A method and system for ordering mitochondrial genetic variations based on neural networks

Abstract

This invention discloses a method and system for ranking mitochondrial genetic variations based on a neural network. The method obtains variation results after gene sequencing; based on the variation results and the HPO list, it obtains the feature vector corresponding to each mitochondrial genetic variation in the test sample; it inputs the feature vector corresponding to each mitochondrial genetic variation in the test sample into a trained neural network model to obtain the prediction result corresponding to each mitochondrial genetic variation in the test sample, and then ranks the prediction results. This invention solves the problems of low efficiency in existing manual variation analysis processes and the incomplete applicability of existing ranking algorithms to mitochondrial genetic variation analysis, thereby improving the efficiency of mitochondrial genetic variation analysis and enhancing the correlation between ranking results and clinical symptoms and the pathogenicity of variations.

Description

Technical Field

[0001] This invention relates to the field of mitochondrial gene detection technology, and in particular to a method and system for sequencing mitochondrial genetic variations based on neural networks. Background Technology

[0002] Mitochondria produce 90% of the energy needed by the human body and are vital energy metabolism organelles in every human cell. As extranuclear genetic material, mitochondrial DNA mutations can lead to various serious inherited metabolic and nervous system diseases, such as Leigh syndrome and mitochondrial myopathy. It is estimated that one in 5,000 people suffers from a hereditary mitochondrial disease, making the analysis of mitochondrial DNA variations of broad medical and social value.

[0003] Unlike autosomes in the cell nucleus, mitochondrial DNA possesses unique genetic characteristics, necessitating targeted optimization of variation analysis workflows. For instance, although mitochondrial DNA is only about 16kb in length, it can exist in 10 to over 1000 copies within a single cell. The heterogeneity of mitochondrial variations can be pathogenic; even variations in a few copies can be disease-causing. For example, mitochondrial DNA can be maternally inherited, lacking introns or exons. These differences mean that commonly used variation analysis standards are not entirely applicable to mitochondria. For instance, ClinGen (the Clinical Genomics Resource Center) expanded the ACMG (American College of Medical Genetics) classification guidelines for genetic variations into the ClinGen Mito Disease ACMG Specifications, a classification guideline specifically for mitochondria. This includes, but is not limited to, removing the PM3 indicator associated with dominant or recessive inheritance, removing the PM1 and PP2 indicators associated with gene variation hotspots, and modifying the PS2 indicator associated with de novo mutations to accommodate the maternally inherited characteristics of mitochondria. These theoretical and practical changes mean that the analysis workflow for autosomal variations is not entirely applicable to mitochondrial variations.

[0004] In practice, automated sorting (or scoring) of variant results is essential. High-throughput sequencing technology can obtain complete mitochondrial variants at low cost, but from an efficiency perspective, it is impossible to examine hundreds or thousands of variant results one by one and classify them by pathogenicity. Existing genetic variant sorting methods, such as Exomiser, rely on weighted averaging of several pathogenicity assessment algorithms, but the weight setting is highly subjective and becomes increasingly difficult to balance as the number of input parameters increases. Furthermore, Exomiser selects parameters specifically for autosomal variants. In addition, there is a lack of integration among pathogenicity scoring methods for mitochondrial variants. For example, APOGEE can predict the harmfulness of missense mutations, and MitoTip can predict the harmfulness of types, but their scores are not comparable and cannot be directly sorted together. Moreover, genetic variant sorting, in addition to calculating variant pathogenicity, also needs to consider phenotypic matching to make the analysis results more relevant to each specific case. Summary of the Invention

[0005] This invention aims to address at least one of the technical problems existing in the prior art. To this end, this invention proposes a neural network-based method and system for mitochondrial genetic variation sorting, which can solve the problems of low efficiency in existing manual variation analysis processes and the inadequacy of existing sorting algorithms for mitochondrial genetic variation analysis. It can improve the efficiency of mitochondrial genetic variation analysis and enhance the correlation between sorting results and clinical symptoms and the pathogenicity of variations.

[0006] In a first aspect, embodiments of the present invention provide a method for sequencing mitochondrial genetic variations based on a neural network, the method comprising:

[0007] Obtain the mutation results after gene sequencing;

[0008] Based on the mutation results and the HPO list, obtain the feature vector corresponding to each mitochondrial genetic variation in the test sample;

[0009] The feature vector corresponding to each mitochondrial genetic variant in the test sample is input into the trained neural network model to obtain the prediction result corresponding to each mitochondrial genetic variant in the test sample, and the prediction result is sorted.

[0010] Compared with the prior art, the first aspect of the present invention has the following beneficial effects:

[0011] This method obtains the mutation results after gene sequencing; based on the mutation results and the HPO list, it obtains the feature vector corresponding to each mitochondrial genetic variant in the test sample; it inputs the feature vector corresponding to each mitochondrial genetic variant in the test sample into a trained neural network model to obtain the prediction result corresponding to each mitochondrial genetic variant in the test sample, and sorts the prediction results.

[0012] This method obtains the mutation results after gene sequencing; based on the mutation results and the HPO list, it acquires the feature vector corresponding to each mitochondrial genetic variant in the test sample; and constructs a feature vector for mitochondrial genetic variants tailored to their characteristics, thus addressing the problem that existing ranking algorithms are not entirely suitable for mitochondrial variant analysis. This method inputs the feature vector corresponding to each mitochondrial genetic variant in the test sample into a trained neural network model to obtain the prediction result for each mitochondrial genetic variant in the test sample, and then ranks the prediction results. The neural network model learns from the feature vectors corresponding to mitochondrial genetic variants obtained from the mutation results and the HPO list, obtaining a large number of correlations between mitochondrial genetic variant feature vectors and clinical phenotypes, thereby improving the correlation between ranking results and clinical symptoms. By ranking the prediction results, this method simplifies the repetitive interpretation process for each variant in traditional genetic variant analysis to first ranking and then interpreting a small number of ranked variants, thus improving the efficiency of mitochondrial genetic variant analysis.

[0013] According to some embodiments of the present invention, the feature vector corresponding to each mitochondrial genetic variant includes: a variant type feature vector for converting variant types into binary vectors; an amino acid change feature vector for converting standard variant representations into vectors; a population frequency feature vector for representing the maximum frequency of each variant in multiple population databases; a variant heterogeneity feature vector for representing the ratio of variant mitochondrial DNA; a maternal genetic background feature vector for representing whether the mother carries the same variant; a database inclusion feature vector for representing whether the variant is included in various mitochondrial databases; a prediction scoring feature vector for scoring the harmfulness prediction scores of different variant types; and a phenotypic association feature vector for representing the matching degree between the variant's associated phenotype and the user-input HPO list.

[0014] According to some embodiments of the present invention, the step of converting the mutation type into a binary vector includes:

[0015] Obtain the number of types of mutations in the mitochondrial genetic variation;

[0016] The mutation type is converted into a binary vector according to the number of mutation types; wherein the dimension value of the binary vector is equal to the number of mutation types.

[0017] According to some embodiments of the present invention, the scoring of harmfulness prediction scores for different variant types includes:

[0018] Obtain the harmfulness prediction score for each type of mitochondrial genetic variation;

[0019] The harmfulness prediction score for each of the aforementioned variant types was normalized:

[0020] x n =xx min / x max -x min

[0021] Where, x n Let x represent the normalized score of the hazard prediction score, and let x represent the hazard prediction score. min x represents the minimum value of the harmfulness prediction score. max This represents the maximum value of the hazard prediction score;

[0022] The harmfulness prediction score for each type of mitochondrial genetic variation is scored as follows:

[0023] S = [x1, x2, ..., x n ]

[0024] Where S represents the score, [x1,x2,...,x n [] represents the set of normalized scores for harmfulness prediction scores of different variant types;

[0025] If the score for any mutation type in the set is missing, the score for that mutation type is recorded as 0.

[0026] According to some embodiments of the present invention, calculating the variational heterogeneity feature vector includes:

[0027] After removing repetitive sequences, the sequencing depth of the variants is obtained;

[0028] Calculate the heterogeneity feature vector of the variant based on the sequencing depth of the variant:

[0029] H=D v / D total

[0030] Among them, D v D represents the sequencing depth of the variant. total H represents the total sequencing depth, and H represents the variant heterogeneity feature vector.

[0031] According to some embodiments of the present invention, before inputting the feature vector corresponding to each mitochondrial genetic variation in the test sample into the trained neural network model, the neural network-based mitochondrial genetic variation ranking method further includes:

[0032] Obtain training samples with multiple known mutation results;

[0033] Based on the mutation results and the HPO list, obtain the feature vector of mitochondrial genetic variation corresponding to each mutation in the training samples of each known mutation result;

[0034] The feature vector corresponding to each mitochondrial genetic variation in the training samples is calibrated to obtain the calibration results.

[0035] Based on the feature vector and calibration results corresponding to each mitochondrial genetic variation in the training samples, a training set and a validation set are constructed.

[0036] A pre-defined neural network model is trained using the training set and the validation set to obtain a trained neural network model.

[0037] According to some embodiments of the present invention, the neural network model includes an input layer, an intermediate layer, and an output layer, wherein the number of nodes in the input layer is the same as the total dimension of the feature vectors of the samples, the number of nodes in the intermediate layer is greater than the number of nodes in the input layer, and the output layer includes one node.

[0038] Secondly, embodiments of the present invention also provide a mitochondrial genetic variation sequencing system based on a neural network, the mitochondrial genetic variation sequencing system based on a neural network comprising:

[0039] The mutation result acquisition unit is used to acquire the mutation results after gene sequencing;

[0040] The feature vector acquisition unit is used to acquire the feature vector of mitochondrial genetic variation corresponding to each variation in the sample to be tested based on the variation results and the HPO list.

[0041] The prediction result acquisition unit is used to input the feature vector corresponding to each mitochondrial genetic variation in the test sample into the trained neural network model, obtain the prediction result corresponding to each mitochondrial genetic variation in the test sample, and sort the prediction results.

[0042] Thirdly, embodiments of the present invention also provide a mitochondrial genetic variation sequencing device based on a neural network, including at least one control processor and a memory for communicatively connecting to the at least one control processor; the memory stores instructions executable by the at least one control processor, the instructions being executed by the at least one control processor to enable the at least one control processor to execute a mitochondrial genetic variation sequencing method based on a neural network as described above.

[0043] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions for causing a computer to execute a neural network-based mitochondrial genetic variation sequencing method as described above.

[0044] It is understood that the beneficial effects of the second to fourth aspects compared with the related technologies are the same as the beneficial effects of the first aspect compared with the related technologies. Please refer to the relevant description in the first aspect above, which will not be repeated here. Attached Figure Description

[0045] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0046] Figure 1 This is a flowchart of a mitochondrial genetic variation sorting method based on a neural network according to an embodiment of the present invention;

[0047] Figure 2 This is a structural diagram of a mitochondrial genetic variation sequencing system based on a neural network, according to an embodiment of the present invention. Detailed Implementation

[0048] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0049] In the description of this invention, the use of terms such as "first," "second," etc., is for the purpose of distinguishing technical features only and should not be construed as indicating or implying relative importance, or implicitly indicating the number of technical features indicated, or implicitly indicating the order of the technical features indicated.

[0050] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, etc., are based on the orientation or positional relationship shown in the drawings and are only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0051] In the description of this invention, it should be noted that, unless otherwise explicitly defined, terms such as "setting," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0052] High-throughput sequencing technology can obtain complete mitochondrial variants at low cost. However, from an efficiency perspective, it is impossible to examine hundreds or thousands of variant results one by one and classify them for pathogenicity. Existing genetic variant ranking methods, such as Exomiser, rely on weighted averaging of several pathogenicity assessment algorithms. However, the weighting is highly subjective and becomes increasingly difficult to balance as the number of input parameters increases. Furthermore, Exomiser selects parameters specifically for autosomal variants. There is a lack of integration among pathogenicity scoring methods for mitochondrial variants. For example, APOGEE can predict the harmfulness of missense mutations, and MitoTip can predict the harmfulness of different types of mutations, but their scores are not comparable and cannot be directly ranked together. In addition, genetic variant ranking needs to consider phenotypic matching in addition to calculating variant pathogenicity, so that the analysis results are closer to each specific case.

[0053] To address the aforementioned issues, this invention obtains the mutation results after gene sequencing; based on the mutation results and the HPO list, it acquires the feature vector corresponding to each mitochondrial genetic variant in the test sample; and constructs a feature vector for mitochondrial genetic variants tailored to their characteristics, thereby solving the problem that existing sorting algorithms are not entirely suitable for mitochondrial variant analysis. This invention inputs the feature vector corresponding to each mitochondrial genetic variant in the test sample into a trained neural network model to obtain the prediction result for each mitochondrial genetic variant in the test sample, and then sorts the prediction results. The neural network model learns from the feature vectors corresponding to mitochondrial genetic variants obtained from the mutation results and the HPO list, obtaining a large number of feature vectors of mitochondrial genetic variants and their correlation with clinical phenotypes, thereby improving the correlation between the sorting results and clinical symptoms. This invention simplifies the repetitive interpretation process for each variant in traditional genetic variant analysis by sorting the prediction results, to first sorting and then interpreting a small number of sorted variants, thus improving the efficiency of mitochondrial genetic variant analysis.

[0054] Reference Figure 1This invention provides a neural network-based method for sequencing mitochondrial genetic variations. This neural network-based method for sequencing mitochondrial genetic variations includes:

[0055] Step S100: Obtain the mutation results after gene sequencing.

[0056] Specifically, the mutation results obtained after gene sequencing are obtained, and these mutation results are annotated. These annotations can be obtained using existing software tools. For example, the VCF format file can be imported into the existing GTX.Digest annotation software (https: / / digest.gtxlab.com / ) to obtain the mutation result annotation file.

[0057] Step S200: Based on the mutation results and the HPO list, obtain the feature vector corresponding to each mitochondrial genetic variation in the sample to be tested.

[0058] Specifically, based on the mutation results obtained after gene sequencing in step S100, and after annotating the mutation results, input parameters that are not applicable to mitochondrial mutations need to be excluded. Input parameters that need to be explicitly excluded from the test sample include: parameters used to determine autosomal dominant or recessive inheritance patterns, and parameters used to determine the gene mutation frequency of mutation hotspot regions.

[0059] Specifically:

[0060] Exclude parameters used to determine autosomal dominant or recessive inheritance patterns, including:

[0061] Examine the genetic pattern parameters in the variation results. If there are fields related to genetic patterns such as INHERITANCE, or other fields containing information on dominant, recessive, or sex-linked inheritance, they should be deleted or replaced with blank values ​​and explicitly not used in subsequent steps. Since mitochondrial inheritance is maternally induced, autosomal genetic pattern parameters are not applicable to mitochondrial analysis. However, if the annotation method for autosomal variations is incorrectly applied in step S100, erroneous genetic pattern parameters may be generated. This step effectively eliminates erroneous parameters.

[0062] Exclude gene variation frequency parameters used to determine variation hotspots, including:

[0063] Examine the gene variation frequency parameters in the obtained variation results. If any fields related to gene variation frequency or other fields marking high-frequency variant genes or high-frequency pathogenic variant genes exist, they should be deleted or replaced with blank values ​​and explicitly not used in subsequent steps. Although this parameter is useful for autosomal variation analysis, mitochondria lack gene recombination mechanisms and nucleosomes, resulting in a higher mutation frequency than autosomes. Therefore, mitochondrial gene variation rates are not suitable as a reference factor for analysis. This step can effectively avoid misleading subsequent analyses.

[0064] After excluding input parameters that are not applicable to mitochondrial variations, the input parameters to be processed are obtained. Input the input parameters to be processed and the HPO list. After preprocessing the input parameters to be processed and the HPO list, the feature vector corresponding to each mitochondrial genetic variation in the test sample is obtained. The feature vector corresponding to each mitochondrial genetic variation in the test sample includes: a variation type feature vector representing the conversion of the variation type into a binary vector; an amino acid change feature vector representing the conversion of the standard variation representation into a vector; a population frequency feature vector representing the maximum frequency of each variation in multiple population databases; a variation heterogeneity feature vector representing the ratio of mitochondrial DNA variations; a maternal genetic background feature vector representing whether the mother carries the same variation; a database inclusion feature vector representing whether the variation is included in various mitochondrial databases; a prediction scoring feature vector for scoring the harmfulness prediction score of different variation types; and a phenotypic association feature vector representing the matching degree between the associated phenotype of the variation and the user-input HPO list. Specifically:

[0065] Construct a mutation type feature vector (T) to represent the mutation type transformed into a binary vector, including:

[0066] Obtain the number of mutation types in mitochondrial genetic variation; convert each mutation type into a binary vector to obtain a mutation type feature vector (T); where the dimension of the binary vector is equal to the number of mutation types. Specifically:

[0067] The input parameter is the mutation type. Assuming there are N mutation types in total, the input parameter is converted into a binary vector with N dimensions. For example, when N=4, "missense mutation" can be converted into a four-dimensional vector [1,0,0,0], "tRNA mutation" can be converted into [0,1,0,0], "insertion / deletion mutation" can be converted into [0,0,1,0], and "nonsense mutation" can be converted into [0,0,0,1].

[0068] Construct an amino acid change feature vector (A) to represent the standard variation representation as a vector, including:

[0069] The input parameter is the standard variant representation (HGVS). The HGVS is converted into a vector representation, where, as shown in Table 1, amino acids are converted into numerical numbers according to the reference table. For example, the HGVS of a certain amino acid variant is p.W163G, which can be converted into the vector [8,9], where amino acid W is denoted as 8 and amino acid G as 9. If the variant type is not a "missense mutation", this vector is denoted as [0,0].

[0070] Table 1

[0071]

[0072]

[0073] Construct a population frequency feature vector (F) to represent the maximum frequency of each variant in multiple population databases, including:

[0074] The input parameters are several population frequencies, such as population frequencies in the population databases gnomAD (The Genome Aggregation Database) and GenBank (US Gene Bank). The maximum frequency of each variant in multiple population databases is obtained, with a value range of [0,1].

[0075] Construct a feature vector (H) representing the heterogeneity of variant mitochondrial DNA ratios, including:

[0076] After removing repetitive sequences, the sequencing depth of the variants is obtained;

[0077] Calculate the variant heterogeneity feature vector based on the sequencing depth of the variant:

[0078] H=D v / D total

[0079] Among them, D v D represents the sequencing depth of the variant. total This represents the total sequencing depth, and H represents the variance heterogeneity feature vector. For example:

[0080] The input parameter is the sequencing depth of the variant, where H is the sequencing depth of the variant after removing repetitive sequences, and D is the sequencing depth of the variant. v D percentage of total sequencing depth total The ratio represents the proportion of the number of copies of the mutation at that site to the number of copies of mitochondrial DNA, and the value ranges from [0,1].

[0081] Construct a maternal genetic background feature vector (M) to represent whether the mother carries the same variant, including:

[0082] The input parameter is whether the mother of each sample carries the same genetic variant. If there is no identical variant, it is set to 0; if there is an identical variant, it is set to the heterogeneity of variants in the mother's samples (H). The value range is [0,1]. This feature is optimized for the maternal inheritance characteristics of mitochondrial genetic variants, which is different from the inheritance pattern of autosomal variants.

[0083] A database inclusion feature vector (D) is constructed to represent whether a variant is included in various mitochondrial databases, including but not limited to whether it is included in OMIM (Online Mendelian Inheritance in Man), commercial mitochondrial panels (e.g., GeneDx 65 Panel), and the PubMed open-source database. The input parameters are the inclusion results from these databases, and the output is a vector of Brownian values. Included variants are denoted as 1, and not included variants as 0. This feature is optimized for databases specific to mitochondrial genetic variations.

[0084] Construct a predictive scoring feature vector (S) for scoring the harmfulness prediction scores of different variant types, including:

[0085] Obtain the harmfulness prediction score for each type of mitochondrial genetic variation;

[0086] The harmfulness prediction score for each variant type was normalized:

[0087] x n =xx min / x max -x min

[0088] Where, x n Let x represent the normalized score of the hazard prediction score. min x represents the minimum value of the hazard prediction score. max This represents the maximum value of the hazard prediction score;

[0089] The harmfulness prediction score for each type of mitochondrial genetic variation was scored as follows:

[0090] S = [x1, x2, ..., x n ]

[0091] Where S represents the score, [x1,x2,...,x n [] represents the set of normalized scores for harmfulness prediction scores of different variant types;

[0092] If a score for any mutation type is missing in the set, the score for that mutation type is recorded as 0. For example:

[0093] The harmfulness prediction scores (x) for different types of mitochondrial variants are integrated into a floating-point vector. Examples include APOGEE (https: / / doi.org / 10.1371%2Fjournal.pcbi.1005628), MitoTip (https: / / doi.org / 10.1371%2Fjournal.pcbi.1005867), and HmtVar (https: / / doi.org / 10.1093%2Fnar%2Fgky1024). Each score is normalized by subtracting the lower limit from the harmfulness prediction score and then dividing by the difference between the upper and lower limits of the harmfulness prediction score.

[0094] Construct a phenotypic association feature vector (P) to represent the matching degree between the associated phenotypic of the variant and the user-input HPO list, including:

[0095] The input parameters are the variant association phenotype (HPOmut) and the user-inputted list of HPOs (HPOinput). The output format is a two-dimensional vector [isHPO, pHPO]. Here, isHPO indicates whether HPOmut intersects with HPOinput, taking a value of 0 or 1; pHPO represents the percentage of this intersection with HPOinput, taking a value in the range [0,1]. The specific phenotypic association scoring formula is as follows:

[0096] pHPO=|HPOmut∩HPOinput| / |HPOinput|

[0097] It should be noted that, as one implementation method, the variant gene name can be extracted from the input file, and the association between the variant name and the phenotype can be extracted from the OMIM database. This implementation method is prior art and will not be described in detail in this embodiment.

[0098] Step S300: Input the feature vector corresponding to each mitochondrial genetic variation in the test sample into the trained neural network model to obtain the prediction result corresponding to each mitochondrial genetic variation in the test sample, and sort the prediction results.

[0099] Specifically, before inputting the feature vector corresponding to each mitochondrial genetic variation in the test sample into the trained neural network model, the following steps are also included:

[0100] Obtain training samples from multiple known variant results. Obtain variant results from several patients and several healthy individuals from user or public databases as training samples.

[0101] Based on the mutation results and the HPO list, obtain the feature vector of mitochondrial genetic variation corresponding to each mutation in the training samples for each known mutation result. Following the method in step S200, the feature vectors corresponding to each mitochondrial genetic variation in the training samples include: a mutation type feature vector representing the conversion of mutation type into a binary vector; an amino acid change feature vector representing the conversion of standard mutation representation into a vector; a population frequency feature vector representing the maximum frequency of each mutation in multiple population databases; a mutation heterogeneity feature vector representing the mitochondrial DNA ratio of the mutation; a maternal genetic background feature vector representing whether the mother carries the same mutation; a database inclusion feature vector representing whether the mutation is included in various mitochondrial databases; a prediction scoring feature vector for scoring the harmfulness prediction score of different mutation types; and a phenotypic association feature vector representing the matching degree between the mutation's associated phenotype and the user-input HPO list.

[0102] The feature vectors corresponding to each mitochondrial genetic variant in the training samples are labeled to obtain the labeling results. The output of the feature vector corresponding to the pathogenic variant of the patient is labeled as 1, and the output of the feature vectors corresponding to all other variants is labeled as 0.

[0103] Based on the feature vectors and calibration results corresponding to each mitochondrial genetic variation in the training samples, a training set and a validation set are constructed. 70% of the samples with calibrated output results are used as the training set, and the remaining 30% are used as the validation set.

[0104] A pre-defined neural network model is trained using a training set and a validation set to obtain a well-trained neural network model.

[0105] It should be noted that the preset neural network model in this embodiment is a BPNN (BackPropagation Neural Network) multilayer neural network, but this preset neural network model can be modified according to actual needs, and this embodiment does not impose specific limitations.

[0106] In this embodiment, the preset neural network model includes an input layer, intermediate layers, and an output layer. The number of nodes in the input layer is the same as the total dimension of the sample's feature vectors, the number of nodes in the intermediate layers is greater than the number of nodes in the input layer, and the output layer includes one node. The purpose of choosing the BPNN network structure in this embodiment is to fully utilize the input parameters while also considering scenarios with limited data. Larger models can theoretically handle more complex situations but require larger datasets, while smaller models would lose input features.

[0107] It should be noted that in this embodiment, the number of intermediate layer nodes is set to twice the number of input layer nodes, but the number of intermediate layer nodes and the number of input layer nodes can be modified according to actual needs, and this embodiment does not impose specific limitations.

[0108] The feature vector corresponding to each mitochondrial genetic variant in the test sample is input into a trained neural network model to obtain the prediction result for each mitochondrial genetic variant in the test sample, and the prediction results are then sorted. Specifically:

[0109] The feature vectors corresponding to each mitochondrial genetic variant in the test sample obtained in step S200 are input into the trained neural network model to obtain a prediction score for each variant. The prediction score is a floating-point number in the range [0,1]. All variant results are sorted in descending order of prediction score.

[0110] It should be noted that this embodiment uses descending order, but the sorting in this embodiment can be modified according to actual needs, and this embodiment does not impose any specific limitations.

[0111] In this embodiment, the input feature vector of the neural network model is optimized to address the characteristics of mitochondrial variations, eliminating input parameters unsuitable for mitochondrial variations. This solves the problem that existing ranking algorithms are not fully applicable to mitochondrial variation analysis. For example, a unified ranking score can be provided for all variation types, requiring only individual review and interpretation according to the unified ranking, without needing to examine each variation type separately. A ranking method based on a BPNN neural network is introduced, simplifying the repetitive interpretation process for each variation in traditional variation analysis to first ranking and then interpreting the few variations ranked first. This improves the efficiency of genetic variation analysis and reduces the subjectivity of weights in other ranking methods. The HPO list in this embodiment provides standard vocabulary and terminology used to describe phenotypic abnormalities in human diseases. Each HPO describes a phenotypic abnormality; therefore, the HPO list can reflect the clinical phenotype of the sample. Thus, the ranking method in this embodiment integrates two factors: phenotypic matching degree (i.e., phenotypic associated feature vectors) and variation pathogenicity (i.e., feature vectors other than phenotypic associated feature vectors), improving the correlation between ranking results and clinical symptoms and variation pathogenicity.

[0112] Reference Figure 2 This invention also provides a mitochondrial genetic variation ranking system based on a neural network. This system includes a variation result acquisition unit 100, a feature vector acquisition unit 200, and a prediction result acquisition unit 300, wherein:

[0113] The mutation result acquisition unit 100 is used to acquire the mutation results after gene sequencing.

[0114] The feature vector acquisition unit 200 is used to acquire the feature vector of each mitochondrial genetic variation in the sample to be tested based on the mutation results and the HPO list.

[0115] The prediction result acquisition unit 300 is used to input the feature vector corresponding to each mitochondrial genetic variation in the test sample into the trained neural network model, obtain the prediction result corresponding to each mitochondrial genetic variation in the test sample, and sort the prediction results.

[0116] It should be noted that since the mitochondrial genetic variation sorting system based on neural networks in this embodiment is based on the same inventive concept as the mitochondrial genetic variation sorting method based on neural networks described above, the corresponding content in the method embodiment is also applicable to this system embodiment, and will not be described in detail here.

[0117] This invention also provides a neural network-based mitochondrial genetic variation sequencing device, comprising: at least one control processor and a memory for communicatively connecting to the at least one control processor.

[0118] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0119] The non-transient software program and instructions required to implement the neural network-based mitochondrial genetic variation sorting method of the above embodiments are stored in memory. When executed by the processor, the neural network-based mitochondrial genetic variation sorting method of the above embodiments is executed, for example, the method described above is executed. Figure 1 The method steps S100 to S300.

[0120] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] This invention also provides a computer-readable storage medium storing computer-executable instructions. These instructions are executed by one or more control processors, causing the processors to perform a neural network-based mitochondrial genetic variation sequencing method described in the above-described method embodiments. For example, they can execute the methods described above. Figure 1 The functions of steps S100 to S300 in the method.

[0122] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0123] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for sequencing mitochondrial genetic variations based on neural networks, characterized in that, The neural network-based mitochondrial genetic variation sequencing method includes: Obtain the mutation results after gene sequencing; Based on the mutation results and the HPO list, the feature vector corresponding to each mitochondrial genetic variant in the test sample is obtained. The feature vector corresponding to each mitochondrial genetic variant includes: a variant type feature vector for converting the variant type into a binary vector; an amino acid change feature vector for converting the standard variant representation into a vector; a population frequency feature vector for representing the maximum frequency of each variant in multiple population databases; a variant heterogeneity feature vector for representing the mitochondrial DNA ratio of the variant; a maternal genetic background feature vector for representing whether the mother carries the same variant; a database inclusion feature vector for representing whether the variant is included in various mitochondrial databases; a prediction scoring feature vector for scoring the harmfulness prediction score of different variant types; and a phenotypic association feature vector for representing the matching degree between the variant's associated phenotype and the user-input HPO list. The feature vector corresponding to each mitochondrial genetic variant in the test sample is input into the trained neural network model to obtain the prediction result corresponding to each mitochondrial genetic variant in the test sample, and the prediction result is sorted.

2. The mitochondrial genetic variation sequencing method based on neural networks according to claim 1, characterized in that, The process of converting mutation types into binary vectors includes: Obtain the number of types of mutations in the mitochondrial genetic variation; The mutation type is converted into a binary vector according to the number of mutation types; wherein the dimension value of the binary vector is equal to the number of mutation types.

3. The mitochondrial genetic variation sequencing method based on neural networks according to claim 1, characterized in that, The scoring of the harmfulness prediction scores for different variant types includes: Obtain the harmfulness prediction score for each type of mitochondrial genetic variation; The harmfulness prediction score for each of the aforementioned variant types was normalized: in, This represents the normalized score of the harmfulness prediction score. This represents the hazard prediction score. This represents the minimum value of the harmfulness prediction score. This represents the maximum value of the hazard prediction score; The harmfulness prediction score for each type of mitochondrial genetic variation is scored as follows: in, Indicates the score. A set of normalized scores representing the harmfulness prediction scores for different types of variants; If the score for any mutation type in the set is missing, the score for that mutation type is recorded as 0.

4. The mitochondrial genetic variation sequencing method based on neural networks according to claim 1, characterized in that, Calculating the variability heterogeneity feature vector includes: After removing repetitive sequences, the sequencing depth of the variants is obtained; Calculate the heterogeneity feature vector of the variant based on the sequencing depth of the variant: in, This indicates the sequencing depth of the variant. H represents the total sequencing depth, and H represents the variant heterogeneity feature vector.

5. The mitochondrial genetic variation sequencing method based on neural networks according to claim 1, characterized in that, Before inputting the feature vector corresponding to each mitochondrial genetic variant in the test sample into the trained neural network model, the neural network-based mitochondrial genetic variant ranking method further includes: Obtain training samples with multiple known mutation results; Based on the mutation results and the HPO list, obtain the feature vector of mitochondrial genetic variation corresponding to each mutation in the training samples of each known mutation result; The feature vector corresponding to each mitochondrial genetic variation in the training samples is calibrated to obtain the calibration results. Based on the feature vector and calibration results corresponding to each mitochondrial genetic variation in the training samples, a training set and a validation set are constructed. A pre-defined neural network model is trained using the training set and the validation set to obtain a trained neural network model.

6. The mitochondrial genetic variation sequencing method based on neural networks according to claim 1, characterized in that, The neural network model includes an input layer, an intermediate layer, and an output layer. The number of nodes in the input layer is the same as the total dimension of the feature vectors of the samples. The number of nodes in the intermediate layer is greater than the number of nodes in the input layer. The output layer includes one node.

7. A mitochondrial genetic variation sequencing system based on a neural network, characterized in that, The neural network-based mitochondrial genetic variation sequencing system includes: The mutation result acquisition unit is used to acquire the mutation results after gene sequencing; The feature vector acquisition unit is used to acquire the feature vector of mitochondrial genetic variation corresponding to each variation in the test sample based on the variation results and the HPO list. The feature vector corresponding to each mitochondrial genetic variation includes: a variation type feature vector for converting the variation type into a binary vector; an amino acid change feature vector for converting the standard variation representation into a vector; a population frequency feature vector for representing the maximum frequency of each variation in multiple population databases; a variation heterogeneity feature vector for representing the ratio of mitochondrial DNA of the variation; a maternal genetic background feature vector for representing whether the mother carries the same variation; a database inclusion feature vector for representing whether the variation is included in various mitochondrial databases; a prediction scoring feature vector for scoring the harmfulness prediction score of different variation types; and a phenotypic association feature vector for representing the matching degree between the associated phenotype of the variation and the HPO list input by the user. The prediction result acquisition unit is used to input the feature vector corresponding to each mitochondrial genetic variation in the test sample into the trained neural network model, obtain the prediction result corresponding to each mitochondrial genetic variation in the test sample, and sort the prediction results.

8. A mitochondrial genetic variation sequencing device based on a neural network, characterized in that, It includes at least one control processor and a memory for communicatively connecting to the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform the neural network-based mitochondrial genetic variation sequencing method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer to perform the neural network-based mitochondrial genetic variation sequencing method as described in any one of claims 1 to 6.

Images