Modification analysis for molecule by signal-based method
The novel method for analyzing tRNA modifications by aligning and classifying series data using machine learning models addresses the challenge of high modification density, achieving precise and accurate analysis of tRNA modifications.
Patent Information
- Application Number
- PCT/JP2024/018330
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-30
- Filing Date
- 2024-05-17
- Publication Date
- 2024-12-05
AI Technical Summary
Current methods for analyzing tRNA modifications at the single-molecule level are not yet practical due to high modification density, leading to insufficient analysis accuracy with existing data analysis methods.
A novel method involving the acquisition of series data on signal values that change based on the type and modification state of each constituent unit, alignment with reference data, and analysis using machine learning models to classify and estimate modification states, even at high modification densities.
This approach allows for high-precision analysis of modification states without lowering accuracy, enabling the detection of multiple modifications simultaneously and identifying differences in modification states between molecules.
Smart Images

Figure JP2024018330_05122024_PF_FP_ABST
Abstract
Description
Signal-based molecular modification analysis
[0001] The present invention relates to a method for analyzing the modification state of a molecule by a signal-based method, and an analysis system, an analysis device, a program, and a diagnostic support method related thereto.
[0002] Proteins are synthesized by translation from mRNA templates. During this process, tRNA functions as an adapter molecule that decodes the 61 genetic codes (codons) encoded in mRNA and converts them into 20 amino acids. After transcription, tRNA undergoes various chemical modifications (tRNA modifications) to mature and function normally.
[0003] To date, approximately 150 types of RNA modifications have been discovered in various species, with approximately 80% of these found in tRNA. RNA modifications not only complement RNA functions but also contribute to regulating gene expression by adding entirely new functions. Various modifications are known to occur in the anticodon and its surrounding area of tRNA, and these tRNA modifications play an important role in accurate codon decoding and maintaining the reading frame during translation. Furthermore, many modifications are present in the core (body) of tRNA, and these are known to be primarily involved in tRNA folding and intracellular stability.
[0004] In recent years, it has become clear that the modification state of tRNA changes dynamically in response to the growth environment and various external stresses. To correctly understand the function of tRNA modifications, it is necessary to accurately and quickly map the location of each RNA modification on each tRNA molecule. While mapping by high-performance liquid chromatography-mass spectrometry (LC / MS) allows for highly sensitive measurements, it requires the isolation and purification of the corresponding RNA, making it difficult to simultaneously map multiple samples or perform periodic measurements.
[0005] Single-molecule sequencers capable of directly analyzing RNA have been developed as next-generation sequencers (NGS). For example, nanopore sequencers are single-molecule sequencers that can directly analyze sequences by measuring the ion current signal of single-stranded DNA or RNA passing through a nanopore made of membrane proteins. Modified bases in RNA strands exhibit characteristic current values when passing through the nanopore, unlike unmodified bases, which may enable simultaneous detection of all modifications.
[0006] In recent years, research into the detection of RNA modifications using nanopore sequencing, targeting mRNA and rRNA, has been rapidly developing (Non-Patent Documents 1-4). Because mRNA and rRNA do not exhibit high RNA modifications, RNA modification analysis can be performed using existing base callers. Analysis of tRNA using techniques similar to those for mRNA and rRNA has also been reported (Non-Patent Documents 5-6).
[0007] Liu, H. et al. Accurate detection of m6A RNA modifications in native RNA sequences. Nature communications 10, 4079, doi:10.1038 / s41467-019-11713-9 (2019).Stoiber, M. et al. De novo Identification of DNA Modifications Enabled by Genome-Guided Nanopore Signal Processing. bioRxiv, 094672, doi:10.1101 / 094672 (2017).Leger, A. et al. RNA modifications detection by comparative Nanopore direct RNA sequencing. Nature communications 12, 7198, doi:10.1038 / s41467-021-27393-3 (2021).Smith, A. M. et al. Reading canonical and modified nucleobases in 16S ribosomal RNA using nanopore native RNA sequencing. PLoS One 14, e0216709, doi:10.1371 / journal.pone.0216709 (2019).Thomas, N. K. et al. Direct Nanopore Sequencing of Individual Full Length tRNA Strands. ACS Nano 15, 16642-16653, doi:10.1021 / acsnano.1c06488 (2021).Lucas, M. C. et al. Quantitative analysis of tRNA abundance and modifications by nanopore RNA sequencing. Nat Biotechnol, doi:10.1038 / s41587-023-01743-6 (2023).
[0008] However, although detection of some tRNA modifications has been successful to date, methods for analyzing tRNA modifications at the single molecule level have not yet reached the practical stage. One reason for this is the high density of diverse modifications present on tRNA. The data analysis methods described in the above prior art documents do not provide sufficient analytical accuracy when diverse modifications are present at high density, so there is a need to develop completely new data analysis methods that are completely different from those used so far.
[0009] Therefore, an object of the present invention is to provide a novel method for analyzing the modification state of a structural unit in a molecule in which one or more types of structural units are bonded together.
[0010] The analytical method of this embodiment is a method for analyzing the modification state of constituent units in a molecule in which multiple types of constituent units are bonded together, and includes the steps of: acquiring, for each of multiple target molecules, series data of signal values that vary depending on the type and modification state of each target molecule obtained by scanning the molecule; aligning each of the series data acquired from the multiple target molecules with reference data to thereby align the multiple series data with each other; and analyzing the modification states of at least some of the multiple target molecules based on the aligned series data.
[0011] This analytical method aligns and compares series of data of signal values that vary depending on the type and modification state of each structural unit obtained by scanning a molecule, and therefore, even when analyzing a molecule in which a variety of modifications are present at high density, the analytical accuracy is not reduced as in conventional analytical methods, and the modification state can be analyzed with high accuracy. Note that the multiple analyte molecules may include one or more types of molecules. The multiple analyte molecules may include multiple molecules of one type, or may include multiple molecules of each type, each of which may include two or more types of molecules.
[0012] Preferably, the above method further includes generating the reference data by mixing a plurality of series of data including the acquired series of data. In this aspect, the reference data used when aligning the plurality of series of data is obtained by actually measuring the series of data and mixing the plurality of series of data, so that reference data with higher alignment accuracy for the plurality of series of data can be generated. Note that in this aspect, it is not necessary to use all of the plurality of acquired series of data for mixing; it is sufficient to mix at least a portion of the acquired series of data (e.g., one-third or more, half or more, two-thirds or more, etc. of the total number of series of data).
[0013] In the above-described embodiment, the plurality of series data to be mixed preferably includes two or more types of series data obtained or estimated from two or more types of molecules having the same structural unit sequence as at least one of the target molecules but with different modification states of the structural units. In this embodiment, the reference data is obtained by mixing the series data based on the two or more types of molecules with different modification states, thereby enabling the creation of reference data with higher alignment accuracy for the plurality of series data.
[0014] In the above-described method, analyzing the modification states of the plurality of analyte molecules may include: calculating a distance between a distribution of signal values at a predetermined point in the plurality of series of data for the first molecule and a distribution of signal values at a point in the plurality of series of data for the second molecule associated with the predetermined point, using the plurality of series of data obtained or estimated from and aligned with a first molecule whose constituent units have a first modification state and the plurality of series of data obtained or estimated from and aligned with a second molecule whose constituent units have a second modification state; and estimating, based on the calculated distance, portions whose modification states differ between the first molecule and the second molecule. This aspect makes it possible to identify portions whose modification states differ between two types of series of data.
[0015] In the above-described method, analyzing the modification states of the plurality of analyte molecules may include estimating the modification state of a first building block in a third molecule, the modification state of which is unknown, using a machine learning model trained with training data including the sequence data obtained from a first molecule, the first building block being in a first modification state at a first position in the sequence of the building blocks of the analyte molecule, and the sequence data obtained from a second molecule, the first building block being in a second modification state. This aspect makes it possible to estimate the modification state of a building block at a specific position in a molecule, the modification state of which is unknown.
[0016] In the above method, analyzing the modification states of the plurality of analyte molecules may include classifying the aligned plurality of series of data into two or more clusters using an unsupervised machine learning model. This embodiment allows molecules with unknown modification states to be classified into two or more clusters.
[0017] In the above aspect, the classifying into clusters may include calculating a similarity between each of the aligned sequence data and classifying the aligned sequence data into two or more clusters based on the similarity. This aspect enables clustering with a smaller amount of calculation than when the sequence data is used directly for clustering.
[0018] In the above method, the molecule may be a nucleic acid. Preferably, the plurality of analyte molecules includes at least two of a nucleic acid derived from a wild-type strain, a nucleic acid with no modification, and a nucleic acid derived from a mutant strain in which the modification state of a specific base has changed from that of the wild-type strain. Preferably, the plurality of analyte molecules includes all of a nucleic acid derived from a wild-type strain, a nucleic acid with no modification, and a nucleic acid derived from a mutant strain in which the modification state of a specific base has changed from that of the wild-type strain. This aspect allows modified portions in nucleic acids derived from wild-type strains to be identified on the lineage data.
[0019] In the above aspect, it is preferable that the plurality of analyte molecules at least include nucleic acids derived from the wild-type strain, and that in the nucleic acids derived from the wild-type strain, the ratio of the number of modified bases to the total number of bases constituting the nucleic acid is 4% or more. This aspect can effectively achieve the effect of the above method, namely, high analytical accuracy when diverse modifications are present at high density.
[0020] In the above aspect, the method may further include calculating redundancy of the sequence data relative to the reference data when aligning each of the plurality of sequence data with the reference data, and analyzing the modification state based on the calculated redundancy. In this aspect, the speed at which the molecule is scanned is expressed by the redundancy of the sequence data. Therefore, this aspect makes it possible to analyze the modification state of the molecular building blocks based on changes in the speed at which the molecule is scanned.
[0021] The diagnostic support method according to this embodiment is a method for supporting the diagnosis of a disease in which a disorder occurs in the modification state of a structural unit in a molecule in which multiple structural units of one or more types are linked together, and includes analyzing a specimen derived from a patient containing the molecule and a specimen derived from a healthy subject containing the molecule by the above-described analytical method, and analyzing the difference in the modification state of the structural units in the molecule between the specimen derived from the patient and the specimen derived from the healthy subject. This diagnostic support method can simplify the diagnosis of the disease.
[0022] In this specification, the term "system" refers to a system that is configured from multiple devices, etc., and that provides specific functions to users. For example, the system may be configured from a server device, a cloud computing system, an ASP (Application Service Provider), a client-server model, etc., but is not limited to these.
[0023] Furthermore, the term "unit" does not necessarily refer to a physical component, but also includes cases where the functions of the component are realized by software. Furthermore, the functions of one component may be realized by two or more physical components, or the functions of two or more components may be realized by one physical component.
[0024] According to the present invention, it is possible to provide a novel method for analyzing the modification state of a structural unit in a molecule in which a plurality of structural units of one or more types are bonded.
[0025] 1 is a schematic diagram of an analysis system according to the present embodiment; FIG. 2 is a functional block diagram of an analysis device according to the present embodiment; FIG. 3 is a diagram showing the physical configuration of an analysis device according to the present embodiment; FIG. 4 is a flowchart of processing in an analysis device according to the present embodiment; FIG. 5 is an example of an analysis result in an analysis method according to the present embodiment; FIG. 6 is an example of an analysis method according to the present embodiment; FIG. 7 shows the design of an adapter when measuring tRNA in an example using a nanopore sequencer; (A) shows an outline of the design of the adapter; (B) shows the design of the 1st adapter; Asn (A) E. coli tRNA modification was detected. Asn (B) tRNA derived from the wild-type strain. Asn tRNA derived from in vitro transcripts and each deletion strain Asn The analysis method according to this embodiment was performed on Escherichia coli tRNA Gly3 This is an example of creating a machine learning model for the modification of Escherichia coli tRNA Pro3 1 shows an example of clustering performed by unsupervised machine learning on modifications of human mt-tRNA. 2 shows an example of a method for calculating redundancy of sequence data relative to reference data in the analysis method according to the embodiment. Leu(UUR) (A) Human mt-tRNA Leu(UUR) (B) Current values and durations of tRNA derived from Hela cells and tRNA derived from a modification enzyme-deficient strain, as well as the Kullback-Leibler (KL) distance of their distributions are shown. (C) Distributions of current values and durations at the points indicated by triangles in (B). (D) U and τm in the series data derived from Hela cells, the series data derived from a modification enzyme-deficient strain, and the series data derived from myoblasts of a MELAS patient. 5The distribution and estimated abundance ratio of U in yeast tRNA by the analysis method according to the embodiment are shown. Gln1a (A) mcm 5 s 2 (B) The current value and duration of tRNA derived from a wild-type strain and tRNA derived from a modification enzyme-deficient strain (Δelp3Δncs6), as well as the Kullback-Leibler (KL) distance of their distribution are shown. (C) The distribution of current value and duration at the peak position of (B) is shown for the series data derived from a wild-type strain and the data derived from modification enzyme-deficient strains (Δelp3Δncs6 and Δncs6). Gln1a The distribution of each modification state and the estimated abundance ratio are shown for the wild-type strain and the modification enzyme-deficient strains (Δelp3Δncs6 and Δncs6).
[0026] Hereinafter, an embodiment of the present invention (hereinafter referred to as "the present embodiment") will be described in detail with reference to the drawings, but the present invention is not limited to this embodiment and various modifications are possible without departing from the spirit of the present invention. In the following description of the drawings, the same or similar parts are denoted by the same or similar reference numerals.
[0027] [Analysis System] Fig. 1 is a diagram showing the general configuration of an analysis system 1 according to this embodiment. The analysis system 1 includes a measurement device 20 and an analysis device 10 that analyzes the results of measurement by the measurement device 20. In the analysis system 1, the measurement device 20 measures an analyte molecule, and the analysis device 10 analyzes the measurement results, thereby analyzing the modification state of the analyte molecule.
[0028] In the analysis system 1, the analysis device 10 is connected to the measurement device 20 via a communication network N. Here, the communication network N may be a wired or wireless communication network. The analysis device 10 does not necessarily have to be an independent device from the measurement device 20, but may be configured integrally with the measurement device 20. In this case, the measurement device may also function as the analysis device 10 by executing an analysis program installed in the measurement device.
[0029] The analysis system 1 according to this embodiment analyzes the modification state of structural units in molecules (hereinafter simply referred to as "molecules") in which multiple structural units of one or more types are bonded together. The molecules analyzed by the analysis system 1 are preferably biopolymers. Examples of biopolymers include peptides, proteins, nucleic acids, polysaccharides, and combinations thereof (e.g., lipoproteins, glycoproteins, nucleoproteins, sugar chains, etc.). Here, the structural units of peptides and proteins are amino acids, the structural units of nucleic acids are nucleotides, and the structural units of polysaccharides are monosaccharides. It is preferable that the biomolecule be capable of forming a single chain.
[0030] As used herein, the term "modification" of a molecule or a constituent unit thereof means that the molecule or a constituent unit thereof is altered from its normal state, e.g., natural or wild-type, including, for example, the addition of an additional group not normally present, the deletion of a group normally present, or the substitution of an atom or atomic group normally present with another atom or atomic group. For example, a modification of an amino acid may be a modification of the side chain of the amino acid. Examples of such amino acid modifications include post-translational modifications such as phosphorylation, methylation, acetylation, and glycosylation. A modification of a nucleotide may be a modification of the sugar moiety of a nucleoside and / or a modification of the base moiety of a nucleoside. A modification of a monosaccharide may be a modification of the hydrocarbon chain. Modifications of nucleotides, etc., include post-transcriptional modifications of RNA molecules, epigenetic modifications of DNA molecules or histone proteins that coat DNA molecules, etc. Examples of sugar chain modifications include phosphorylation, methylation, acetylation, glycosylation, sulfonation, etc.
[0031] The molecules for which the analysis system 1 analyzes the modification state may be any biomolecule capable of being single-stranded, but nucleic acids are preferred, and tRNA is particularly preferred. tRNA often has a variety of modifications present at high density, making analysis using the analysis system 1 according to this embodiment particularly effective. The molecules for which the analysis system 1 analyzes the modification state are preferably molecules derived from wild-type strains, in which the ratio of the number of modified structural units to the total number of structural units constituting the molecule is 4% or more, 5% or more, 6% or more, 7% or more, 8% or more, 9% or more, or 10% or more. For modifications in nucleic acids, see Boccaletto, P. et al. MODOMICS: a database of RNA modification pathways. 2021 update. Nucleic Acids Res 50, D231-D235, doi:10.1093 / nar / gkab1083 (2022), for example.
[0032] The analysis device 10 according to this embodiment analyzes measurement results using a method completely different from that of conventional analysis devices, and analyzes the modification state of the target molecule. For example, conventional analysis devices for analyzing the modification state of nucleic acids estimate base sequences from measured current signals by base calling, align the estimated base sequences based on similarity to a reference sequence, and obtain feature quantities for each base. Because modified bases exhibit different current values than unmodified bases, the accuracy of mapping to the reference sequence based on similarity is low for modified bases. When analyzing the modification state of mRNA and rRNA, because RNA modifications are rare and most bases are unmodified, accurate mapping of the portions other than modified bases ultimately ensures accurate mapping of modified bases. However, in cases such as tRNA, where there are many types and numbers of RNA modifications and multiple modified bases are densely packed in a short region, such existing strategies cannot be applied.
[0033] On the other hand, the analysis device 10 according to the present embodiment does not estimate constituent units from signal values as in conventional analysis devices, but aligns multiple measured series data and performs signal-based analysis. Therefore, alignment can be performed with high accuracy even when a variety of modifications are present at high density, thereby enabling highly accurate analysis.
[0034] The measurement device 20 and the analysis device 10 will be described in detail below, and the processing flow in the analysis system 1 according to this embodiment will be explained.
[0035] (Measuring Device) The measuring device 20 is a device that measures molecules each composed of multiple bonds of one or more types of structural units to obtain sequence data. To obtain the sequence data, the measuring device 20 scans the molecule and obtains signal values that vary depending on the type and modification state of each structural unit. The sequence data obtained by the measuring device 20 is typically, but not necessarily, time-series data.
[0036] The measuring device 20 measures a plurality of molecules to be analyzed. Therefore, a plurality of series of data can be obtained by measuring the molecules to be analyzed using the measuring device 20. The measuring device 20 may measure a plurality of molecules of one type, or may measure a plurality of molecules of two or more types.
[0037] An example of the measurement device 20 is a nanopore sequencer. A nanopore sequencer measures the current value when a measurement molecule passes through a nanopore made of a membrane protein or the like, which is large enough for a single measurement molecule to pass through, thereby obtaining signal values that reflect the type and modification state of each constituent unit of the measurement molecule. The sequence data obtained by the nanopore sequencer is time-series data. For example, when a molecule passes through a nanopore from one end to the other, a signal value corresponding to the molecular portion passing through the nanopore (i.e., a portion consisting of one or more constituent units) is obtained, and time-series data is output as the molecular portion moves from one end to the other.
[0038] A commercially available measuring device may be used as the measuring device 20. For example, a nanopore sequencer from ONT (Oxford nanopore technologies) may be used.
[0039] (Analysis Device) The analysis device 10 analyzes the series data measured by the measurement device 20 and analyzes the modification state of the measured molecules. Fig. 2 is a functional block diagram of the analysis device 10. The analysis device 10 includes an acquisition unit 11, a reference data generation unit 12, an alignment unit 13, and an analysis unit 14. Note that the analysis device 10 does not necessarily have to include the reference data generation unit 12, for example, and may have another functional configuration not shown in Fig. 2.
[0040] The acquiring unit 11 acquires the sequence data measured by the measurement device 20. The acquiring unit 11 may be realized by the communication unit 10d described below. The acquiring unit 11 may acquire the sequence data measured by the measurement device 20 for each measurement, or may acquire multiple sequence data obtained by multiple measurements all at once.
[0041] The reference data generation unit 12 generates reference data to be used for alignment in the alignment unit 13. The reference data generation unit 12 generates reference data that can be aligned with each of the multiple series of data measured by the measurement device 20. The reference data may include, for example, any of the multiple series of data measured by the measurement device 20, or may include series of data estimated from the sequence information of the molecules measured by the measurement device 20. The reference data may be, for example, any of the multiple series of data measured by the measurement device 20, or may be series of data estimated from the sequence information of the molecules measured by the measurement device 20, or may be series of data obtained by mixing two or more of the multiple series of data measured by the measurement device 20 and, optionally, series of data estimated from the sequence information of the molecules measured by the measurement device 20. The reference data generation unit 12 may be realized by a program stored in RAM 10b or ROM 10c (described below) and executed by the CPU 10a.
[0042] The alignment unit 13 aligns the multiple sequential data sets acquired by the acquisition unit 11 based on the reference data generated by the reference data generation unit 12. In this specification, "aligning" sequential data sets means aligning the overall lengths of sequential data sets that are difficult to compare directly because the lengths and numbers of corresponding data points differ, even though the order of each data point corresponds to each other, by associating each data point with each other while maintaining the order of each data point. Taking the alignment of time series data as an example, this means identifying corresponding data points for two sequential data sets and adjusting the time of each data point so that the times of the corresponding data points are the same, thereby aligning the overall lengths of the data sets.
[0043] The alignment unit 13 may be realized by an input unit 10e (described later) and a program stored in a RAM 10b or a ROM 10c (described later) and executed by a CPU 10a. The alignment unit 13 may perform alignment using, for example, dynamic programming, or may perform alignment using a dynamic time warping (DTW) method or the like. Although there are no particular limitations on the DTW, soft DTW is preferable. Furthermore, the sequence data obtained by the measurement device 20 may be interpreted as data undergoing state transitions according to a hidden Markov model, and alignment may be performed using a Viterbi algorithm.
[0044] When aligning multiple pieces of sequential data based on the reference data, the alignment unit 13 may calculate the redundancy of the sequential data with respect to the reference data. Here, the redundancy of the sequential data with respect to the reference data refers to the length of the portion of the sequential data that is associated with a specific portion of the reference data (the number of data points). For example, when comparing a case in which two or more points of sequential data are associated with one point of reference data with a case in which one point of sequential data is associated with two or more points of reference data, it can be said that the former has higher redundancy of the sequential data than the latter.
[0045] The redundancy of the sequential data relative to the reference data is calculated locally for each part of the reference data and the sequential data, since the sequential data is associated with a specific part of the reference data. A specific example of a method for calculating the redundancy of the sequential data will be described later.
[0046] The analysis unit 14 analyzes the modification state of molecules corresponding to at least a portion of the acquired sequence data based on the plurality of sequence data aligned by the alignment unit 13. Because the plurality of sequence data acquired by the acquisition unit 11 is aligned by the alignment unit 13, the analysis unit 14 can analyze the modification state with high accuracy. Specifically, the analysis unit 14 detects modifications based on changes in signal values and / or changes in redundancy using the aligned plurality of sequence data. Details of the analysis method in the analysis unit 14 will be described later. The analysis unit 14 may be realized by an input unit 10e (described later) and a program stored in RAM 10b or ROM 10c (described later) and executed by the CPU 10a.
[0047] The analysis device 10 may output the analysis results obtained by the analysis unit 14 in an appropriate format. For example, the analysis device 10 may display the analysis results on a display unit 10f (described later) or transmit the analysis results to another terminal via a communication unit 10d. The analysis results may be in the form of a three-dimensional or two-dimensional graph, numerical data, or the modification state of a structural unit in a molecule.
[0048] FIG. 3 is a diagram showing the physical configuration of the analysis device 10. The analysis device 10 includes a CPU (Central Processing Unit) 10a, which corresponds to a processor; a RAM (Random Access Memory) 10b and a ROM (Read Only Memory) 10c, which correspond to storage units; a communication unit 10d; an input unit 10e; and a display unit 10f. These components are connected via a bus so that they can transmit and receive data to and from each other. While this example describes a case in which the analysis device 10 is configured by a single computer, the analysis device 10 may also be realized by combining multiple computers. The configuration shown in FIG. 3 is merely an example, and the analysis device 10 may have other components or may not have some of these components.
[0049] The CPU 10a is a control unit that controls the execution of programs stored in the RAM 10b or the ROM 10c and performs data calculations and processing. The CPU 10a is a calculation unit that executes a program (analysis program) that analyzes the series data acquired by the measurement device 20. The CPU 10a receives various data from the input unit 10e and the communication unit 10d, and displays the results of calculations on the data on the display unit 10f or stores them in the RAM 10b or the ROM 10c.
[0050] The RAM 10b is a rewritable storage device and may be implemented, for example, by a semiconductor memory device. The RAM 10b may store an analysis program executed by the CPU 10a. Note that these are merely examples, and the RAM 10b may store other data.
[0051] The ROM 10c is a memory from which data can be read, and may be configured with, for example, a semiconductor memory element. The ROM 10c may store, for example, data that is not rewritten.
[0052] The communication unit 10d is an interface that connects the analysis device 10 to other devices, and may be connected to a communication network N such as the Internet.
[0053] The input unit 10e receives data input from a user and may include, for example, a keyboard and a touch panel.
[0054] The display unit 10f visually displays the results of calculations performed by the CPU 10a and may be configured with, for example, an LCD (Liquid Crystal Display). The display unit 10f may display the results of analysis performed by the analysis unit 14.
[0055] The analysis program may be provided by being stored in a computer-readable storage medium such as RAM 10b or ROM 10c, or may be provided via a communication network connected by communication unit 10d. In analysis device 10, CPU 10a executes the analysis program to realize the operations of acquisition unit 11, reference data generation unit 12, alignment unit 13, and analysis unit 14 described with reference to FIG. 2. Note that these physical configurations are merely examples and do not necessarily have to be independent configurations. For example, analysis device 10 may include an LSI (Large-Scale Integration) in which CPU 10a is integrated with RAM 10b and / or ROM 10c.
[0056] (Processing Flow) <Overview> Fig. 4 is a flowchart of processing in the analysis device 10 according to this embodiment. First, the acquisition unit 11 acquires multiple pieces of sequence data from the measurement device 20 (S10). Prior to step S10, the measurement device 20 measures one or more types of analyte molecules and measures the sequence data. When there are two or more types of analyte molecules, the two or more types of analyte molecules have the same sequence of constituent units but differ in the modification state of the constituent units. For example, when the analyte molecules are nucleic acids, the measurement device 20 measures two or more types of nucleic acids that have the same base sequence but differ in the modification state.
[0057] When the measuring device 20 measures two or more types of analyte molecules, the two or more types of analyte molecules may be measured one by one to obtain multiple series data labeled for each type, or the two or more types of analyte molecules may be measured all at once to obtain multiple series data without labels.
[0058] Next, the reference data generating unit 12 generates reference data for aligning the sequence data acquired in step S10 (S11). The reference data to be generated is not particularly limited as long as it can be aligned with each of the acquired sequence data.
[0059] In step S11, the reference data generator 12 may generate reference data using, for example, one of the following methods: (1) A method of selecting one of the sequential data acquired in step S10 as reference data. In this method, reference data may be selected randomly from the sequential data acquired in step S10, or may be selected based on the length of the sequential data. For example, the central sequential data when the acquired sequential data is sorted in order of length may be selected as reference data.
[0060] (2) A method for generating reference data based on a known sequence of the constituent units of a molecule to be analyzed. This method generates virtual sequence data predicted from the sequence of the constituent units of the molecule to be analyzed, and uses this as reference data. For example, when the molecule to be analyzed is a nucleic acid, software can be used to virtually predict the current value generated by a nanopore sequencer based on the base sequence. Note that reference data can be generated using this method only when the correspondence between the sequence data and the measured signal value is known. For example, when the molecule to be analyzed is a nucleic acid, sequence data for unmodified nucleic acids can be predicted using known methods.
[0061] (3) A method for generating reference data by mixing at least a portion of the sequence data obtained in step S10. In this method, multiple sequence data obtained from one type of analyte molecule may be mixed, or multiple sequence data obtained from two or more types of analyte molecules may be mixed. Mixing multiple sequence data may be performed using an algorithm that averages multiple sequence data. Examples of such algorithms include DTW barycenter averaging (DBA). Furthermore, when the sequence data obtained by the measurement device 20 is interpreted as data that undergoes state transitions according to a hidden Markov model, the parameters of the hidden Markov model may be estimated using the Baum-Welch algorithm.
[0062] In this method, it is preferable to mix two or more types of sequence data obtained from two or more types of molecules that have the same sequence of building blocks but differ in the modification state of the building blocks. The reference data obtained in this manner has the average characteristics of the two or more target molecules, and therefore tends to be able to be aligned with multiple sequence data with high accuracy. Furthermore, in this method, it is more preferable to mix two or more types of sequence data obtained from at least two types of molecules derived from a wild-type strain, molecules without modifications, and molecules derived from a mutant strain in which the modification state of a specific base has been changed from that of the wild-type strain. It is even more preferable to mix two types of sequence data obtained from molecules derived from a wild-type strain and molecules without modifications. It is not necessary to use all of the sequence data obtained in step S10 for mixing; it is sufficient to mix at least a portion of the acquired sequence data (e.g., at least 1 / 3, at least half, at least 2 / 3, etc.)
[0063] (4) When the sequence of the constituent units of the analyte molecule is known, virtual sequence data is generated based on the sequence in a manner similar to (2) above, and the virtual sequence data is then mixed with at least a portion of the sequence data obtained in step S10 in a manner similar to (3) above to generate reference data. In this method, the molecule from which virtual sequence data is generated may be an unmodified molecule. The sequence data mixed with the virtual sequence data may be one or more types of sequence data obtained in step S10.
[0064] Among the above methods (1) to (4), from the viewpoint of enabling highly accurate alignment in step S12 described below, the above methods (3) and (4) are preferred, and the above method (3) is more preferred. In the above method (3), a method of mixing two or more types of series data obtained from two or more types of molecules having the same sequence of constituent units but different modification states of the constituent units is particularly preferred.
[0065] Next, the alignment unit 13 aligns the plurality of sequence data acquired in step S10 based on the reference data generated in step S11 (S12). In step S12, the plurality of sequence data are aligned using the reference data as a reference. More specifically, by aligning each of the plurality of sequence data with the reference data, the lengths of the plurality of sequence data correspond to the lengths of the reference data, making it easier to compare the sequence data. The alignment can be performed as described above.
[0066] In step S12, in addition to the plurality of series data acquired in step S10, the virtual series data generated in step S11 and the series data stored in RAM 10b or ROM 10c may be aligned with the reference data.
[0067] Furthermore, redundancy of the sequence data with respect to the reference data may be calculated when aligning the plurality of sequence data based on the reference data in step 12. The calculation of redundancy will be described with reference to FIG.
[0068] Fig. 11(A) is a conceptual diagram of aligning reference data and sequence data. In Fig. 11(A), each point in the reference data is associated with each point in the sequence data, and this association is represented by a line segment connecting the points. The data shown in Fig. 11(A) conceptually shows, for example, a portion where four points in the sequence data are associated with one point in the reference data, or a portion where three points in the reference data are associated with one point in the sequence data. Furthermore, the value indicating how many points in the sequence data are associated with one point in the reference data is shown as an "insertion number," and the value indicating how many points in the reference data are associated with one point in the sequence data is shown as a "deletion number."
[0069] By calculating these Insertion number and Deletion number, it is possible to calculate the redundancy of the sequential data relative to the reference data. The larger the Deletion number corresponding to each point in the sequential data, the lower the redundancy of the sequential data relative to the reference data (conversely, the more redundant the reference data is with respect to the sequential data). The larger the Insertion number corresponding to a point in the reference data associated with each point in the sequential data, the higher the redundancy of the sequential data relative to the reference data.
[0070] Next, referring to Fig. 11(B), an example of calculating the redundancy of sequential data with respect to reference data using a correspondence matrix between reference data and sequential data to be aligned will be described. The correspondence matrix can be calculated, for example, by a soft DTW algorithm using the reference data and sequential data.
[0071] FIG. 11B shows a correspondence matrix in which the strength of the correspondence between each point of the reference data and each point of the aligned sequence data is probabilistically expressed as a value between 0 and 1. The matrix is expressed as P=(p ij ) (i = integers from 1 to n (number of data included in the sequence data); j = integers from 1 to m (number of data included in the reference data)). For example, if the first point in the reference data is associated with the first and second points in the sequence data, the elements in the first and second columns of the first row of matrix P are 1, and the other elements are 0. In the correspondence matrix of FIG. 11(B), areas with probabilities closer to 1 are shown in darker colors.
[0072] Using such a matrix P, the deletion number at the ith point of the sequence data can be expressed by the following formula (I), and the insertion number at the jth point of the reference data can be expressed by the following formula (II).
[0073]
[0074] Furthermore, by calculating a weighted average of the deletion numbers of the sequential data using the following formula (III) and matching the length of the deletion number with the length of the reference data, it is possible to calculate a Duration that reflects both the deletion number and the insertion number using the following formula (IV). Duration is an index that indicates how locally redundant a point of reference data associated with the jth point of reference data is with respect to the reference data. When Duration is 0, this means that the sequential data and the reference data are associated one-to-one, when Duration is less than 0, this means that the reference data is more redundant than the sequential data, and when Duration is greater than 0, this means that the sequential data is more redundant than the reference data.
[0075]
[0076] By using the above formulas (I) to (IV), it is possible to display the local redundancy of the sequence data and the reference data as a graph, with the sequence data and the reference data corresponding to the horizontal axis, as shown in FIG. 11(B).
[0077] 4 , after step S12, the analysis unit 14 analyzes the modification state of at least a portion of the target molecule based on the plurality of series data aligned in step S12 (S13). Since the series data are aligned in step S12, changes in signal values due to the type of target molecule and / or changes in the redundancy of the series data relative to the reference data can be easily detected statistically in step S13. An example of step S13 is described below.
[0078] Analyzing the modification state of at least some of the molecules to be analyzed based on aligned multiple series data includes analyzing the modification state by comparing the magnitude and distribution of signal values of the multiple series data, analyzing the modification state by comparing the amount and distribution of redundancy of the multiple series data relative to reference data, and analyzing the modification state using a machine learning model that uses aligned multiple series data as input values, as described in detail below.
[0079] Modification State Analysis Step (First Aspect) Step S13 may include, for example, calculating a distance between a distribution of signal values at a predetermined point in the plurality of sequence data for the first molecule and a distribution of signal values at a point in the plurality of sequence data for the second molecule associated with the predetermined point, using a plurality of sequence data obtained and aligned from a first molecule whose constituent units have a first modification state and a plurality of sequence data obtained and aligned from a second molecule whose constituent units have a second modification state, and estimating a portion of the first molecule whose modification state differs from that of the second molecule based on the calculated distance. Here, the sequence data for the first molecule and / or the sequence data for the second molecule may be virtual sequence data generated by a method similar to that described in step S11(2). Preferably, at least one of the sequence data is the sequence data obtained in step S10, and more preferably, both are the sequence data obtained in step S10.
[0080] FIG. 5 shows an example of an analysis in this embodiment in which the first molecule is a wild-type (natural) molecule and the second molecule is an in vitro transcript-derived (synthesized in a test tube) molecule. In FIG. 5 , the upper row shows multiple aligned series data obtained from wild-type molecules and multiple aligned series data obtained from in vitro transcript-derived molecules. The middle row shows the signal value distribution of the series data of wild-type molecules (WT) and the signal value distribution of the series data of in vitro transcript-derived molecules (IVT) at several points in the series data. In the graph in the middle row, the vertical axis represents signal value and the horizontal axis represents frequency. The difference in signal value distribution between WT and IVT indicates that the modification state of the wild-type molecule and the in vitro transcript-derived molecule is different at that point. The lower row shows the distance between the signal value distribution of the series data of wild-type molecules and the signal value distribution of the series data of in vitro transcript-derived molecules at all points in the series data. As highlighted by the triangles at the bottom of this distance graph, the portion where the calculated distance exceeds a predetermined threshold can be estimated as the modification position of the molecule derived from the wild-type strain.
[0081] In this aspect, as described above, the distribution distance of signal values of the sequence data for each type of molecule is calculated at a predetermined point in the plurality of sequence data for two or more aligned molecules. In the method of this embodiment, since the plurality of sequence data are aligned, it is easy to compare corresponding points of the sequence data. Taking the graph in the upper part of Figure 5 as an example, by comparing points having the same horizontal axis value, the distribution distance of signal values of the sequence data for each type of molecule can be easily calculated. In this aspect, the distribution distance of signal values of the plurality of sequence data for two or more molecules may be calculated at all points of the sequence data (all horizontal axis values in the upper part of Figure 5).
[0082] The distance between signal value distributions is a value that quantitatively represents the difference between two or more distributions, and can be calculated by a known method. Examples of such methods include a method that quantifies the difference between two probability distributions. Specific examples include the Kullback-Leibler (KL) distance and the Jensen-Shannon (JS) distance.
[0083] In this way, by calculating the distribution distance of the signal values of two or more types of molecules, it is possible to estimate locations that require attention as areas with different modification states. Therefore, hereinafter, data obtained by calculating the distribution distance of the signal values of two or more types of molecules and associating them with each point of the sequence data may be referred to as an "attention map."
[0084] (Second Aspect) Step S13 may include, for example, estimating the modification state of a first building block in a third molecule, the modification state of which is unknown, using a machine learning model trained with training data including sequence data obtained from a first molecule, the modification state of which is a first modification state of a first building block located at a first position in the sequence of building blocks, and sequence data obtained from a second molecule, the modification state of which is a second modification state of the first building block.
[0085] In this embodiment, for example, the first molecule may be a molecule derived from a wild-type strain, and the second molecule may be a molecule derived from an in vitro transcript or a molecule derived from a deletion strain in which an enzyme involved in the modification at the first position in a wild-type strain has been deleted.
[0086] In this embodiment, the training data for the machine learning model is sequence data labeled as the first or second molecule. Similarly to the first embodiment, the machine learning model may be weighted using an attention map created from molecules derived from a wild-type strain and molecules derived from in vitro transcripts or molecules derived from a mutant strain in which the modification state of a structural unit at a specific position has changed from that of the wild-type strain. In this case, portions of the attention map that are farther apart may be weighted more strongly.
[0087] (Third Aspect) Step S13 may include classifying the aligned sequence data into two or more types of clusters using, for example, an unsupervised machine learning model. In this aspect, the aligned sequence data may be clustered directly, or the aligned sequence data may be clustered based on the similarity between the aligned sequence data and the other sequence data after calculating the similarity.
[0088] The similarity between two series data may be calculated by a known method, but for example, the distance (absolute value of the difference in signal values) at each point of the aligned series data may be calculated, and the sum of the distances for all points may be used as an index of similarity. When clustering based on similarity, for example, taking the case of clustering N series data as an example, the similarity between each molecule and N series data including itself may be quantified (for example, by calculating the distance as described above), and N N-component vectors with the similarity as components may be generated, and these vectors may be clustered. Note that N is not particularly limited as long as it is 2 or greater.
[0089] The method for clustering vectors whose components are sequence data or similarities is not particularly limited, and methods such as hierarchical clustering, kmeans, and DBSCAN (Density-based spatial clustering of applications with noise) may be used.
[0090] Furthermore, for example, when the origin of the sequence data is known when the data is acquired, the machine learning model may be weighted using an attention map created from molecules derived from a first strain and molecules derived from a second strain, as in the first aspect above. The first molecule may be a molecule derived from a wild-type strain, and the second molecule may be a molecule derived from an in vitro transcript or a molecule derived from a deletion strain in which an enzyme involved in the modification at the first position in the wild-type strain has been deleted.
[0091] (Fourth Aspect) Step S13 may involve analyzing the modification state based on, for example, the redundancy of the sequence data calculated in step S12. For example, taking the case where the molecule to be analyzed is RNA, when RNA is measured by nanopore sequencing, the rate at which the RNA passes through the nanopore varies depending on whether the RNA is modified. One factor contributing to this is, for example, the rate at which the RNA duplex is unwound by the helicase to form a single strand as the RNA passes through the nanopore, which varies depending on whether the RNA is modified. For example, it has been reported that when RNA with a 2'O-methyl modification passes through the helicase, the passage time of the base being read in the nanopore (9-11 bases downstream) increases (Stephenson W, Razaghi R, Busan S, Weeks KM, Timp W, Smibert P (2022) Direct detection of RNA modifications and structure using single-molecule nanopore sequencing. Cell Genom 2).
[0092] Thus, since the scanning speed of the analyte molecule can vary depending on the modification state of the analyte molecule, the sequence data acquired by scanning the analyte molecule contains information about the modification state of the analyte molecule not only in terms of the magnitude of the signal value but also in terms of the length of the sequence data, i.e., the redundancy relative to the reference data. Therefore, in this embodiment, the modification state is analyzed based on the redundancy of the sequence data calculated in step S12.
[0093] In this aspect, for example, analysis may be performed in the same manner as in the first aspect, except that instead of calculating the distance of the distribution of signal values of the sequence data, a distance of redundancy in the sequence data is calculated. That is, in this aspect, for example, for a plurality of sequence data obtained and aligned from a first molecule whose constituent units have a first modification state and a plurality of sequence data obtained and aligned from a second molecule whose constituent units have a second modification state, calculating redundancy of the sequence data relative to reference data as in step S12, calculating the distance between the redundancy distribution at a predetermined point in the plurality of sequence data related to the first molecule and the redundancy distribution at a point in the plurality of sequence data related to the second molecule associated with the predetermined point, and estimating portions where the modification states differ between the first molecule and the second molecule based on the calculated distance. Here, the sequence data of the first molecule and / or the sequence data of the second molecule may be virtual sequence data generated by a method similar to that described in step S11(2) above, but preferably at least one of them is sequence data obtained in step S10, and more preferably both are sequence data obtained in step S10.
[0094] Alternatively, in this aspect, for example, the analysis of modification states may be performed by combining the analysis of the first aspect with an analysis based on redundancy in sequential data. For example, for a first molecule whose constituent units have a first modification state and a second molecule whose constituent units have a second modification state, the method may include estimating a difference in modification state between the first molecule and the second molecule based on a signal value distribution at a first point and a redundancy distribution at a second point in a plurality of sequential data related to the first molecule, and a signal value distribution at a plurality of sequential data points related to the second molecule associated with the first point and a redundancy distribution at a plurality of sequential data points related to the second molecule associated with the second point. Here, the first point and the second point may be the same or different. For example, the first point may be the point where the distance in the distribution of signal values between the first molecule and the second molecule is the largest (the peak portion of the graph of the distance in the distribution of signal values), and the second point may be the point where the distance in the distribution of redundancy between the first molecule and the second molecule is the largest (the peak portion of the graph of the distance in the distribution of redundancy).
[0095] Specifically, at a given point in the aligned sequence data, signal values may be plotted on a first axis (e.g., the vertical axis) and redundancy on a second axis (e.g., the horizontal axis), creating a scatter plot reflecting the signal values and local redundancy. The difference in modification state between the first molecule and the second molecule may then be estimated based on the scatter plot. Furthermore, the abundance ratio of the first molecule to the second molecule in the sample may be estimated from each plot in the scatter plot. One such method involves clustering the plots in the scatter plot. Specifically, for example, the current value and duration values may be assumed to be distributed according to a bivariate normal distribution for each modification state, and the modification state of each plot in the scatter plot may be labeled using a bivariate Gaussian mixture model.
[0096] Alternatively, for example, as described in the first and present aspects above, with respect to a first molecule whose constituent units have a first modification state and a second molecule whose constituent units have a second modification state, the method may include calculating the distance between the distribution and redundancy of signal values at a predetermined point in a plurality of series data related to the first molecule and the distribution and redundancy of signal values at a point in a plurality of series data related to the second molecule associated with the predetermined point, and estimating the portions in which the modification states differ between the first molecule and the second molecule based on the two calculated distances.
[0097] <Analysis Example> Figure 6 shows an example of analysis using the analysis system 1 according to this embodiment. In the method shown in Figure 6, sequence data is acquired by the measurement device 20 for molecules derived from in vitro transcripts (IVT) derived from wild-type strains, molecules derived from deletion strains (deletion strains A to C) in which an enzyme involved in modification at a specific position in a wild-type strain has been deleted, and a sample to be analyzed. Here, the modification state of the sample to be analyzed is unknown, and the sequences of the building blocks of all molecules are identical.
[0098] In the method shown in Figure 6, reference data is created by mixing molecules derived from wild-type strains with molecules derived from in vitro transcripts (IVT), and all of the above-mentioned sequence data are aligned based on this reference data. The wild-type molecules and the molecules derived from each deletion strain are analyzed, for example, using the method described in the modification state analysis step above, to create corresponding attention maps and machine learning models. By analyzing the target sample based on the created attention map and machine learning model, the modification state of the target sample can be analyzed molecule by molecule.
[0099] Although Figure 6 shows deletion strains A to C, the number of molecules derived from deletion strains compared with molecules derived from wild-type strains may be two or less, or may be four or more. Furthermore, the reference data does not have to be a mixture of molecules derived from wild-type strains and molecules derived from in vitro transcripts; for example, it may be a mixture of multiple series of data for molecules derived from wild-type strains. Furthermore, instead of and / or in addition to molecules derived from deletion strains, molecules derived from modified strains to which an enzyme involved in a modification at a specific position not present in the wild-type strain has been added may be used for analysis.
[0100] [Diagnostic Support Method] The analysis system 1 according to this embodiment can be used in a method for supporting the diagnosis of a disease that causes a disorder in the modification state of a structural unit in a molecule in which a plurality of structural units of one or more types are bonded. The diagnostic support method according to this embodiment includes analyzing a specimen derived from a patient containing a molecule in which a plurality of structural units of one or more types are bonded, and a specimen derived from a healthy subject containing the molecule, using the analysis system 1 according to this embodiment, and analyzing the difference in the modification state of the structural units in the molecule between the specimen derived from the patient and the specimen derived from the healthy subject.
[0101] In the diagnostic support method according to this embodiment, an attention map may be created based on, for example, sequence data obtained from a patient-derived specimen and sequence data obtained from a healthy subject, which can assist in easily determining whether a patient has a disorder in the modification state of the structural units of the molecule.
[0102] In the diagnostic assistance method according to this embodiment, a specimen derived from a patient and a specimen derived from a healthy subject may be analyzed using, for example, a method similar to the analytical example described with reference to Figure 6, thereby estimating how the modification state of molecules contained in the specimen derived from the patient differs from that of the specimen derived from the healthy subject.
[0103] [Additional Notes] The present invention includes the following embodiments. [1] A method for analyzing modification states of structural units in molecules in which a plurality of structural units of one or more types are bonded together, comprising: acquiring, for each of a plurality of target molecules, series of data of signal values that vary depending on the type and modification state of each structural unit obtained by scanning the molecule; aligning each of the series of data obtained from the plurality of target molecules with reference data to mutually align the plurality of series of data; and analyzing the modification states of at least some of the target molecules based on the aligned series of data. [1-1] A method for analyzing modification states of bases in nucleic acids, comprising: acquiring, for each of a plurality of target molecules, series of data of signal values that vary depending on the type and modification state of each base obtained by scanning the nucleic acid; aligning each of the series of data obtained from the plurality of nucleic acids with reference data to mutually align the plurality of series of data; and analyzing the modification states of at least some of the plurality of nucleic acids based on the aligned series of data. (Hereinafter, [1] and [1-1] are collectively referred to as [1].) [2] The analytical method according to [1], further comprising generating the reference data by mixing a plurality of series of data including the acquired series of data. [3] The analytical method according to [2], wherein the mixed plurality of series of data comprises two or more types of series of data obtained or estimated from two or more types of molecules having the same sequence of structural units as at least one of the target molecules, but differing in the modification state of the structural units.[4] The analytical method according to any one of [1] to [3], wherein analyzing the modification states of the plurality of analyte molecules includes: calculating a distance between a distribution of signal values at a predetermined point in the plurality of series data for the first molecule and a distribution of signal values at a point in the plurality of series data for the second molecule associated with the predetermined point, using the plurality of series data obtained or estimated from and aligned with a first molecule whose constituent units have a first modification state and the plurality of series data obtained or estimated from and aligned with a second molecule whose constituent units have a second modification state; and estimating a portion where the modification state differs between the first molecule and the second molecule based on the calculated distance. [5] The analytical method of any one of [1] to [4], wherein analyzing the modification states of the plurality of analyte molecules comprises estimating the modification state of a first building block in a third molecule, the modification state of which is unknown, using a machine learning model trained with training data including the sequence data obtained from a first molecule, in which a first building block located at a first position in a sequence of the building blocks of the analyte molecule is in a first modification state, and the sequence data obtained from a second molecule, in which the modification state of the first building block is in a second modification state. [6] The analytical method of any one of [1] to [5], wherein analyzing the modification states of the plurality of analyte molecules comprises classifying the aligned plurality of series of data into two or more clusters using an unsupervised machine learning model. [7] The analytical method of [6], wherein classifying the aligned plurality of series of data into clusters comprises calculating a similarity between the aligned plurality of series of data and classifying the aligned plurality of series of data into two or more clusters based on the similarity. [8] The analytical method according to any one of [1] to [7], wherein the molecules are nucleic acids. [9] The analytical method according to [8], wherein the plurality of analyte molecules include at least two of nucleic acids derived from a wild-type strain, nucleic acids having no modifications, and nucleic acids derived from a modified strain in which the modification state of a specific base has changed from that of the wild-type strain.[9-1] The analytical method according to any one of [1] to [7], wherein the plurality of analyte molecules include at least two of molecules derived from a wild-type strain, molecules without modifications, and molecules derived from a mutant strain in which the modification state of a specific base has changed from that of the wild-type strain. (Hereinafter, [9] and [9-1] are collectively referred to as [9].)
[10] The analytical method according to [9], wherein the plurality of analyte molecules include at least nucleic acids derived from the wild-type strain, and in the nucleic acids derived from the wild-type strain, the ratio of the number of modified bases to the total number of bases constituting the nucleic acid is 4% or more. [10-1] The analytical method according to [9-1], wherein the plurality of analyte molecules include at least molecules derived from the wild-type strain, and in the molecules derived from the wild-type strain, the ratio of the number of modified structural units to the total number of structural units constituting the molecule is 4% or more. (Hereinafter,
[10] and [10-1] are collectively referred to as
[10] .)
[11] The analysis method according to any one of [1] to
[10] , further comprising calculating redundancy of the series of data with respect to the reference data when aligning each of the plurality of series of data with the reference data, and analyzing the modification states based on the calculated redundancy.
[12] A system for analyzing modification states of structural units in a molecule in which a plurality of structural units of one or more types are bonded together, comprising: a measurement device that acquires, for each of a plurality of analyte molecules, series of data of signal values that vary depending on the type and modification state of each structural unit obtained by scanning the molecule; an acquisition unit that acquires the series of data acquired by the measurement device; an alignment unit that aligns each of the plurality of series of data acquired from the plurality of analyte molecules with reference data, thereby aligning the plurality of series of data with each other; and an analysis unit that analyzes the modification states of at least some of the plurality of analyte molecules based on the aligned plurality of series of data.
[13] An analytical device for analyzing modification states of structural units in a molecule in which a plurality of structural units of one or more types are bonded together, comprising: an acquisition unit that acquires, for each of a plurality of analyte molecules, series data of signal values that change depending on the type and modification state of each structural unit obtained by scanning the molecule; an alignment unit that aligns the series data acquired from the plurality of analyte molecules with reference data, thereby aligning the plurality of series data with each other; and an analysis unit that analyzes the modification states of at least some of the plurality of analyte molecules based on the aligned series data.
[14] A program that causes a computer to function as an analytical device that analyzes modification states of structural units in molecules in which a plurality of structural units of one or more types are bonded, the program causing the computer to function as: an acquisition unit that acquires, for each of a plurality of analyte molecules, series of data from a measurement device that acquires series of data of signal values that vary depending on the type and modification state of each structural unit obtained by scanning the molecule, the said series of data, an alignment unit that aligns the plurality of series of data acquired from the plurality of analyte molecules with each other by aligning each of the plurality of series of data with reference data, and an analysis unit that analyzes the modification states of at least some of the plurality of analyte molecules based on the aligned plurality of series of data.
[15] A method for assisting in the diagnosis of a disease in which a disorder occurs in the modification state of structural units in molecules in which a plurality of structural units of one or more types are bonded, the method comprising: analyzing a patient-derived specimen containing the molecule and a healthy subject-derived specimen containing the molecule by the analytical method according to any one of [1] to
[11] ; and analyzing a difference in the modification state of structural units in the molecule between the patient-derived specimen and the healthy subject-derived specimen.
[0104] The present invention will be described in more detail below using examples. The present invention is not limited to the following examples. In particular, in the following examples, the modification state of tRNA is analyzed, but the molecule to be analyzed is not limited to tRNA.
[0105] 1. Nanopore Sequencing of tRNA (Overview) To perform nanopore sequencing on RNA samples, an adapter (second adapter) provided by ONT (Oxford Nanopore Technologies) was ligated (Figure 7(A)). The second adapter is bound to a motor protein required for the RNA strand to pass through the nanopore. Due to differences in the protruding sequences, the second adapter was not ligated directly to the tRNA. Instead, an adapter (first adapter) that mediates between the tRNA and the second adapter was ligated first. The tRNA ligated to the first adapter was then purified by electrophoresis and gel excision, and the second adapter was then ligated. Subsequent procedures followed the Direct RNA Sequencing manual published by ONT.
[0106] (Design of the 1st adapter) A 1st adapter was designed, consisting of a 3' adapter ligated to the 3' end of the tRNA and a 5' adapter ligated to the 5' end (Figure 7(B)). The last 10 bases of the 3' adapter were DNA, and the rest were RNA. To ligate with the tRNA, the 3' adapter was designed to form a complementary strand to the overhanging 3' end sequence of the mature tRNA molecule. Since there are five possible patterns for the overhanging sequence at the 3' end of the mature tRNA depending on the tRNA species (ACCA, UCCA, GCCA, CCCA, CCA), the corresponding 5' adapters were designed as shown in the table below. The DNA sequence was set to TAGTAGGTTC (SEQ ID NO: 1) so that the DNA sequence at the end of the 3' adapter would form a complementary strand with the 2' adapter during ligation with the 2' adapter. A spacer was incorporated into the center of the 3' adapter, which is not involved in ligation but is used to generate a characteristic current value during nanopore sequencing. The spacer was used for adapter trimming of the current signal during data analysis.
[0107]
[0108] (Ligation reaction of 1st adapter) First, annealing of the 1st adapter was performed. The 5' adapter and the 3' adapter were mixed in a ratio of 3:2 and annealed at 75°C for 3 minutes. The temperature was then lowered to 25°C at a rate of 1.2°C / min, and the adapter ligation reaction solution was mixed.
[0109]
[0110] The mixture was then incubated at 22°C for 6 hours. Next, 1 / 20 equivalent of 0.5 mol / L EDTA was added, followed by 1 equivalent of PCI reagent, and the mixture was stirred. The mixture was centrifuged at 15,000 g for 10 minutes at 4°C, and the upper layer was collected. The mixture was then precipitated with ethanol and dissolved in water to obtain tRNA ligated with the first adapter.
[0111] (Excision and Purification of Ligated tRNA) A 7M urea 8% acrylamide gel was prepared and left overnight. 10 μL of tRNA was mixed with 10 μL of 2x loading buffer to prepare a sample. The sample solution was heat-shocked at 65°C for 5 minutes, and the gel was placed in an electrophoresis apparatus and pre-run at 100 V for 10 minutes. The wells were washed, and the sample and marker were loaded. The marker was prepared by mixing 1 μL of DynaMarker RNA Low2 with the loading buffer. Electrophoresis was performed at 100 V for 10 minutes and at 250 V for 40 minutes. The gel was removed, stained with toluidine blue, and washed with pure water. The fragment between 100 nt and 200 nt, containing the luggage spaced tRNA, was excised from the gel. The excised fragment was crushed in a 1.5 mL tube. 250 μL of Elution Buffer was added, stirred at 1100 rpm for 1 hour, centrifuged, and the supernatant was collected. The process of adding Elution Buffer, stirring, centrifugation, and collecting the supernatant was repeated twice. The solution together with the gel was placed in a spin column and centrifuged at 12,000 g for 2 minutes at 4°C. The filtrate was collected and precipitated with ethanol. Small molecules were removed by drop dialysis using a membrane filter, and the ligated tRNA was purified by dissolving in pure water.
[0112] (Ligation reaction of second adapter) The ligation reaction solution was mixed as shown in the table below and reacted at 22°C for 20 minutes. One equivalent of RNA Clean XP was added and gently mixed using a rotator for 5 minutes. The beads were adsorbed with a magnet, the liquid was removed, 75 μL of WSB was added, and the mixture was tapped. This process was repeated, and then 21 μL of ELB was added in place of the 75 μL of WSB, followed by pipetting. After leaving the mixture at room temperature for 10 minutes, the beads were adsorbed with a magnet and the liquid was collected.
[0113]
[0114] (Nanopore sequencing) Nanopore sequencing was performed according to the Direct RNA sequencing manual published by ONT. The reagents used above are listed below. T4 RNA Ligase 2 (10 units / μL) (NEB M0239L) DynaMarker RNA Low 2 (BioDynamics Laboratories DM152) NEBNext Quick Ligation Reaction Buffer (NEB B6058) T4 DNA Ligase (2,000 units / μL) (NEB M0202M) RNA CleanXP (Beckman-Coulter A63987) Direct RNA sequencing kit (Oxford nanopore technologies SQK-RNA002) Reagents with the following composition:
[0115]
[0116] 2. Detection of modifications in E. coli tRNA E. coli tRNA Asn The modification of E. coli tRNA was detected. Asn As shown in Figure 8(A), seven types of RNA modifications are present at nine positions in tRNAs derived from E. coli wild-type strains (hereinafter also referred to as "WT strains"), various tRNA modification enzyme-deficient strains (hereinafter referred to as "KO strains"), and in vitro transcripts (hereinafter referred to as "IVT"). Asnwas obtained or prepared, and current signals were measured by nanopore sequencing to obtain time series data.
[0117] The time series data from the WT strain and the time series data from the IVT strain were mixed at a 1:1 ratio, and reference data was created by DTW barycenter averaging. All measured time series data was aligned using this reference data. DTW was used for alignment. The time series data from the WT strain and the time series data from the IVT strain were compared, and the Kullback-Leibler (KL) distance was calculated for each point of the time series data. When the data was plotted, multiple peaks were confirmed (WT vs. IVT graph in Figure 8(B)).
[0118] Next, the time series data from the WT strain was compared with the time series data from various KO strains, and the Kullback-Leibler (KL) distance was calculated for each point in the time series data. The data were plotted, and peaks were detected at positions corresponding to each modification site (Figure 8(B)). A peak was detected for dihydrouridine (D16) at position 16 when comparing the WT strain with the ΔdusC strain, and a peak was detected for D20 when comparing the WT strain with the ΔdusA strain. Similarly, a peak was detected for queusine (Q34) at position 34, pseudouridine (Ψ39) at position 39, and 7-methylguanosine (m) at position 46. 7 The modifications at Ψ55 and Ψ46 were detected by comparing the WT strain with the KO strains, Δtgt, ΔtruA, ΔtrmB, and ΔtruB. These KL distance peaks were in good agreement with the peaks observed in the comparison of the time-series data derived from the WT strain with the time-series data derived from IVT, demonstrating that the analysis method according to this embodiment can analyze the modification state of tRNA.
[0119] 3. Detection of modifications on individual tRNA molecules (Ψ55) Next, analysis was performed using a supervised machine learning model (Figure 9). Gly3 Ψ55 of the wild-type nucleotide sequence was selected. LC / MS analysis revealed that this site is Ψ-modified at a rate of 99% or more in the wild-type strain.
[0120] First, the WT strain, the ΔtruB strain (KO strain), and the IVT-derived tRNA Gly3 The time series data were obtained by nanopore sequencing and measured using nanopore sequencing. Using a method similar to that described in "2. Detection of E. coli tRNA Modifications," the time series data from the WT strain and the ΔtruB strain were compared, and the Kullback-Leibler (KL) distance was calculated for each point in the time series data to create an attention map. The peak positions of the attention map were extracted as features, and the time series data from the WT strain were labeled as Ψ55, and the time series data from the ΔtruB strain were labeled as unmodified uridine (U55). A machine learning model (RandomForestClassifier (Breiman, L. Random Forests. Machine Learning 45, 5-32, doi:10.1023 / a:1010933404324 (2001))) was trained to classify these data using training data. As a result, an accuracy of 94.25% was achieved on the test data (Figure 9, top panel).
[0121] Furthermore, tRNA from the WT strain and the ΔtruB strain Gly3 Using samples containing these at an appropriate ratio, time-series data was obtained by nanopore sequencing, and then the tRNA having Ψ55 in each sample was analyzed. Gly3 The proportion of tRNA derived from the WT strain was estimated using the machine learning model. Gly3 and tRNA from the ΔtruB strain Gly3 The actual mixture ratio of tRNA with Ψ55 estimated by the machine learning model Gly3 The results showed good agreement with the proportions of α- and β-modifications (lower panel of Figure 9). These results demonstrate that the analysis method according to this embodiment makes it possible to accurately determine modification states from individual time-series data.
[0122] 4. Detection of modifications for each tRNA molecule (detection of multiple modification states by clustering current signals) Next, for cases where multiple modification states were present, the modification states were classified using an unsupervised machine learning model (Figure 10). Pro3A modified uridine was selected at position 34 of the uridine-containing 5-methoxycarbonyl methoxyuridine (mcmo 5 U34), 5-carboxymethoxyuridine (cmo 5 It has been found that three types of uridine exist in the ribosomal tRNA: 5-methoxycarbonylmethoxyuridine (U34), and unmodified uridine (U34) (Sakai, Y., Miyauchi, K., Kimura, S. & Suzuki, T. Biogenesis and growth phase-dependent alteration of 5-methoxycarbonylmethoxyuridine in tRNA anticodons. Nucleic Acids Res 44, 509-523, doi:10.1093 / nar / gkv1470 (2016)).
[0123] First, the WT strain tRNA Pro3 From the time series data, the above three types (mcmo 5 U34, cmo 5 To classify the modification status of tRNAs (U34, U35, and U34), we used tRNAs derived from the WT strain, the KO strain ΔcmoM, and IVT. Pro3 was obtained or prepared, and the current signal was measured by nanopore sequencing to obtain time series data. 5 Methylate U34 and mcmo 5 An enzyme that forms U34, and ΔcmoM strain-derived tRNA Pro3 Included in 5 U34 disappears, cmo 5 Only U34 and U34 are present. On the other hand, IVT tRNA Pro3 Only U34 exists in the
[0124] The time series data from the WT strain, ΔcmoM strain, and IVT were aligned using a method similar to that described in "2. Detection of E. coli tRNA Modifications." For all combinations of 1,000 pieces of time series data from the WT strain, ΔcmoM strain, and IVT (3,000 pieces of data in total), the sum of the distances between each point in the two time series data was calculated, and a 3,000 x 3,000 distance matrix was created. The calculated sum of the distances corresponds to the similarity between the two time series data, with a smaller sum of the distances indicating a higher similarity.
[0125] The distance matrix was subjected to hierarchical clustering using Ward's method. When the data were divided into three clusters 1, 2, and 3, all the time series data from IVT were classified into cluster 3, the time series data from the ΔcmoM strain into clusters 1 and 3, and the time series data from the WT strain into clusters 1, 2, and 3. From these results, it was found that cluster 3 was U34 and cluster 2 was cmo. 5 U34, cluster 1 is mcmo 5 The abundance ratio of each modification state in wild-type tRNA was calculated from the ratio of the number of time-series data contained in each cluster, and the calculated percentages were in good agreement with the modification rates measured by LC / MS. This demonstrates that the analytical method according to this embodiment enables the detection and quantification of tRNA modifications.
[0126] 5. Detection and quantification of taurine modification in human mitochondrial tRNA Mutations in the human mitochondrial tRNA (mt-tRNA) gene are known to cause mitochondrial diseases. A typical form of this disease, MELAS (Mitochondrial myopathy, encephalopathy, lactic acidosis, and stroke-like episodes), is caused by mutations in mt-tRNA. Leu(UUR)It is caused by a point mutation in the gene (Goto Y, Nonaka I, Horai S (1990) A mutation in the tRNA(Leu)(UUR) gene associated with the MELAS subgroup of mitochondrial encephalomyopathies. Nature 348: 651-653). In addition, mutant mt-tRNA extracted from cells derived from MELAS patients Leu(UUR) In this case, 5-taurinomethyluridine (τm 5 It has been reported that a significant decrease in the modification rate of mt-tRNA (Suzuki T, Nagao A, Suzuki T (2011) Human mitochondrial tRNAs: biogenesis, function, structural aspects, and diseases. Annual review of genetics 45: 299-329; Yasukawa T, Suzuki T, Suzuki T, Ueda T, Ohta S, Watanabe K (2000) Modification defect at anticodon wobble nucleotide of mitochondrial tRNAs(Leu)(UUR) with pathogenic mutations of mitochondrial myopathy, encephalopathy, lactic acidosis, and stroke-like episodes. J Biol Chem 275: 4251-4257). Leu(UUR) The secondary structure of is shown in Figure 12(A).
[0127] From this, it can be seen that MELAS mutations (A3243G, T3271C, etc.) Leu(UUR) This destabilizes the three-dimensional structure of τm, preventing the recognition of tRNA-modifying enzymes. 5 This is thought to cause a decrease in the modification rate of U34. 5U34 is essential for decoding the UUG codon, and defects in this modification are thought to be one of the causes of mitochondrial dysfunction (Suzuki T, et al. (2011) cited above; Kirino Y, Yasukawa T, Ohta S, Akira S, Ishihara K, Watanabe K, Suzuki T (2004) Codon-specific translational defect caused by a wobble modification deficiency in mutant tRNA from a human mitochondrial disease. Proc Natl Acad Sci USA 101: 15070-15075).
[0128] Therefore, in this example, mt-tRNA Leu(UUR) τm 5 We created a model to classify the modification status of U34 and demonstrated that it is possible to measure the modification rate from nanopore sequencing data of total tRNA extracted from myoblasts derived from MELAS patients. Leu(UUR) and τm 5 mt-tRNA derived from a strain lacking U-modifying enzyme (GTPBP3) Leu(UUR) and were used to measure current signals by nanopore sequencing in accordance with the method described above, and time-series data were obtained.
[0129] Using a method similar to that described in "2. Detection of E. coli tRNA modifications," reference data was created based on the time-series data derived from HeLa cells and the time-series data derived from the modification enzyme-deficient strain, and all of the measured time-series data was aligned using this reference data ( FIG. 12(B) ). During alignment, the duration was calculated for each point in the time-series data based on the above-mentioned formulas (I) to (IV) ( FIG. 12(B) ).
[0130] Using the same method as described in "2. Detection of E. coli tRNA modifications," the Kullback-Leibler (KL) distance was calculated for each point in the time series data of current value based on the aligned all time series data from HeLa cells and the aligned all time series data from the modification enzyme-deficient strain ( FIG. 12(B) ). Similarly, the Kullback-Leibler (KL) distance was calculated for each point in the time series data of Duration calculated above ( FIG. 12(B) ).
[0131] As a result, in the Kullback-Leibler (KL) distance of the current value and the Kullback-Leibler (KL) distance of the duration, τm 5 A peak presumably derived from U34 was confirmed (indicated by a triangle in Figure 12(B)). When the current value and duration values were extracted from the peaks of the Kullback-Leibler (KL) distance of the current value and the Kullback-Leibler (KL) distance of the duration, and the values were plotted as a scatter diagram, it was confirmed that the distributions of the data derived from HeLa cells and the data derived from the modification enzyme-deficient strain were clearly separated (Figure 12(C)).
[0132] Based on the data in FIG. 12(C), the distribution of U (uridine) on the scatter plot of Duration × current value and τm 5 The parameters of the Gaussian distribution for U were calculated using the expectation-maximization (EM) algorithm. Specifically, we assumed that the current value and the Duration value were distributed according to a bivariate normal distribution for each modification state, and used a bivariate Gaussian mixture model. As a result, the modification rate of U derived from HeLa cells was 91.2% (228 of 250 reads were within the τm 5 The modification rate of U derived from the modification enzyme-deficient strain was 1.7% (9 out of 541 reads were τm 5 U. In FIG. 12(D), the "★" indicates the center of the normal distribution, and the circle indicates a range twice the standard deviation.
[0133] Furthermore, when this model was applied to sequencing data from myoblasts of MELAS patients, the modification rate was 5.3% (4 of 76 reads were τm 5 U.) (Figure 12(D)). Thus, we successfully measured the decrease in taurine modification rate due to the MELAS mutation through nanopore sequencing of human total tRNA.
[0134] 6. Multi-step modification mcm in yeast tRNA 5 s 2 Detection and quantification of U In eukaryotes such as humans and yeast, tRNAs corresponding to amino acids such as glutamic acid (Glu), lysine (Lys), and glutamine (Gln) contain 5-methoxycarbonylmethyl-2-thiourdine (mcm) at the first letter (position 34) of the anticodon. 5 s 2 U) modification. 5 s 2 U is a multi-step modification formed by a complex biosynthetic pathway as shown in Figure 13(A). 5 s 2 U is difficult to detect using conventional methods (e.g., conventional RNA sequencing or primer extension), and LC / MS analysis is typically performed after isolating and purifying tRNA to detect this modification (Yoshida M, Kataoka N, Miyauchi K, Ohe K, Iida K, Yoshida S, Nojima T, Okuno Y, Onogi H, Usui T et al (2015) Rectifier of aberrant mRNA splicing recovers tRNA modification in familial dysautonomia. Proc Natl Acad Sci USA 112: 2764-2769).
[0135] In humans, mcm 5 s 2It has been reported that abnormal splicing of the ELP1 gene, which is essential for U biosynthesis, causes familial dysautonomia, and that abnormalities in the ELP3 gene cause ALS, suggesting the importance of this modification (Anderson SL, Coli R, Daly IW, Kichula EA, Rork MJ, Volpi SA, Ekstein J, Rubin BY (2001) Familial dysautonomia is caused by mutations of the IKAP gene. Am J Hum Genet 68: 753-758; Simpson CL, Lemmens R, Miskiewicz K, Broom WJ, Hansen VK, van Vught PW, Landers JE, Sapp P, Van Den Bosch L, Knight J et al (2009) Variants of the elongator protein 3 (ELP3) gene are associated with motor neuron degeneration. Hum Mol Genet 18: 472-481).
[0136] Therefore, in this example, yeast tRNA Gln1a Targeting 5 s 2 U and one of its intermediates, 5-methoxycarbonylmethylurdine (mcm 5 We created a model to classify three types: modified U, unmodified U, and unmodified U.
[0137] First, wild-type (WT) tRNA Gln1a and tRNA derived from a modification enzyme-deficient strain Gln1a The deletion strains used were an ncs6 deletion strain (Δncs6) and an elp3 and ncs6 double deletion strain (Δelp3Δncs6). Using these tRNAs, current signals were measured by nanopore sequencing according to the method described above to obtain time-series data.
[0138] The time-series data from the WT strain and the time-series data from Δelp3Δncs6 were mixed at a 1:1 ratio, and reference data was generated using DTW barycenter averaging. All measured time-series data were aligned using this reference data ( Figure 13(B) ). Duration was calculated for each point in the time-series data using a method similar to that described in "5. Detection and quantification of taurine modifications in human mitochondrial tRNA" ( Figure 13(B) ).
[0139] Next, using the same method as described in "2. Detection of E. coli tRNA Modifications," the Kullback-Leibler (KL) distance was calculated for each point of the aligned time-series data of current values derived from the Δelp3Δncs6 strain, with the aligned time-series data of current values derived from the WT strain as the reference ( FIG. 13(B) ). Similarly, the Kullback-Leibler (KL) distance was also calculated for each point of the time-series data of Duration calculated above ( FIG. 13(B) ). Similarly, the Kullback-Leibler (KL) distance was also calculated for the aligned time-series data of current values derived from the Δncs6 strain and the time-series data of Duration, with the aligned time-series data of current values derived from the WT strain as the reference.
[0140] Next, peaks were observed in the two Kullback-Leibler (KL) distances calculated from the time-series data derived from Δelp3Δncs6 and the two Kullback-Leibler (KL) distances calculated from the time-series data derived from Δncs6 (Fig. 13(B)). When the current and duration values were extracted from the peaks of the current and duration distances and plotted as a scatter diagram, three types of distributions were confirmed among the data derived from the WT strain, Δncs6, and Δelp3Δncs6 (Fig. 13(C)).
[0141] Based on the data in FIG. 13(C), the distribution of U (uridine) on the scatter plot of Duration × current value, mcm 5 Distribution of U and mcm 5 s2 The Gaussian distribution parameters for the U distribution were calculated using the expectation-maximization (EM) algorithm. Specifically, we assumed that the current and duration values were distributed according to a bivariate normal distribution for each modification state, and used a bivariate Gaussian mixture model. The results showed that the abundance of each modification state was generally consistent with the results measured by LC / MS (Figure 14). Specifically, in the wild-type strain, mcm 5 s 2 U, Δelp3Δncs6 U, Δncs6 mcm 5 U was detected with high accuracy. In Fig. 14, "★" indicates the center of the normal distribution, and the circle indicates the range twice the standard deviation.
[0142] In this way, the analysis method of this embodiment can detect mcm, which is difficult to detect by conventional RNA sequencing. 5 s 2 It was shown that U can be detected. 5 It was also shown that detection of U and U is possible.
[0143] 1...analysis system, 10...analysis device, 10a...CPU, 10b...RAM, 10c...ROM, 10d...communication unit, 10e...input unit, 10f...display unit, 11...acquisition unit, 12...reference data generation unit, 13...alignment unit, 14...analysis unit, 20...measuring device
Claims
1. A method for analyzing the modification state of structural units in a molecule in which multiple structural units of one or more types are bonded together, comprising: acquiring, for each of a plurality of target molecules, series of data of signal values which vary depending on the type and modification state of each structural unit obtained by scanning the molecule; aligning each of the series of data acquired from the plurality of target molecules with reference data, thereby aligning the plurality of series of data with each other; and analyzing the modification states of at least some of the plurality of target molecules based on the aligned series of data.
2. The analysis method according to claim 1, further comprising: generating the reference data by mixing a plurality of sequence data including the acquired sequence data.
3. The analytical method described in claim 2, wherein the multiple series data to be mixed include two or more types of series data obtained or estimated from two or more types of molecules having the same sequence of structural units as at least one of the target molecules, but differing in the modification state of the structural units.
4. The analytical method according to claim 1, wherein analyzing the modification states of the plurality of analyte molecules includes: calculating a distance between a distribution of signal values at a predetermined point of the plurality of series data related to the first molecule and a distribution of signal values at a point of the plurality of series data related to the second molecule corresponding to the predetermined point, using the plurality of series data obtained or estimated from and aligned on a first molecule, the modification state of the constituent units of which is a first modification state, and the plurality of series data obtained or estimated from and aligned on a second molecule, the modification state of the constituent units of which is a second modification state; and estimating a portion where the modification state differs between the first molecule and the second molecule based on the calculated distance.
5. The analytical method of claim 1, wherein analyzing the modification states of the plurality of analyte molecules comprises estimating the modification state of a first building block in a third molecule, in which the modification state of the first building block is unknown, using a machine learning model trained with teacher data including the sequence data obtained from a first molecule, in which a modification state of a first building block present at a first position in a sequence of building blocks of the analyte molecule is a first modification state, and the sequence data obtained from a second molecule, in which a modification state of the first building block is a second modification state.
6. The analytical method according to claim 1, wherein analyzing the modification states of the plurality of analyte molecules comprises classifying the aligned plurality of series data into two or more clusters using an unsupervised machine learning model.
7. The method of claim 6, wherein the classifying into clusters comprises: calculating a similarity between each of the aligned plurality of sequence data; and classifying the aligned plurality of sequence data into two or more types of clusters based on the similarity.
8. The method of claim 1, wherein the molecule is a nucleic acid.
9. The analytical method according to claim 8, wherein the plurality of analyte molecules include at least two of nucleic acids derived from a wild-type strain, nucleic acids having no modifications, and nucleic acids derived from a mutant strain in which the modification state of a specific base has been changed from that of the wild-type strain.
10. The analytical method according to claim 9, wherein the plurality of analyte molecules at least comprise nucleic acid derived from the wild strain, and the ratio of the number of modified bases to the total number of bases constituting the nucleic acid in the nucleic acid derived from the wild strain is 4% or more.
11. The analysis method according to claim 1, further comprising: calculating redundancy of the sequence data with respect to the reference data when aligning each of the plurality of sequence data with the reference data; and performing an analysis of the modification state based on the calculated redundancy.
12. A system for analyzing the modification state of structural units in a molecule in which one or more types of structural units are bonded together, comprising: a measuring device that acquires, for each of a plurality of analyte molecules, series data of signal values that vary depending on the type and modification state of each structural unit obtained by scanning the molecule; and an analyzing device comprising: an acquisition unit that acquires the series data acquired by the measuring device; an alignment unit that aligns each of the plurality of series data acquired from the plurality of analyte molecules with reference data, thereby aligning the plurality of series data with each other; and an analysis unit that analyzes the modification state of at least some of the plurality of analyte molecules based on the aligned plurality of series data.
13. An analytical device for analyzing the modification state of structural units in a molecule in which multiple structural units of one or more types are bonded together, comprising: an acquisition unit that acquires, for each of a plurality of analyte molecules, series data of signal values that vary depending on the type and modification state of each structural unit obtained by scanning the molecule; an alignment unit that aligns the plurality of series data acquired from the plurality of analyte molecules with reference data, thereby aligning the plurality of series data with each other; and an analysis unit that analyzes the modification state of at least some of the plurality of analyte molecules based on the aligned plurality of series data.
14. A program that causes a computer to function as an analytical device that analyzes the modification state of structural units in a molecule in which one or more types of structural units are bonded together, comprising: an acquisition unit that acquires series data from a measurement device that acquires series data of signal values that vary depending on the type and modification state of each structural unit obtained by scanning the molecule for each of a plurality of analyte molecules; an alignment unit that aligns the plurality of series data acquired from the plurality of analyte molecules with reference data, thereby aligning the plurality of series data with each other; and an analysis unit that analyzes the modification state of at least some of the plurality of analyte molecules based on the aligned plurality of series data.
15. A method for assisting in the diagnosis of a disease that causes a disorder in the modification state of a structural unit in a molecule in which one or more types of structural units are linked together, the method comprising: analyzing a specimen derived from a patient containing the molecule and a specimen derived from a healthy subject containing the molecule by the analytical method described in any one of claims 1 to 11; and analyzing the difference in the modification state of the structural units in the molecule between the specimen derived from the patient and the specimen derived from the healthy subject.
Citation Information
Patent Citations
Color image reader
JP1996237436A
Base modification analysis using electrical signals
US20220328135A1
Microrna analysis using tunneling current
WO2020218489A1
