Methods, devices, electronic equipment, and computer-readable media for identifying heterozygous frameshift mutations in gene editing target site sequencing results based on machine learning.
By using machine learning methods to automatically identify heterozygous frameshift mutations in gene editing target site sequencing results, the problem of low efficiency in manual identification in existing technologies is solved, and efficient and accurate automated identification is achieved.
Patent Information
- Application Number
- CN202411230365.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2044-09-04
AI Technical Summary
In existing technologies, the identification of heterozygous frameshift mutations in gene editing target site sequencing results relies on manual identification, resulting in low controllability and being time-consuming and labor-intensive, lacking efficient automated identification methods.
Using machine learning methods, a classification model is constructed by extracting sample feature values from sequencing results to automatically identify heterozygous frameshift mutations. This includes sample feature extraction, data standardization, and model training, achieving automated and efficient identification of heterozygous frameshift mutations.
It improves the accuracy and efficiency of heterozygous frameshift mutation identification, reduces labor costs, shortens data analysis time, and provides fast and accurate classification results.
Smart Images

Figure CN119229964B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of detecting heterozygous frameshift mutations in gene editing, and particularly to a method, apparatus, electronic device, and computer-readable medium for determining heterozygous frameshift mutations in gene editing target site sequencing results based on machine learning. Background Technology
[0002] Gene editing technology, especially the CRISPR / Cas9 gene editing system, has been widely used in gene editing breeding of plants and animals. Detection of target site editing is essential in large-scale gene editing breeding. PCR amplification and first-generation sequencing technologies are among the important methods for detecting whether target site editing has occurred.
[0003] Current sequencing result identification methods primarily rely on manual identification of bimodal features within the target site region of the sequencing peak diagram to determine whether a sample has a heterozygous frameshift mutation. This is extremely labor-intensive in practical applications. Furthermore, manual identification of heterozygous frameshift mutations in sequencing peak diagrams depends on individual experience and the level of care taken in the judgment, resulting in low predictability and being time-consuming and labor-intensive. Clearly, existing classification methods have multiple drawbacks. Therefore, there is an urgent need to provide a batch-accurate method for determining the occurrence of frameshift mutations based on peak diagrams, in order to improve identification efficiency and reduce manual costs. Summary of the Invention
[0004] To address the technical problem that the determination of heterozygous frameshift mutations in existing technologies relies on manual methods, resulting in low controllability and being time-consuming and labor-intensive, this invention provides a method, apparatus, electronic device, and computer-readable medium for determining heterozygous frameshift mutations in gene editing target site sequencing results based on machine learning, which can improve identification efficiency and reduce the cost of manual identification.
[0005] To achieve the above-mentioned technical objectives, the technical solution of the present invention is as follows:
[0006] A machine learning-based method for determining heterozygous frameshift mutations in gene editing target site sequencing results includes the following steps:
[0007] Step 1: Extract the main peak sequence, secondary peak sequence, and consensus sequence from the sequencing results labeled with heterozygous frameshift mutation tags. After determining the target site region, extract the sequence feature values of the sequences surrounding the target site region to form sample features. The sample features include:
[0008] Feature 1) The length of homology between the consistent sequence and the target site sequence;
[0009] Feature 2) The number of degenerate bases upstream of the start position and downstream of the end position of the target site region in the consistent sequence within a preset range;
[0010] Feature 3) The difference and ratio of the number of degenerate bases downstream of the termination position and upstream of the start position in the target site region within the same preset range in the consistent sequence;
[0011] Feature 4) Consistency between the secondary peak sequence and the primary peak sequence within the same preset range upstream of the start position and downstream of the end position in the target site region;
[0012] Step 2: Construct a mapping from sample features to sample labels, and standardize the sample features based on a supervised classification algorithm in machine learning to establish a dataset for training. Then train the model and evaluate the accuracy of the trained model. Repeat the training until a model that meets the preset accuracy requirements is obtained.
[0013] Step 3: Extract the sequence feature values of the sequencing results of the sample to be predicted and input them into the model obtained in Step 2 to obtain the judgment result of whether the sample to be predicted has undergone heterozygous frameshift mutation.
[0014] The method described above, in step 1, obtains the sequencing results labeled with heterozygous frameshift mutations through the following steps:
[0015] First-generation sequencing was performed on the gene editing target site region, and then the sequencing results were used to determine whether a heterozygous frameshift mutation occurred and to label it.
[0016] The method further includes, before performing first-generation sequencing on the gene editing target site region, a step of designing PCR primers in regions at least 200 bp upstream and downstream of the gene editing target site and performing sequence amplification.
[0017] The method described above, which determines whether a heterozygous frameshift mutation has occurred based on sequencing results, is based on the following steps:
[0018] Based on the target site sequence in the sequencing results file, the target site is located by finding a single peak on one side. The target site is then judged according to its location: if there are two peaks in the middle of the target site that continue to the other side, it is judged as a heterozygous frameshift mutation; otherwise, it is judged as a non-heterozygous frameshift mutation.
[0019] The method described above, when labeling, labels heterozygous frameshift mutations and non-heterozygous frameshift mutations as one of the categories in the binary classification label.
[0020] In the method described above, step 1 involves extracting the main peak sequence, secondary peak sequence, and consensus sequence through the following steps:
[0021] The highest peak in the sequencing results is taken as the main peak. Then, all sequences lower than 0.2 times the main peak are filtered out as background interference. The highest peak among the remaining sequences is taken as the secondary peak. The sequence identity of the two sequences is obtained by comparing the main peak sequence and the secondary peak sequence.
[0022] In the method described above, step 1 involves determining the target site region by finding regions in the consistent sequence that are homologous to the target site sequence, and determining the start and end positions of the target site region in the consistent sequence.
[0023] The method, in step 1, includes the following features: Feature 2) of the sample features:
[0024] ① The number of degenerate bases within 30-80 bp upstream of the start position of the target site region in the homogeneous sequence;
[0025] ② The number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence.
[0026] In the method described above, in step 1, feature 3) of the sample features includes:
[0027] ① The difference between the number of degenerate bases within 30-80 bp downstream of the termination position and the number of degenerate bases within 30-80 bp upstream of the start position in the target site region of the homogeneous sequence;
[0028] ② The ratio of the number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence to the number of degenerate bases within 30-80 bp upstream of the start position.
[0029] The method, in step 1, includes the following features: Feature 4) of the sample features:
[0030] ① The consistency between the 30-80bp upstream of the target site region in the secondary peak sequence and the 30-80bp upstream of the target site region in the main peak sequence;
[0031] ② The consistency between the 30-80bp downstream of the target site region termination position in the secondary peak sequence and the 30-80bp downstream of the target site region termination position in the main peak sequence.
[0032] In the method described above, step 2 involves data standardization through mean-variance normalization, using the following formula:
[0033]
[0034] Where x scale Here, x represents the standardized eigenvalues. mean s is the characteristic mean, and s is the characteristic variance.
[0035] The method described above, in step 2, includes training the model and evaluating the accuracy of the trained model, which includes:
[0036] The obtained dataset is used for model training, with 80% of the data used as the training set to train the model and the remaining 20% used as the validation set to obtain evaluation results.
[0037] A machine learning-based device for determining heterozygous frameshift mutations in gene editing target site sequencing results includes the following modules:
[0038] The sample feature extraction module is used to extract the main peak sequence, secondary peak sequence, and consensus sequence from sequencing results that have been labeled with heterozygous frameshift mutation tags, and to extract sequence feature values of sequences surrounding the target site region after determining the target site region, so as to form sample features; wherein the sample features include:
[0039] Feature 1) The length of homology between the consistent sequence and the target site sequence;
[0040] Feature 2) The number of degenerate bases upstream of the start position and downstream of the end position of the target site region in the consistent sequence within a preset range;
[0041] Feature 3) The difference and ratio of the number of degenerate bases downstream of the termination position and upstream of the start position in the target site region within the same preset range in the consistent sequence;
[0042] Feature 4) Consistency between the secondary peak sequence and the primary peak sequence within the same preset range upstream of the start position and downstream of the end position in the target site region;
[0043] The model building module is used to construct a mapping from sample features to sample labels, and to standardize the sample features based on supervised classification algorithms in machine learning, thereby establishing a dataset for training. Then, the model is trained and the accuracy of the trained model is evaluated. The training is repeated until a model that meets the preset accuracy requirements is obtained.
[0044] The heterozygous frameshift mutation determination module is used to extract the sequence feature values of the sequencing results of the sample to be predicted, and input them into the model obtained by the model building module, so as to obtain the determination result of whether the sample to be predicted has undergone frameshift mutation.
[0045] In the aforementioned apparatus, the sample feature extraction module, the sequencing results labeled with heterozygous frameshift mutations involve first-generation sequencing of the gene editing target site region, followed by determination of whether a heterozygous frameshift mutation has occurred and subsequent labeling based on the sequencing results. Before first-generation sequencing of the gene editing target site region, PCR primers are designed in regions at least 200 bp upstream and downstream of the gene editing target site, and sequence amplification is performed. Determining whether a heterozygous frameshift mutation has occurred based on the sequencing results file involves locating the target site by finding a single peak on one side of the target site, and then judging based on the located target site: if a target site has a double peak in the middle that extends to the other side, it is judged as a heterozygous frameshift mutation; otherwise, it is judged as a non-heterozygous frameshift mutation. During labeling, heterozygous frameshift mutations and non-heterozygous frameshift mutations are respectively labeled as one category in a binary classification label.
[0046] In the aforementioned apparatus, the sample feature extraction module extracts the main peak sequence, secondary peak sequence, and consistency sequence through the following steps:
[0047] The highest peak in the sequencing results is taken as the main peak. Then, all sequences lower than 0.2 times the main peak are filtered out as background interference. The highest peak among the remaining sequences is taken as the secondary peak. The sequence identity of the two sequences is obtained by comparing the main peak sequence and the secondary peak sequence.
[0048] In the aforementioned apparatus, the sample feature extraction module determines the target site region by finding regions in the consistent sequence that are homologous to the target site sequence, and by determining the start and end positions of the target site region in the consistent sequence.
[0049] In the aforementioned apparatus, in the aforementioned sample feature extraction module, feature 2) of the sample features includes:
[0050] ① The number of degenerate bases within 30-80 bp upstream of the start position of the target site region in the homogeneous sequence;
[0051] ② The number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence;
[0052] Feature 3) includes:
[0053] ① The difference between the number of degenerate bases within 30-80 bp downstream of the termination position and the number of degenerate bases within 30-80 bp upstream of the start position in the target site region of the homogeneous sequence;
[0054] ② The ratio of the number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence to the number of degenerate bases within 30-80 bp upstream of the start position.
[0055] Feature 4) includes:
[0056] ① The consistency between the 30-80bp upstream of the target site region in the secondary peak sequence and the 30-80bp upstream of the target site region in the main peak sequence;
[0057] ② The consistency between the 30-80bp downstream of the target site region termination position in the secondary peak sequence and the 30-80bp downstream of the target site region termination position in the main peak sequence.
[0058] In the aforementioned apparatus and model building module, data standardization is performed through mean-variance normalization, using the following formula:
[0059]
[0060] Where x scale Here, x represents the standardized eigenvalues. mean s is the characteristic mean, and s is the characteristic variance.
[0061] In the aforementioned apparatus, the model building module includes training the model and evaluating the accuracy of the trained model, which includes:
[0062] The obtained dataset is used for model training, with 80% of the data used as the training set to train the model and the remaining 20% used as the validation set to obtain evaluation results.
[0063] An electronic device, comprising:
[0064] One or more processors;
[0065] Storage device for storing one or more programs.
[0066] When the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method.
[0067] A computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method.
[0068] The technical advantages of this invention lie in its use of sequence feature values to train a classification model. This method fully utilizes the information contained in first-generation sequencing results in sequence feature selection, ensuring high accuracy and stability of the classification results. This invention exhibits high accuracy in sample classification. Compared to traditional manual identification methods, the entire classification process does not rely on networks or manual operation, enabling rapid and accurate identification of heterozygous frameshift mutation events. This significantly reduces manual time costs, improves identification efficiency, and lowers data analysis costs.
[0069] The present invention will now be described in detail with reference to the embodiments shown in the accompanying drawings. The drawings are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort. Attached Figure Description
[0070] Figure 1 The diagram shows a flowchart of the method for determining heterozygous frameshift mutations in gene editing target site sequencing results based on machine learning, as described in this invention.
[0071] Figure 2 The figure shows the sequencing results and feature extraction regions of non-heterozygous frameshift mutation samples in the gene editing target site sequencing results of the present invention based on machine learning.
[0072] Figure 3 The figure shows the sequencing results and feature extraction region of the heterozygous frameshift mutation sample in the gene editing target site sequencing results of the present invention based on machine learning.
[0073] Figure 4 The diagram shows the flowchart of the classification model construction method for judging heterozygous frameshift mutations in gene editing target site sequencing results based on machine learning, according to the present invention.
[0074] Figure 5 The diagram shows a flowchart of the classification process for samples in the machine learning-based gene editing target site sequencing results of this invention. Detailed Implementation
[0075] The following will provide a clear and complete description of the concept, specific operation, and technical effects of the present invention in conjunction with the embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention.
[0076] Example 1
[0077] See Figure 1-5 This embodiment proposes a machine learning-based method for determining heterozygous frameshift mutations in gene editing target site sequencing results, including the following steps:
[0078] PCR primers were designed in regions at least 200 bp upstream and downstream of the gene editing target site, and sequence amplification and first-generation sequencing were performed.
[0079] Open the sequencing results file using AB1 format file reading software (such as SnapGene), locate the target site sequence, and manually identify the single peak on one side of the target site to pinpoint its location. Then, classify the located target sites: if a double peak appears in the middle of the target site and continues to the other side, it is considered a heterozygous frameshift mutation event; otherwise, it is considered a non-heterozygous frameshift mutation event. Next, assign binary labels to the sequencing results, such as labeling heterozygous frameshift mutations as 1 and non-heterozygous frameshift mutations as 0.
[0080] The sangerseqR package in R was used to read the ab1 file from the first-generation sequencing. The makeBaseCalls method was used to filter background interference below 0.2 times the main peak (ratio = 0.20), extracting the main peak and secondary peak sequences from the sequencing results. The highest peak in the sequencing results was taken as the main peak, and all sequences below 0.2 times the main peak were filtered out as background interference. The highest remaining sequence was then identified as the secondary peak. Sequence alignment of the main peak and secondary peak sequences was then performed to obtain the identical sequences between the two sequences. The sequence consisting of the most frequent nucleotides or amino acids at each corresponding position in both sequences was considered the identical sequence.
[0081] Then, the sequence alignment tool blast+ software is used to find target site regions that are homologous to the target site sequence in the homologous sequence, and the start and end positions of the target site regions in the homologous sequence are extracted.
[0082] Next, sequence feature values of the sequences surrounding the target site region are extracted to form sample features. The sequence feature values include the following extracted features: (1) the length of homology between the consistent sequence and the target site sequence. (2) the number of degenerate bases within a preset range upstream of the start position of the target site region in the consistent sequence. (3) the number of degenerate bases within a preset range downstream of the end position of the target site region in the consistent sequence. (4) the difference between the number of degenerate bases within a preset range downstream of the end position of the target site region in the consistent sequence and the number of degenerate bases within the same preset range upstream of the start position. (5) the ratio of the number of degenerate bases within a preset range downstream of the end position of the target site region in the consistent sequence to the number of degenerate bases within the same preset range upstream of the start position. (6) the consistency between the sequence within a preset range upstream of the start position of the target site region in the secondary peak sequence and the sequence within the same preset range upstream of the start position of the target site region in the primary peak sequence. (7) the consistency between the sequence within a preset range downstream of the end position of the target site region in the secondary peak sequence and the sequence within the same preset range downstream of the end position of the target site region in the primary peak sequence. The preset range can be set to 30-80bp, but it can be set to 50bp during actual operation.
[0083] After extraction, the sample features are standardized using the mean-variance normalization method to obtain the dataset for model training. Data standardization is performed using mean-variance normalization, with the following formula:
[0084]
[0085] Where x scale Here, x represents the standardized eigenvalues. mean s is the characteristic mean, and s is the characteristic variance.
[0086] A model is built based on a supervised classification algorithm in machine learning, and the obtained dataset is used for model training. 80% of the data in the dataset is used for training the model, and the remaining 20% is used for validation and evaluation. If the evaluation results do not meet the preset requirements, training is repeated until the requirements are met, thus obtaining a classification model that can be used for judgment.
[0087] Then, the sequence feature values of the sequencing results of the sample to be predicted are extracted and input into the constructed classification model to determine whether the sample to be predicted has undergone frameshift mutation, and the judgment result is output.
[0088] In determining heterozygous frameshift mutations in gene editing target site sequencing results, this method obtains the dataset used for model training through manual identification of a small number of samples, and then uses it to determine heterozygous frameshift mutations in gene editing target site sequencing results of a large number of samples.
[0089] Example 2
[0090] This embodiment proposes a machine learning-based device for determining heterozygous frameshift mutations in gene editing target site sequencing results, including:
[0091] The sample feature extraction module is used to extract the main peak sequence, secondary peak sequence, and consensus sequence from sequencing results that have been labeled with heterozygous frameshift mutation tags, and to extract sequence feature values of sequences surrounding the target site region after determining the target site region, so as to form sample features; wherein the sample features include:
[0092] (1) The length of homology between the homologous sequence and the target site sequence. (2) The number of degenerate bases within a preset range upstream of the start position of the target site region in the homologous sequence. (3) The number of degenerate bases within a preset range downstream of the end position of the target site region in the homologous sequence. (4) The difference between the number of degenerate bases within a preset range downstream of the end position of the target site region in the homologous sequence and the number of degenerate bases within the same preset range upstream of the start position. (5) The ratio of the number of degenerate bases within a preset range downstream of the end position of the target site region in the homologous sequence to the number of degenerate bases within the same preset range upstream of the start position. (6) The homology between the preset range upstream of the start position of the target site region in the secondary peak sequence and the same preset range upstream of the start position of the target site region in the primary peak sequence. (7) The homology between the preset range downstream of the end position of the target site region in the secondary peak sequence and the same preset range downstream of the end position of the target site region in the primary peak sequence. The preset range can be set to 30-80 bp, and can be set to 50 bp in specific operation.
[0093] The sequencing results labeled with heterozygous frameshift mutations in the sample feature extraction module involve first-generation sequencing of the gene editing target site region, followed by determination of whether a heterozygous frameshift mutation has occurred and subsequent labeling based on the sequencing results. Before first-generation sequencing of the gene editing target site region, PCR primers are designed for regions at least 200 bp upstream and downstream of the gene editing target site, and sequence amplification is performed. Determining whether a heterozygous frameshift mutation has occurred based on the sequencing results file involves locating the target site by finding a single peak on one side of the target site. The determination is based on the located target site: if a target site has a double peak in the middle that extends to the other side, it is identified as a heterozygous frameshift mutation; otherwise, it is identified as a non-heterozygous frameshift mutation. During labeling, heterozygous frameshift mutations and non-heterozygous frameshift mutations are labeled as one category in a binary classification tag.
[0094] The main peak sequence, secondary peak sequence, and consistency sequence in the sample feature extraction module are determined by identifying the highest peak in the sequencing results as the main peak, filtering out all sequences below 0.2 times the size of the main peak as background interference, and then identifying the highest remaining sequence as the secondary peak. Sequence alignment between the main and secondary peak sequences is then used to obtain the consistency sequence of the two sequences. Determining the target site region involves locating regions homologous to the target site sequence within the consistency sequence and identifying the start and end positions of the target site region within the consistency sequence.
[0095] The model building module is used to construct a mapping from sample features to sample labels, and to standardize the sample features based on supervised classification algorithms in machine learning, thereby establishing a dataset for training. Then, the model is trained and the accuracy of the trained model is evaluated. The training is repeated until a model that meets the preset accuracy requirements is obtained.
[0096] Data standardization is achieved through mean-variance normalization, using the following formula:
[0097]
[0098] Where x scale Here, x represents the standardized eigenvalues. mean s is the characteristic mean, and s is the characteristic variance.
[0099] Training the model and evaluating its accuracy involves using the obtained dataset for model training, where 80% of the data in the dataset is used as the training set to train the model, and the remaining 20% of the data is used as the validation set to obtain the evaluation results.
[0100] The heterozygous frameshift mutation determination module is used to extract the sequence feature values of the sequencing results of the sample to be predicted, and input them into the model obtained by the model building module, so as to obtain the determination result of whether the sample to be predicted has undergone frameshift mutation.
[0101] According to embodiments of the present invention, the present invention also provides an electronic device and a computer-readable medium.
[0102] Electronic devices include:
[0103] One or more processors;
[0104] Storage device for storing one or more programs.
[0105] When the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method.
[0106] In practical use, users can interact with servers, which are also electronic devices, via a network to receive or send messages. Terminal devices are generally various electronic devices equipped with a display and used through a human-computer interface, including but not limited to smartphones, tablets, laptops, and desktop computers. Various specific application software can be installed on these terminal devices as needed, including but not limited to web browsers, instant messaging software, social media platforms, and shopping apps.
[0107] A server is a network service provider that offers various services, such as a backend server that provides corresponding judgment services for sequencing results of samples to be predicted transmitted from terminal devices. This allows for the training of a historical model using the received training dataset, and the determination of heterozygous frameshift mutations based on the trained model, ultimately returning the final judgment result to the terminal device.
[0108] The hybrid code shift mutation determination method provided in this embodiment is generally executed by the server. In practical applications, under necessary conditions, the terminal device can also directly perform the hybrid code shift mutation determination. Correspondingly, the hybrid code shift mutation determination device can be set on the server. Similarly, under necessary conditions, the hybrid code shift mutation determination device can also be set in the terminal device.
[0109] Similarly, the computer-readable medium of the present invention stores a computer program that, when executed by a processor, implements the heterogeneous frameshift mutation determination method of the embodiments of the present invention. Although the description of the invention has been quite detailed, it is not intended to limit it to any of these details or embodiments, but should be considered as providing a broad possible interpretation of the claims by referring to the appended claims in consideration of the prior art, thereby effectively covering the intended scope of the invention.
[0110] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. Any invention that achieves the technical effects of the present invention by the same means should fall within the protection scope of the present invention.
Claims
1. A method for determining heterozygous frameshift mutations in gene editing target site sequencing results based on machine learning, characterized in that, Includes the following steps: Step 1: Extract the main peak sequence, secondary peak sequence and consensus sequence from the sequencing results that have been labeled with heterozygous frameshift mutation tags, and extract the sequence feature values of the sequences around the target site region after determining the target site region to form sample features; in Sample features include: Feature 1) The length of homology between the consistent sequence and the target site sequence; Feature 2) The number of degenerate bases upstream of the start position and downstream of the end position of the target site region in the consistent sequence within a preset range; Feature 3) The difference and ratio of the number of degenerate bases downstream of the termination position and upstream of the start position in the target site region within the same preset range in the consistent sequence; Feature 4) Consistency between the secondary peak sequence and the primary peak sequence within the same preset range upstream of the start position and downstream of the end position in the target site region; Step 2: Construct a mapping from sample features to sample labels, and standardize the sample features based on a supervised classification algorithm in machine learning to establish a dataset for training. Then train the model and evaluate the accuracy of the trained model. Repeat the training until a model that meets the preset accuracy requirements is obtained. Step 3: Extract the sequence feature values of the sequencing results of the sample to be predicted and input them into the model obtained in Step 2 to obtain the judgment result of whether the sample to be predicted has undergone heterozygous frameshift mutation.
2. The method according to claim 1, characterized in that, In step 1, the sequencing results with heterozygous frameshift mutation tags were obtained through the following steps: First-generation sequencing was performed on the gene editing target site region, and then the sequencing results were used to determine whether a heterozygous frameshift mutation occurred and to label it.
3. The method according to claim 2, characterized in that, Before performing first-generation sequencing on the gene editing target site region, the process also includes designing PCR primers in regions at least 200 bp upstream and downstream of the gene editing target site and performing sequence amplification.
4. The method according to claim 2, characterized in that, Determining whether a heterozygous frameshift mutation has occurred based on sequencing results is done through the following steps: Based on the target site sequence in the sequencing results file, the target site is located by finding a single peak on one side. The target site is then judged according to its location: if there are two peaks in the middle of the target site that continue to the other side, it is judged as a heterozygous frameshift mutation; otherwise, it is judged as a non-heterozygous frameshift mutation.
5. The method according to claim 2, characterized in that, When labeling, heterozygous frameshift mutations and non-heterozygous frameshift mutations are labeled as one of the two categories in the binary classification label.
6. The method according to claim 1, characterized in that, In step 1, the main peak sequence, secondary peak sequence, and consensus sequence are extracted through the following steps: The highest peak in the sequencing results is taken as the main peak. Then, all sequences lower than 0.2 times the main peak are filtered out as background interference. The highest peak among the remaining sequences is taken as the secondary peak. The sequence identity of the two sequences is obtained by comparing the main peak sequence and the secondary peak sequence.
7. The method according to claim 1, characterized in that, In step 1, determining the target site region involves finding regions in the consistent sequence that are homologous to the target site sequence, and determining the start and end positions of the target site region in the consistent sequence.
8. The method according to claim 1, characterized in that, In step 1, among the sample features, feature 2) includes: ① The number of degenerate bases within 30-80 bp upstream of the start position of the target site region in the homogeneous sequence; ② The number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence.
9. The method according to claim 1, characterized in that, In step 1, among the sample features, feature 3) includes: ① The difference between the number of degenerate bases within 30-80 bp downstream of the termination position and the number of degenerate bases within 30-80 bp upstream of the start position in the target site region of the homogeneous sequence; ② The ratio of the number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence to the number of degenerate bases within 30-80 bp upstream of the start position.
10. The method according to claim 1, characterized in that, In step 1, among the sample features, feature 4) includes: ① The consistency between the 30-80bp upstream of the target site region in the secondary peak sequence and the 30-80bp upstream of the target site region in the main peak sequence; ② The consistency between the 30-80bp downstream of the target site region termination position in the secondary peak sequence and the 30-80bp downstream of the target site region termination position in the main peak sequence.
11. The method according to claim 1, characterized in that, In step 2, data standardization is performed through mean-variance normalization, using the following formula: Where x scale Here, x represents the standardized eigenvalues. mean s is the characteristic mean, and s is the characteristic variance.
12. The method according to claim 1, characterized in that, Step 2, training the model and evaluating the accuracy of the trained model, includes: The obtained dataset is used for model training, with 80% of the data used as the training set to train the model and the remaining 20% used as the validation set to obtain evaluation results.
13. A device for determining heterozygous frameshift mutations in gene editing target site sequencing results based on machine learning, characterized in that, Includes the following modules: The sample feature extraction module is used to extract the main peak sequence, secondary peak sequence and consensus sequence from the sequencing results that have been labeled with heterozygous frameshift mutation tags, and after determining the target site region, extract the sequence feature values of the sequences surrounding the target site region to form sample features; in Sample features include: Feature 1) The length of homology between the consistent sequence and the target site sequence; Feature 2) The number of degenerate bases upstream of the start position and downstream of the end position of the target site region in the consistent sequence within a preset range; Feature 3) The difference and ratio of the number of degenerate bases downstream of the termination position and upstream of the start position in the target site region within the same preset range in the consistent sequence; Feature 4) Consistency between the secondary peak sequence and the primary peak sequence within the same preset range upstream of the start position and downstream of the end position in the target site region; The model building module is used to construct a mapping from sample features to sample labels, and to standardize the sample features based on supervised classification algorithms in machine learning, thereby establishing a dataset for training. Then, the model is trained and the accuracy of the trained model is evaluated. The training is repeated until a model that meets the preset accuracy requirements is obtained. The heterozygous frameshift mutation determination module is used to extract the sequence feature values of the sequencing results of the sample to be predicted, and input them into the model obtained by the model construction module, so as to obtain the determination result of whether the sample to be predicted has undergone heterozygous frameshift mutation.
14. The apparatus according to claim 13, characterized in that, In the sample feature extraction module, the sequencing results that have been labeled with heterozygous frameshift mutations are obtained by performing first-generation sequencing on the gene editing target site region, and then determining whether a heterozygous frameshift mutation has occurred based on the sequencing results and labeling it. Before performing first-generation sequencing on the gene editing target site region, PCR primers are designed for regions at least 200 bp upstream and downstream of the gene editing target site, and the sequences are amplified. To determine whether a heterozygous frameshift mutation has occurred based on the sequencing results, the target site sequence is obtained, and a single peak on one side of the target site is searched to locate the target site. The target site is then judged based on the located target site: if there are two peaks in the middle of the target site that continue to the other side, it is judged as a heterozygous frameshift mutation; otherwise, it is judged as a non-heterozygous frameshift mutation. When labeling, heterozygous frameshift mutations and non-heterozygous frameshift mutations are labeled as one of the two categories in a binary classification label.
15. The apparatus according to claim 13, characterized in that, In the sample feature extraction module, the main peak sequence, secondary peak sequence, and consistency sequence are extracted through the following steps: The highest peak in the sequencing results is taken as the main peak. Then, all sequences lower than 0.2 times the main peak are filtered out as background interference. The highest peak among the remaining sequences is taken as the secondary peak. The sequence identity of the two sequences is obtained by comparing the main peak sequence and the secondary peak sequence.
16. The apparatus according to claim 13, characterized in that, In the sample feature extraction module, determining the target site region involves finding regions in the consistent sequence that are homologous to the target site sequence, and determining the start and end positions of the target site region in the consistent sequence.
17. The apparatus according to claim 13, characterized in that, In the sample feature extraction module, feature 2) includes: ① The number of degenerate bases within 30-80 bp upstream of the start position of the target site region in the homogeneous sequence; ② The number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence; Feature 3) includes: ① The difference between the number of degenerate bases within 30-80 bp downstream of the termination position and the number of degenerate bases within 30-80 bp upstream of the start position in the target site region of the homogeneous sequence; ② The ratio of the number of degenerate bases within 30-80 bp downstream of the termination position of the target site region in the homogeneous sequence to the number of degenerate bases within 30-80 bp upstream of the start position; Feature 4) includes: ① The consistency between the 30-80bp upstream of the target site region in the secondary peak sequence and the 30-80bp upstream of the target site region in the main peak sequence; ② The consistency between the 30-80bp downstream of the target site region termination position in the secondary peak sequence and the 30-80bp downstream of the target site region termination position in the main peak sequence.
18. The apparatus according to claim 13, characterized in that, In the model building module, data standardization is performed through mean-variance normalization, using the following formula: Where x scale Here, x represents the standardized eigenvalues. mean s is the characteristic mean, and s is the characteristic variance.
19. The apparatus according to claim 13, characterized in that, The model building module includes training the model and evaluating the accuracy of the trained model, which includes: The obtained dataset is used for model training, with 80% of the data used as the training set to train the model and the remaining 20% used as the validation set to obtain evaluation results.
20. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-12.
21. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-12.
Citation Information
Patent Citations
Genetic mutation detection using deep learning
CN113574605A
False positive nucleotide variation site filtering method and computing device
CN114613430A
