A multifunctional bioactive peptide prediction method and system

Abstract

A method and system for predicting multifunctional bioactive peptides, comprising the following steps: S1. Preprocessing data; converting amino acid sequence data into digital data; S2. Feature extraction of the digital data; first using a multi-scale TCN to obtain information about sequence patterns of different lengths in the sequence, then using a GRU to obtain the relationship between each sequence pattern, and finally obtaining a feature representation with global information; S3. Target classification; using a fully connected neural network with classification function to classify the multifunctional bioactive peptide data; S4. Model training; dividing the multifunctional bioactive peptide dataset into a training set and a validation set using a five-fold cross-validation method, and finally obtaining the optimal model; S5. Using the model for prediction. This application is based on multi-label learning and time-recurrent neural networks. We propose a multi-label predictor to identify the various functions of bioactive peptides, including anti-cancer, anti-diabetic, anti-hypertensive, anti-inflammatory, and anti-microbial.

Description

Technical Field

[0001] The present invention belongs to the field of bioinformatics computing, and in particular relates to a multifunctional bioactive peptide prediction method and system. Background Art

[0002] Bioactive peptides play a crucial role in regulating various bodily functions, such as lowering blood pressure, blood sugar levels, and cholesterol. Researchers have extensively explored bioactive peptides, and a growing body of experimental data is becoming available. Consequently, over the past few decades, a growing number of bioactive peptides with diverse functions have been identified. Due to their efficacy, short half-life, and low toxicity, peptide drug development has gained increasing market approval. To advance the development of peptide therapeutics, the identification of peptide function is crucial.

[0003] Given the challenges of bioactive peptide screening, machine learning can be a powerful tool for identifying potential functions of bioactive peptides. However, previous methods have primarily focused on predicting single peptide functions. Over the past few decades, numerous multi-label classification algorithms have been proposed across various fields, including text, images, music, and bioinformatics. In recent years, deep learning has shown great potential by combining representation learning and classification learning into its models. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a multifunctional bioactive peptide prediction method and system. The specific technical solutions are as follows:

[0005] A method for predicting multifunctional bioactive peptides comprises the following steps:

[0006] S1. Preprocess the data; convert the amino acid sequence data in the multifunctional bioactive peptide dataset into digital data;

[0007] S2. Feature extraction of digitized data: first, a multi-scale TCN is used to obtain information about sequence patterns of different lengths in the sequence, and then a GRU is used to obtain the relationship between each sequence pattern, ultimately obtaining a feature representation with global information.

[0008] S3, target classification: using a fully connected neural network (FCNN) with classification function to classify the multifunctional bioactive peptide data;

[0009] S4. Model training: The multifunctional bioactive peptide dataset is divided into a training set and a validation set according to a five-fold cross-validation method. The training set is used to construct and train the multifunctional bioactive peptide prediction model, and the validation set is used to adjust the model parameters to ultimately obtain the optimal model.

[0010] S5. Use the model to make predictions; input the sequence data into the prediction model of step S4 to make predictions and obtain the functions of the peptide sequence.

[0011] The data converted in step S1 includes dictionary encoding, sequence completion and word embedding;

[0012] Dictionary encoding: Each peptide sequence in the dataset is composed of twenty amino acids: A, C, D, E, F, G, H, I, K, L, M, N, P, Q, R, S, T, V, W, and Y. The amino acids in the peptide sequence are represented by numbers 1, 2, 3, ..., 20 respectively;

[0013] Sequence padding: Use the zero-padding method to fill each sequence with the number 0 to the set length of the sequence in the data set;

[0014] Word embedding: Convert words into vector representations that represent the connections between each element in the sequence.

[0015] The length of the sequence is set to 517.

[0016] In step S4, 80% of the multifunctional bioactive peptide dataset is divided into training set and validation set according to a five-fold cross-validation method, and the remaining 20% is used as a test set.

[0017] A system for realizing the above-mentioned multifunctional bioactive peptide prediction method comprises:

[0018] The data preprocessing module converts the amino acid sequence data in the multifunctional bioactive peptide dataset into digital data and performs sequence completion;

[0019] The feature extractor, which includes a temporal convolutional neural network and a gated recurrent unit (GRU) for feature extraction, first uses a multi-scale TCN to obtain information about sequence patterns of different lengths in the sequence, then uses a GRU to obtain the relationship between each sequence pattern, and finally obtains a feature representation with global information;

[0020] The target classifier is a fully connected neural network with classification function, which classifies the multifunctional bioactive peptide data;

[0021] The data set includes a training set and a validation set divided from the multifunctional bioactive peptide data set according to a five-fold cross-validation method. The training set is used to construct and train a multifunctional bioactive peptide prediction model, and the validation set is used to adjust the parameters of the model. The model predicts the functions of peptide sequences.

[0022] The advantages of the present invention are: Based on multi-label learning and time-recurrent neural networks, we propose a multi-label predictor to identify multiple functions of bioactive peptides, including anti-cancer, anti-diabetic, anti-hypertensive, anti-inflammatory and anti-microbial. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of the overall framework flow of the device of the present invention;

[0024] Figure 2 This is a schematic diagram of the prediction model for multifunctional bioactive peptides. DETAILED DESCRIPTION

[0025] like Figure 1-2 As shown, a method for predicting multifunctional bioactive peptides comprises the following steps:

[0026] S1. Preprocess the data; convert the amino acid sequence data in the multifunctional bioactive peptide dataset into digital data;

[0027] The conversion in step S1 includes dictionary encoding, sequence completion and word embedding;

[0028] Dictionary encoding: Since each peptide sequence in the dataset is composed of twenty amino acids, A, C, D, E, F, G, H, I, K, L, M, N, P, Q, R, S, T, V, W, and Y, this step represents the amino acids in the peptide sequence with numbers 1, 2, 3, ..., 20 respectively;

[0029] Sequence padding: TCN input requires a fixed-length sequence. However, the lengths of sequences in the dataset vary. Therefore, we use the zero-padding method to pad each sequence with the number 0 to the set length in the dataset. The set length for sequence padding is 517.

[0030] Word embedding: Since digital representation cannot reflect the positional relationship between each element in the sequence, word embedding converts words into vector representation, which can effectively represent the connection between each element in the sequence.

[0031] S2. Feature extraction of digitized data: first, a multi-scale TCN is used to obtain information about sequence patterns of different lengths in the sequence, and then a GRU is used to obtain the relationship between each sequence pattern, ultimately obtaining a feature representation with global information.

[0032] S3, target classification: using a fully connected neural network (FCNN) with classification function to classify the multifunctional bioactive peptide data;

[0033] S4. Model Training: The multifunctional bioactive peptide dataset was divided into training and validation sets using a five-fold cross-validation method. The training set was used to construct and train the multifunctional bioactive peptide prediction model, while the validation set was used to adjust the model parameters and ultimately obtain the optimal model. This experiment used the MLBP dataset, which contains 646 anticancer peptides, 514 antidiabetic peptides, 868 antihypertensive peptides, 1,678 anti-inflammatory peptides, and 2,409 antimicrobial peptides. 80% of these peptides were used as the training set, and 20% as the test set.

[0034] S5. Use the model to make predictions; input the sequence data into the prediction model of step S4 to make predictions and obtain the functions of the peptide sequence.

[0035] A system for realizing the above-mentioned multifunctional bioactive peptide prediction method comprises:

[0036] The data preprocessing module converts the amino acid sequence data in the multifunctional bioactive peptide dataset into digital data and performs sequence completion;

[0037] The feature extractor, which includes a temporal convolutional neural network and a gated recurrent unit (GRU) for feature extraction, first uses a multi-scale TCN to obtain information about sequence patterns of different lengths in the sequence, then uses a GRU to obtain the relationship between each sequence pattern, and finally obtains a feature representation with global information;

[0038] The target classifier is a fully connected neural network with classification function, which classifies the multifunctional bioactive peptide data;

[0039] The data set includes a training set and a validation set divided from the multifunctional bioactive peptide data set according to a five-fold cross-validation method. The training set is used to construct and train a multifunctional bioactive peptide prediction model, and the validation set is used to adjust the parameters of the model. The model is used to predict the functions of peptide sequences.

[0040] This application evaluates the above model using the following evaluation metrics: precision, recall, accuracy, absolute true, and absolute false. The calculation method is as follows:

[0041]

[0042] Where N is the total number of multifunctional bioactive peptides, M represents the total number of bioactive peptide types, ∪ represents the union in set theory, ∩ represents the intersection in set theory, |||| represents the operation of calculating the number of elements, and L i represents the true label of the i-th sample, represents the predicted label of the i-th sample, and:

[0043]

[0044] In order to verify the superiority of the multifunctional bioactive peptide prediction model (PrMLBP), we selected some excellent tools for comparison, including: the deep learning-based method MLBP, the second-order algorithm CLR, the high-order algorithm RAKEL, the low-rank learning algorithm RBRL and MLDF. Table 1 fully shows the performance comparison of the PrMLBP model with the other five models under five evaluation indicators (Precision, Coverage, Accuracy, Absolute true and Absolute false). It can be seen that the PrMLBP model achieved the best performance in the four indicators of Precision, Coverage, Accuracy and Absolute true, and also performed well in Absolute false. Accuracy and Absolute true are the two most important evaluation indicators for measuring multi-label learning models. The Accuracy and Absolute true performance of the PrMLBP model are improved compared with the other five models, reflecting the effectiveness of the PrMLBP model.

[0045]

[0046] Table 1

[0047] The present invention adopts a multi-label classification method for multifunctional bioactive peptides with five functions including anti-cancer, anti-diabetes, anti-hypertension, anti-inflammatory and antimicrobial, to construct an optimal model to predict and classify the multifunctional bioactive peptides, contributing to the rapid identification of the functions of bioactive peptides.

[0048] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting multifunctional bioactive peptides, characterized in that: The following steps are involved: S1. Preprocess the data; convert the amino acid sequence data in the multifunctional bioactive peptide dataset into digital data; S2. Feature extraction of digitized data: first, a multi-scale TCN is used to obtain information about sequence patterns of different lengths in the sequence, and then a GRU is used to obtain the relationship between each sequence pattern, ultimately obtaining a feature representation with global information. S3, target classification: using a fully connected neural network (FCNN) with classification function to classify the multifunctional bioactive peptide data; S4. Model training: The multifunctional bioactive peptide dataset is divided into a training set and a validation set according to a five-fold cross-validation method. The training set is used to construct and train the multifunctional bioactive peptide prediction model, and the validation set is used to adjust the model parameters to ultimately obtain the optimal model. S5. Use the model to make predictions; input the sequence data into the prediction model of step S4 to make predictions and obtain the functions of the peptide sequence.

2. A multifunctional bioactive peptide prediction method according to claim 1, characterized in that: The data converted in step S1 includes dictionary encoding, sequence completion and word embedding; Dictionary encoding: Each peptide sequence in the dataset is composed of twenty amino acids: A, C, D, E, F, G, H, I, K, L, M, N, P, Q, R, S, T, V, W, and Y. The amino acids in the peptide sequence are represented by numbers 1, 2, 3, ..., 20 respectively; Sequence padding: Use the zero-padding method to fill each sequence with the number 0 to the set length of the sequence in the data set; Word embedding: Convert words into vector representations that represent the connections between each element in the sequence.

3. A multifunctional bioactive peptide prediction method according to claim 2, characterized in that: The length of the sequence is set to 517.

4. A multifunctional bioactive peptide prediction method according to claim 1, characterized in that: In step S4, 80% of the multifunctional bioactive peptide dataset is divided into training set and validation set according to a five-fold cross-validation method, and the remaining 20% is used as a test set.

5. A system for implementing the multifunctional bioactive peptide prediction method according to any one of claims 1 to 4, characterized in that: include The data preprocessing module converts the amino acid sequence data in the multifunctional bioactive peptide dataset into digital data and performs sequence completion; The feature extractor, which includes a temporal convolutional neural network and a gated recurrent unit (GRU) for feature extraction, first uses a multi-scale TCN to obtain information about sequence patterns of different lengths in the sequence, then uses a GRU to obtain the relationship between each sequence pattern, and finally obtains a feature representation with global information; The target classifier is a fully connected neural network with classification function, which classifies the multifunctional bioactive peptide data; The data set includes a training set and a validation set divided from the multifunctional bioactive peptide data set according to a five-fold cross-validation method. The training set is used to construct and train a multifunctional bioactive peptide prediction model, and the validation set is used to adjust the parameters of the model. The model predicts the functions of peptide sequences.

Images