention]

[0013] According to the present invention, it is possible to provide an estimation device, an estimation method, and a program that enable users to easily estimate the uses of biological species without requiring specialized knowledge of bioinformatics. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a diagram illustrating an example of a usage environment and configuration of an estimation device 100. FIG. [Figure 2] 10 is a diagram for explaining the shaping process of text information by the shaping unit 110. FIG. [Figure 3] 10 is a diagram for explaining the conversion process of the formatted text information 152 by the conversion unit 120. FIG. [Figure 4] 10 is a diagram for explaining the process of estimating the use of koji mold by the estimation unit 130. FIG. [Figure 5] This is a diagram for explaining an overview of the phylogenetic tree used to generate the trained model 156. [Figure 6] FIG. 10 is a diagram showing an example of the configuration of a usage association table 158. [Figure 7] 3 is a flowchart showing an example of the flow of processing executed by the estimation device 100. DETAILED DESCRIPTION OF THE INVENTION

[0015] [overview] An estimation device 100 according to an embodiment of the present invention will be described below with reference to the drawings. In the following embodiment, as an example, a case where the biological species for which a use is to be estimated is Aspergillus oryzae will be described. However, the present invention is not limited to cases where the biological species is Aspergillus oryzae, and can be more generally applied to biological species for which a phylogenetic tree, which will be described later, can be created.

[0016] FIG. 1 is a diagram illustrating an example of the usage environment and configuration of the estimation device 100. The estimation device 100 operates in cooperation with, for example, a next-generation sequencer 10 and a terminal device 20. The next-generation sequencer 10 performs genome analysis on a sample of Aspergillus oryzae, the use of which is to be estimated, and acquires genome sequence information representing the base sequence of the nucleic acid of the sample as text information. Upon acquiring the text information representing the genome sequence information, the next-generation sequencer 10 transmits the acquired text information to the estimation device 100 via a network NW. Alternatively, the text information acquired by the next-generation sequencer 10 may be manually transmitted to the estimation device 100 by a user of the estimation device 100 (for example, by connecting the next-generation sequencer 10 and the estimation device 100 via a universal serial bus (USB)). The next-generation sequencer 10 may be a known genome analysis device.

[0017] The terminal device 20 is a computer device such as a personal computer, smartphone, or tablet terminal. The terminal device 20 communicates with the estimation device 100 via the network NW, receives the estimation results of the uses of koji mold by the estimation device 100, and outputs them on a screen such as a display. This allows users of the estimation device 100 to understand the uses of koji mold. Here, users of the estimation device 100 are, for example, koji starter manufacturers and brewers who develop new strains or find uses for existing strains.

[0018] The estimation device 100 is a server device such as a web server. The estimation device 100 includes, for example, a shaping unit 110, a conversion unit 120, an estimation unit 130, and a storage unit 150. Each of the shaping unit 110, the conversion unit 120, and the estimation unit 130 is realized by, for example, a hardware processor such as a CPU (Central Processing Unit) executing a program (software). Some or all of these components may be realized by hardware (including circuitry) such as an LSI (Large Scale Integration), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a GPU (Graphics Processing Unit), or may be realized by a combination of software and hardware. The program may be stored in advance in a storage device (a storage device having a non-transitory storage medium) such as a hard disk drive (HDD) or flash memory, or may be stored in a removable storage medium (a non-transitory storage medium) such as a DVD or CD-ROM, and installed by inserting the storage medium into a drive device. The storage unit 150 is realized by a storage device such as a HDD, flash memory, or RAM (Random Access Memory). The storage unit 150 stores, for example, post-formatting text information 152, image information 154, a trained model 156, and an application association table 158.

[0019] 2 is a diagram illustrating the text information shaping process performed by the shaping unit 110. The shaping unit 110 receives text information representing the genome sequence information of the koji mold whose use is to be estimated from the next-generation sequencer 10, and shapes the text information into a predetermined number of rows and columns. The upper part of FIG. 2 shows the raw text information obtained from the next-generation sequencer 10, and the lower part of FIG. 2 shows the shaped text information obtained by shaping the obtained text information into a predetermined number of rows and columns.

[0020] In Figure 2, "A (a)" in the base sequence represents adenine, "T (t)" represents thymine, "G (g)" represents guanine, and "C (c)" represents cytosine. The distinction between uppercase and lowercase letters indicates the accuracy of the analysis by the next-generation sequencer 10; for each base, if the analysis accuracy is high, it is output as an uppercase letter, whereas if the analysis accuracy is low, it is output as a lowercase letter. This accuracy determination is based on a function pre-installed in the next-generation sequencer 10. Furthermore, although not shown in Figure 2, regions where the analysis by the next-generation sequencer 10 fails are output as gap regions, with blanks.

[0021] After shaping the text information into a predetermined number of rows and columns, the shaping unit 110 stores the shaped text information in the storage unit 150 as shaped text information 152. By shaping the text information into a predetermined number of rows and columns in this manner, the shaping unit 110 can generate a trained model 156 (described later) and perform estimation using the trained model 156.

[0022] 3 is a diagram illustrating the conversion process of the post-shaping text information by the conversion unit 120. The conversion unit 120 associates a brightness value with each character in the post-shaping text information 152, and converts the post-shaping text information 152 into image information 154. More specifically, for example, the conversion unit 120 converts the post-shaping text information 152 into image information 154 by associating adenine with 50, thymine with 100, guanine with 150, cytosine with 200, and gap regions with 255 in the base sequence using a 256-level grayscale, and stores the image information in the storage unit 150. This provides a grayscale image that represents the base sequence of the nucleic acid of the koji mold, the use of which is to be estimated.

[0023] Note that the correspondence between each character in the base sequence and a brightness value is merely an example, and each character in the base sequence may be associated with a different grayscale brightness value. In this way, by converting the shaped text information 152 into grayscale image information 154 instead of colorscale, the processing load for generating the trained model 156 and for estimation using the trained model 156 can be reduced.

[0024] Furthermore, in the above description, uppercase and lowercase letters in a base sequence are not distinguished from each other and are associated with the same brightness value. However, the present invention is not limited to such a configuration. For example, different brightness values may be associated with uppercase and lowercase letters, or brightness values may be associated only with uppercase letters, omitting lowercase letters with low accuracy. Furthermore, learning, as described below, may be performed by, for example, assigning a higher weight to uppercase letters with high accuracy (and conversely, assigning a lower weight to lowercase letters with low accuracy). Furthermore, for example, gap regions may not be assigned brightness values and may be deleted when converted to an image.

[0025] FIG. 4 is a diagram illustrating the process of estimating the use of koji mold by the estimation unit 130. The estimation unit 130 inputs image information 154 obtained by the conversion process by the conversion unit 120 into the trained model 156, thereby estimating the use of koji mold associated with the image information 154. Here, the trained model 156 is a machine learning model trained to output the use of the koji mold when image information having brightness values of the predetermined number of rows and columns and representing the base sequence of the nucleic acid of the koji mold is input. In FIG. 4, the trained model 156 is trained to output the position of the koji mold on the phylogenetic tree (TK-6 is used as an example in FIG. 4) as the use of the koji mold. The reason for such training is that strains that are close in position on the phylogenetic tree tend to have similar uses.

[0026] FIG. 5 is a diagram for explaining an overview of the phylogenetic tree used to generate the trained model 156. The phylogenetic tree shown in FIG. 5 is obtained by collecting a large number of koji mold samples and then performing comparative genome analysis, in which koji mold strains with similar genome structures are classified into the same clade CL (closely related group). As a conceptual diagram, FIG. 5 shows how strains are classified into clades CL1 to CL4. Strains belonging to the same clade tend to be close to each other on the phylogenetic tree and therefore have similar uses.

[0027] The trained model 156 is generated by training a machine learning model such as a convolutional neural network using, as training data, a combination of grayscale image information 154 corresponding to koji mold samples whose positions on the phylogenetic tree are known by genome analysis and the positions on the phylogenetic tree. Alternatively, the trained model 156 may be generated by clustering the image information 154 of each koji mold sample whose positions on the phylogenetic tree are known using an unsupervised learning method such as a self-organizing map. In this way, the estimation unit 130 can input image information 154 of the koji mold whose use is to be estimated into the trained model 156, thereby estimating the position of the koji mold on the phylogenetic tree.

[0028] FIG. 6 is a diagram showing an example of the configuration of the use association table 158. FIG. 6 shows, for example, a table in which positions on a phylogenetic tree and their uses are associated and stored in advance. The estimation unit 130 can identify the use of the koji mold whose use is to be estimated by referring to the use association table 158 using the position on the phylogenetic tree output by the trained model 156 as a key. For example, as shown in FIG. 4, when TK-6 is output from the trained model 156, the estimation unit 130 can identify "sake" as the use of TK-6 by referring to the use association table 158 using TK-6 as a key.

[0029] When the estimation unit 130 identifies the use of the koji mold whose use is to be estimated, it transmits the identified use to the terminal device 20 via the network NW and outputs it, for example, on a screen such as a display of the terminal device 20. This allows the user of the estimation device 100 to understand the use of the koji mold.

[0030] In the above embodiment, as an example, the trained model 156 is trained to take image information 154 of the koji mold whose use is to be estimated as input and output the position of the koji mold on the phylogenetic tree. However, the present invention is not limited to such a configuration. For example, the trained model 156 may be trained to take image information 154 as input and output the clade of the koji mold on the phylogenetic tree. Furthermore, for example, the trained model 156 may be trained to take image information 154 as input and output the use of the koji mold (e.g., sake, miso, soy sauce, etc.) as is.

[0031] Next, the flow of processing executed by the estimation device 100 will be described with reference to Fig. 7. Fig. 7 is a flowchart showing an example of the flow of processing executed by the estimation device 100. First, the shaping unit 110 acquires text information representing the base sequence of nucleic acid of the koji mold whose use is to be estimated from the next-generation sequencer 10 (step S100). Next, the shaping unit 110 acquires shaped text information 152 by shaping the acquired text information into a predetermined number of rows and columns (step S102).

[0032] The conversion unit 120 associates a brightness value with each character in the shaped text information 152 and converts the text information into image information 154 (step S104). Next, the estimation unit 130 inputs the converted image information 154 into the trained model 156 and estimates the position of the koji mold on the phylogenetic tree (step S106). Next, the estimation unit 130 outputs the estimation result to the terminal device 20 (step S108). This ends the processing of this flowchart.

[0033] According to the present embodiment described above, text information representing the base sequence of nucleic acid of a biological species for which use is to be estimated is acquired and shaped into a predetermined number of rows and columns, and each character of the shaped text information is converted into image information by associating it with a brightness value, and the image information is input into a trained model, thereby estimating the use of the biological species. This makes it possible to provide an estimation device, estimation method, and program that enable easy estimation of the use of a biological species without requiring the user to have specialized knowledge in bioinformatics.

[0034] The above describes the form for carrying out the present invention using an embodiment, but the present invention is not limited to such an embodiment, and various modifications and substitutions can be made within the scope that does not deviate from the gist of the present invention. [Explanation of symbols]

[0035] 10 Next-generation sequencer 20 Terminal equipment 100 Estimator 110 Plastic surgery department 120 Conversion Unit 130 Estimation part 150 Storage section 152 Formatted text information 154 Image Information 156 trained models 158 Usage mapping table

Claims

1. a formatting unit that acquires text information representing the base sequence of a nucleic acid of a biological species that is a target for use estimation, and formats the text information into a predetermined number of rows and columns; a conversion unit that converts the text information into image information by associating a luminance value with each character of the formatted text information; an estimation unit that estimates the purpose of the biological species by inputting the image information into a trained model; The trained model is a machine learning model trained to output the purpose of a biological species when image information having brightness values of the predetermined number of rows and columns and representing a base sequence of a nucleic acid of the biological species is input. Estimation device.

2. the text information includes adenine, thymine, guanine, cytosine, and a gap representing a space as each character; the conversion unit converts the text information into image information by associating adenine, thymine, guanine, cytosine, and gaps included in the text information with brightness values; The estimation device according to claim 1 .

3. the conversion unit associates a luminance value with each character of the formatted text information and converts the text information into grayscale image information. The estimation device according to claim 1 .

4. the biological species is Aspergillus oryzae, The trained model is trained to output the use of the koji mold. The estimation device according to claim 1 .

5. The trained model is trained to output the position of the koji mold on a phylogenetic tree as the use of the koji mold. The estimation device according to claim 4 .

6. The computer obtaining text information representing the base sequence of a nucleic acid of a biological species for which a use is to be estimated, and shaping the text information into a predetermined number of rows and columns; Associating a luminance value with each character of the formatted text information, and converting the text information into image information; The image information is input into a trained model to estimate the purpose of the biological species; The trained model is a machine learning model trained to output the purpose of a biological species when image information having brightness values of the predetermined number of rows and columns and representing a base sequence of a nucleic acid of the biological species is input. Estimation method.

7. On the computer, obtaining text information representing the base sequence of a nucleic acid of a biological species for which a use is to be estimated, and shaping the text information into a predetermined number of rows and columns; Associating a luminance value with each character of the formatted text information, and converting the text information into image information; The image information is input into a trained model to estimate the use of the biological species; The trained model is a machine learning model trained to output the purpose of a biological species when image information having brightness values of the predetermined number of rows and columns and representing a base sequence of a nucleic acid of the biological species is input. program.

Images