P450 gene automatic extraction and identification method based on omics data

Abstract

The invention provides an omics data-based P450 gene automatic extraction and identification method, which comprises the following steps: carrying out homogenization treatment on different omics amino acid sequence data to obtain a plurality of gene homologous groups; acquiring P450 and non-P450 amino acid sequence data, and summarizing the data into a data set; splitting the data set into different groups based on gene evolution clustering to obtain grouping information, taking the grouping information as auxiliary features of classification and identification, and adding the grouping information into the data set; dividing the data set into training data and test data, training the classification model through the training data, and performing performance evaluation on the trained classification model through the test data; and inputting the plurality of genes into a trained classification model in a homologous grouping manner, and outputting a P450 gene identification result. The method can assist a machine learning algorithm in deeply and accurately analyzing the gene sequence, so that the accuracy and efficiency of gene identification are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of gene identification, and in particular to a method for automatic extraction and identification of P450 genes based on omics data. Background Art

[0002] The identification of P450 genes has far-reaching significance in the biomedical field. P450 genes are not only involved in the detoxification of drugs in the human liver and bile acid metabolism, but also in the biosynthesis of anti-tumor drugs such as paclitaxel, anti-malarial drugs such as artemisinin, antibacterial drugs such as penicillin, and lipid-lowering statins. Identifying P450 genes in transcriptomes of different states of the human body can help us better understand the metabolic differences of individuals and predict their responses to the environment, thereby guiding clinical medication and improving the safety and effectiveness of drug treatment; identifying P450 genes in the genomes of different species can help us better understand the distribution characteristics of P450 biological groups, explore the structural and functional diversity of P450, reveal its pathways and mechanisms of participation in metabolism, and provide new ideas and methods for the screening of innovative biological drugs, disease treatment and prevention.

[0003] In animal genomes, the number of P450 gene family members is huge and widely distributed. For example, there are hundreds of different P450 sequences in mammals and thousands of P450 in wheat. At the same time, the similarity of P450 sequences between different species is low, usually less than 40%. When conducting cross-species comparisons or whole genome identification, it is difficult to find all relevant P450 sequences through simple sequence alignment or search methods. Traditional gene identification methods mainly rely on manual operations, which is particularly time-consuming and labor-intensive when processing massive amounts of data. More importantly, human factors may cause errors and inconsistencies in the data processing process, which seriously affects the accuracy and reliability of subsequent analysis results.

[0004] To overcome the above limitations, it is particularly important to develop an efficient and accurate automated gene identification process. The automated gene identification process can fully utilize the advantages of computer technology and algorithms to quickly process massive amounts of omics data, accurately identify, extract and analyze relevant genes, improve the efficiency and accuracy of data processing, and reduce errors and inconsistencies caused by human factors. Summary of the invention

[0005] The purpose of the present invention is to provide a method for automatic extraction and identification of P450 genes based on omics data, comprehensively consider the differences in various omics data types, fully learn the characteristics of P450 genes of different species through amino acid word frequency, and assist P450 rapid identification based on gene evolution clustering information. Finally, P450 genes are accurately identified by integrating multi-model prediction results to overcome the limitations of traditional P450 gene extraction and identification.

[0006] To achieve the above object, the present invention provides a method for automatic extraction and identification of P450 genes based on omics data, the method comprising: The amino acid sequence data of different omics were homogenized to obtain multiple gene homology groups; Obtain P450 and non-P450 amino acid sequence data and summarize them into a data set; Splitting the data set into different groups based on gene evolution clustering, obtaining group information, using the group information as an auxiliary feature for classification and identification, and adding it to the data set; Divide the data set into training data and test data, train the classification model with the training data, and evaluate the performance of the trained classification model with the test data; Multiple gene homology groups were input into the trained classification model, and the P450 gene identification results were output.

[0007] Furthermore, the homogenization of different omics amino acid sequence data to obtain multiple gene homology groups specifically includes: Determine the properties of amino acid sequence data from different omics to distinguish whether the amino acid sequence data is protein data or nucleotide data; For nucleotide data, determine whether it is annotated nucleotide data, translate the annotated nucleotide data into protein data and perform normalization; For unannotated nucleotide data, it is further determined whether it is derived from eukaryotes or prokaryotes; Similar sequence fragments were extracted from unannotated nucleotide data from eukaryotes and prokaryotes, and gene structures were predicted for the similar sequence fragments to obtain potential P450 protein sequences.

[0008] Furthermore, uncommon amino acids in potential P450 protein sequences and non-underline characters in fasta tags are identified, replaced and formatted, wherein the uncommon amino acids are amino acids that do not belong to the standard 20 amino acids.

[0009] Furthermore, the acquisition of P450 and non-P450 amino acid sequence data and aggregating them into a data set specifically includes: The acquired P450 and non-P450 amino acid sequence data were converted into text content, and the CountVectorizer tool for text feature extraction in Scikit-learn was used to extract phrases of single amino acid, two consecutive amino acids, and three consecutive amino acids according to ngram_range=(1, 3); Using regular expressions to identify each amino acid phrase, so that only amino acid phrases will be used as building blocks of n-grams; Generate a phrase-document matrix to record the occurrence of each n-gram in the text, and convert the amino acid phrases into word frequency representation using bag-of-words encoding.

[0010] Furthermore, the P450 and non-P450 amino acid sequence data were initially cleaned and formatted before being converted into text content.

[0011] Furthermore, the data set is split into different groups based on gene evolution clustering to obtain grouping information, specifically: the data set is processed by orthofinder software to obtain clustering information based on sequence evolution, and the data belonging to the same cluster in the clustering information are designated as the same auxiliary group to obtain grouping information.

[0012] Furthermore, the grouping information is represented in a hot single encoding manner.

[0013] Furthermore, the grouping information is added to the data set, specifically, the hot singleton encoding of the grouping information and the bag-of-words model encoding of the amino acid word frequency in the data set are concatenated into a feature vector.

[0014] Furthermore, the performance of the trained classification model is evaluated through test data, and the performance includes precision, recall, F1 score and accuracy.

[0015] Furthermore, multiple gene homology groups are input into a trained classification model to output the P450 gene identification results, specifically including: using a linear regression model, a Bayesian model, an extreme gradient boosting model, a support vector machine model and a random forest model of a natural language processing method based on amino acid sequence features as base learners, training multiple base learners respectively, and constructing an extreme gradient boosting model as a meta-learner at the same time, inputting multiple gene homology groups into multiple base learners respectively, using the output of the base learner as the input of the meta-learner, making a final prediction, and outputting the P450 gene identification results.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The method provided by the present invention combines bioinformatics data processing with machine learning technology based on amino acid word frequency combined with gene evolution clustering grouping information to construct a word bag, which is used for P450 gene identification. The method can assist machine learning algorithms in performing in-depth and accurate analysis of gene sequences, thereby significantly improving the accuracy and efficiency of gene identification.

[0017] 2. Compared with traditional gene extraction methods, the method provided by the present invention has a simpler and clearer process, which can reduce the time and labor costs of gene identification, making the method more economical and practical, and easy to promote on a large scale.

[0018] 3. The method is based on automated bioinformatics processing and rapid machine learning to identify P450 sequences, and can efficiently process large-scale biological data sets. The method is not only suitable for basic scientific research, but also has good clinical application prospects. At the same time, the method can also be applied to the identification of other genes besides P450 genes, providing strong support for disease diagnosis, drug development, genetic counseling, precision medicine and personalized treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 It is a schematic diagram of the overall process of a method for automatic extraction and identification of P450 genes based on omics data provided by an embodiment of the present invention.

[0021] Figure 2 4 is a schematic diagram of confusion matrix evaluation provided by an embodiment of the present invention.

[0022] Figure 3 Schematic diagram of the ROC curve provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0023] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The enumerated embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0024] Reference Figure 1 This embodiment provides a method for automatic extraction and identification of P450 genes based on omics data, the method comprising: S101. Perform homogenization processing on amino acid sequence data of different omics to obtain multiple gene homology groups.

[0025] S102. Obtain P450 and non-P450 amino acid sequence data, and summarize them into a data set.

[0026] Exemplarily, in this step, P450 amino acid sequence data are collected by manual screening, and an equal amount of non-P450 amino acid sequence data are collected from the uniport public database.

[0027] S103, splitting the data set into different groups based on gene evolution clustering, obtaining group information, using the group information as an auxiliary feature for classification and identification, and adding it to the data set.

[0028] S104, dividing the data set into training data and test data, training the classification model through the training data, and evaluating the performance of the trained classification model through the test data.

[0029] S105, inputting multiple gene homology groups into the trained classification model, and outputting the P450 gene identification results.

[0030] As a possible implementation, the homogenization of different omics amino acid sequence data to obtain multiple gene homology groups specifically includes: S201. Determine the attributes of different omics amino acid sequence data to distinguish whether the amino acid sequence data is protein data or nucleotide data.

[0031] S202: For the nucleotide data, determine whether it is annotated nucleotide data, translate the annotated nucleotide data into protein data, and perform normalization.

[0032] S203. For the unannotated nucleotide data, further determine whether it is derived from eukaryotes or prokaryotes.

[0033] S204. Extract similar sequence fragments from unannotated nucleotide data from eukaryotes and prokaryotes, predict gene structures of the similar sequence fragments, and obtain potential P450 protein sequences.

[0034] In this embodiment, the P450 dataset is compared based on sequence similarity and homology to extract similar genome sequence fragments. The P450 dataset used for the similar fragment comparison is the P450 sequences of all biological groups that are artificially screened and summarized.

[0035] In this embodiment, Augustus software can be used to predict the gene structure of similar sequence fragments from unannotated nucleotide data of eukaryotes; Prokka software can be used to predict the gene structure of similar sequence fragments from unannotated nucleotide data of prokaryotes.

[0036] In this embodiment, after obtaining the potential P450 protein sequence, the uncommon amino acids in the potential P450 protein sequence and the non-underline characters in the fasta tag are also identified, replaced and formatted to ensure uniformity and accuracy in subsequent data analysis. The uncommon amino acids are amino acids that do not belong to the standard 20 amino acids. Exemplary, the replacement can be to replace the uncommon amino acids with X and replace the non-underline characters with underscores.

[0037] This implementation first extracts similar sequence fragments from species genome data and then performs gene structure prediction, which can reduce the time consumption caused by prediction.

[0038] As a possible implementation, the obtaining of P450 and non-P450 amino acid sequence data and aggregating them into a data set specifically includes the following operations: S301, converting the acquired P450 amino acid sequence data and non-P450 amino acid sequence data into text content, and using CountVectorizer, a tool for text feature extraction in Scikit-learn, to extract phrases of a single amino acid, two consecutive amino acids, and three consecutive amino acids according to ngram_range=(1,3).

[0039] S302: Use regular expressions to identify each amino acid phrase, so that only the amino acid phrase will be used as a construction unit of n-gram.

[0040] S303, generating a phrase-document matrix, recording the occurrence of each n-gram in the text, and converting the amino acid phrase into a word frequency representation using a bag-of-words encoding method.

[0041] In this embodiment, before converting the P450 and non-P450 amino acid sequence data into text content, they are preliminarily cleaned and formatted to prepare for subsequent model training. Exemplarily, the process of the preliminarily cleaning includes reading the data file and then performing data cleaning on it, such as deleting rows with empty content or labels.

[0042] In this embodiment, by manually screening protein sequences of P450 databases from multiple categories and species, the extensiveness of P450 training data can be ensured, and the sequence characteristics of different species can be fully displayed. At the same time, sequence feature extraction is performed based on amino acid word frequency, so that the model can fully learn the characteristics of P450 of different species.

[0043] As another possible implementation, the data set is split into different groups based on gene evolution clustering to obtain grouping information, specifically: the data set is processed by orthofinder software to obtain clustering information based on sequence evolution, and the data belonging to the same cluster in the clustering information are designated as the same auxiliary group to obtain grouping information.

[0044] In this embodiment, auxiliary grouping is achieved by clustering based on gene evolution. This grouping can better display the homology information of genes compared with sequence similarity grouping. The above combination of multi-category data training, word frequency feature extraction, and evolutionary grouping auxiliary can further improve the accuracy of the model for P450 identification.

[0045] In this implementation, the group information is represented in a one-hot encoding manner.

[0046] The above implementation is described below by using an example as shown in Table 1.

[0047] Table 1 Dataset examples

[0048] The amino acid word frequency is represented by the bag-of-words model encoding: P450=[ [1,1,1,1,1,0,0,0]#Sequence 1 [……]#Other P450 sequences ] UnP450=[ [1,0,0,0,1,1,1,1]#Sequence 2 [……] #Other non-P450 sequences ] The group information is represented by hot encoding: P450=[ [1,0]#Sequence 1#Auxiliary grouping is 1 [……]#Other P450 sequences ] UnP450=[ [0,1]#Sequence 2#Auxiliary grouping is 2 [……] #Other non-P450 sequences ] Use FeatureUnion to merge the results by column to form a matrix for subsequent use by the model. The matrix can be expressed as: P450=[ [1,1,1,1,1,0,0,0,1,0] #Sequence 1 [……]#Other P450 sequences ] UnP450=[ [1,0,0,0,1,1,1,1,0,1] #Sequence 2 [……] #Other non-P450 sequences ] In another optional embodiment, the grouping information is added to the data set, specifically, the hot unique encoding of the grouping information and the bag-of-words encoding of the amino acid word frequency in the data set are concatenated into a feature vector, based on this method, all the bag-of-words encodings in the data set are concatenated into a feature vector, and on this basis, the data set is divided into training data and test data. Exemplarily, the ratio of training data to test data can be 8:2.

[0049] In this embodiment, the performance of the trained classification model is evaluated by test data, and the performance includes precision, recall, F1 score and accuracy. Specifically, in order to confirm the accuracy of the model, the classification model is used to predict the test data, generate prediction labels, calculate and output the classification results, including precision, recall, F1 score and accuracy, calculate and output the confusion matrix, and display the accuracy and error distribution of the model prediction results.

[0050] As another possible implementation method, multiple gene homology groups are input into a trained classification model to output the P450 gene identification result, specifically including: using a linear regression model, a Bayesian model, an extreme gradient boosting model, a support vector machine model and a random forest model of a natural language processing method based on amino acid sequence features as base learners, training multiple base learners respectively, and constructing an extreme gradient boosting model as a meta-learner at the same time, inputting multiple gene homology groups into multiple base learners respectively, using the output of the base learner as the input of the meta-learner, making a final prediction, and outputting the P450 gene identification result.

[0051] In this implementation, the meta-learner uses grid search to try all possible parameter combinations to find the optimal solution for training and optimization of the final integrated model.

[0052] In an experimental example of the present invention, the data set used for testing includes manually collected multi-class P450 data and non-P450 data randomly extracted from the uniport database, totaling 114,992 protein sequences. 2 / 10 of them were randomly selected to test the reliability and accuracy of the results. For the integrated model composed of the base learner and the meta learner, the classification accuracy evaluation results are shown in Table 2.

[0053] Table 2 Classification accuracy evaluation results of the integrated model

[0054] The accuracy of the model refers to the proportion of samples that the model predicts correctly. The 0.98 in Table 2 means that approximately 98% of the samples in the corresponding total number of samples are correctly predicted by the model.

[0055] like Figure 2 As shown in the figure, from the numbers on the diagonal, the integrated model achieves accurate prediction for about 98% of the sequences in both categories.

[0056] AUC (Area Under the Curve) is the area under the ROC curve (Receiver Operating Characteristic Curve). The ROC curve represents the performance of a classifier by plotting the True Positive Rate (TPR, also known as the recall rate or sensitivity) against the False Positive Rate (FPR) at different thresholds. Figure 3 As shown, the AUC values ​​of both UnP450 and P450 labels are 0.98, which indicates that the model provided in this embodiment can distinguish all positive and negative classes.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for automatic extraction and identification of P450 genes based on omics data, characterized in that: The method comprises: The amino acid sequence data of different omics were homogenized to obtain multiple gene homology groups; Obtain P450 and non-P450 amino acid sequence data and summarize them into a data set; Splitting the data set into different groups based on gene evolution clustering, obtaining group information, using the group information as an auxiliary feature for classification and identification, and adding it to the data set; Divide the data set into training data and test data, train the classification model with the training data, and evaluate the performance of the trained classification model with the test data; Multiple gene homology groups were input into the trained classification model, and the P450 gene identification results were output.

2. The method for automatic extraction and identification of P450 genes based on omics data according to claim 1, characterized in that: The homogenization of different omics amino acid sequence data to obtain multiple gene homology groups specifically includes: Determine the properties of amino acid sequence data from different omics to distinguish whether the amino acid sequence data is protein data or nucleotide data; For nucleotide data, determine whether it is annotated nucleotide data, translate the annotated nucleotide data into protein data and perform normalization; For unannotated nucleotide data, it is further determined whether it is derived from eukaryotes or prokaryotes; Similar sequence fragments were extracted from unannotated nucleotide data from eukaryotes and prokaryotes, and gene structures were predicted for the similar sequence fragments to obtain potential P450 protein sequences.

3. The method for automatic extraction and identification of P450 genes based on omics data according to claim 2, characterized in that: Uncommon amino acids in potential P450 protein sequences and non-underline characters in fasta tags are identified, replaced and formatted. The uncommon amino acids are amino acids that do not belong to the standard 20 amino acids.

4. The method for automatic extraction and identification of P450 genes based on omics data according to claim 1, characterized in that: The obtaining of P450 and non-P450 amino acid sequence data and summarizing them into a data set specifically includes: The acquired P450 and non-P450 amino acid sequence data were converted into text content, and the CountVectorizer tool for text feature extraction in Scikit-learn was used to extract phrases of single amino acid, two consecutive amino acids, and three consecutive amino acids according to ngram_range=(1, 3); Using regular expressions to identify each amino acid phrase, so that only amino acid phrases will be used as building blocks of n-grams; Generate a phrase-document matrix to record the occurrence of each n-gram in the text, and convert the amino acid phrases into word frequency representation using bag-of-words encoding.

5. The method for automatic extraction and identification of P450 genes based on omics data according to claim 4, characterized in that: The P450 and non-P450 amino acid sequence data were initially cleaned and formatted before being converted into text content.

6. The method for automatic extraction and identification of P450 genes based on omics data according to claim 4, characterized in that: The method of dividing the data set into different groups based on gene evolution clustering to obtain grouping information is specifically as follows: processing the data set by orthofinder software to obtain clustering information based on sequence evolution, and designating data belonging to the same cluster in the clustering information as the same auxiliary group to obtain grouping information.

7. The method for automatic extraction and identification of P450 genes based on omics data according to claim 6, characterized in that: The grouping information is represented in a one-hot encoding manner.

8. The method for automatic extraction and identification of P450 genes based on omics data according to claim 7, characterized in that: The grouping information is added to the data set, specifically, the hot unique encoding of the grouping information and the bag-of-words model encoding of the amino acid word frequency in the data set are concatenated into a feature vector.

9. The method for automatic extraction and identification of P450 genes based on omics data according to claim 1, characterized in that: The performance of the trained classification model is evaluated by testing data, and the performance includes precision, recall, F1 score and accuracy.

10. The method for automatic extraction and identification of P450 genes based on omics data according to claim 1, characterized in that: Multiple gene homology groups are input into a trained classification model, and the P450 gene identification results are output, specifically including: using a linear regression model, a Bayesian model, an extreme gradient boosting model, a support vector machine model and a random forest model of a natural language processing method based on amino acid sequence features as base learners, training multiple base learners respectively, and constructing an extreme gradient boosting model as a meta-learner at the same time, inputting multiple gene homology groups into multiple base learners respectively, using the output of the base learner as the input of the meta-learner, making a final prediction, and outputting the P450 gene identification results.

Images