Antibacterial Peptide Generation Method, Device and Computer Equipment Based on Generative Model

Abstract

The present application relates to a method, apparatus and computer equipment for generating antimicrobial peptides based on a generative model, and involves the fields of artificial intelligence and bioinformatics. The method comprises obtaining a set of existing antimicrobial peptide sequences; using a pre-trained generative model to predict novel antimicrobial peptides based on the existing antimicrobial peptide sequences to obtain predicted sequences; inputting the predicted sequences into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequences have antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set; and outputting the predicted sequences having antimicrobial activity as the final target antimicrobial peptide sequence. This method solves the problem of low efficiency in antimicrobial peptide design and generation, and can achieve de novo generation and determination of antimicrobial peptides with the help of a generative model and classifier, thereby improving generation efficiency and accuracy.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence and bioinformatics, and in particular to a method, apparatus, and computer device for generating antimicrobial peptides based on a generative model. Background Art

[0002] Antimicrobial peptides (AMPs) are a class of short peptide molecules composed of amino acids. Due to their broad-spectrum antimicrobial properties, they have a wide range of applications in anti-infection therapy, food preservation, and biopharmaceuticals. Current AMP databases include both natural and synthetic AMPs. Natural AMPs are metabolites produced by organisms with antimicrobial properties, while synthetic AMPs are proteins synthesized using existing synthetic techniques and share the same biochemical properties as natural AMPs. Given the limited number of natural AMPs, researchers are hoping to design and develop a wider range of AMPs through synthetic methods.

[0003] However, traditional antimicrobial peptide design mostly relies on expert experience and manual trial and error, which has a long production cycle, high cost, and a low success rate.

[0004] There is currently no effective solution to the problem of low efficiency in the design and generation of antimicrobial peptides in related technologies. Summary of the Invention

[0005] In this embodiment, a method, apparatus, and computer device for generating antimicrobial peptides based on a generative model are provided to solve the problem of low efficiency in antimicrobial peptide design in related technologies.

[0006] In a first aspect, this embodiment provides a method for generating antimicrobial peptides based on a generative model, the method comprising:

[0007] Obtaining a collection of existing antimicrobial peptide sequences;

[0008] Using a pre-trained generative model, predicting a novel antimicrobial peptide based on the existing antimicrobial peptide sequence to obtain a predicted sequence;

[0009] Inputting the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set;

[0010] The predicted sequence with antimicrobial activity is output as the final target antimicrobial peptide sequence.

[0011] In some embodiments, obtaining a collection of existing antimicrobial peptide sequences includes:

[0012] Initial antimicrobial peptide data were extracted from public databases;

[0013] The initial antimicrobial peptide data are cleaned to obtain a set of existing antimicrobial peptide sequences.

[0014] In some embodiments, a pre-trained generative model is used to predict novel antimicrobial peptides based on the existing antimicrobial peptide sequences to obtain predicted sequences, including:

[0015] Based on the multiple sequence alignment technology, the existing set of antimicrobial peptide sequences is expanded to obtain an antimicrobial peptide multiple sequence alignment dataset;

[0016] The antimicrobial peptide multiple sequence alignment dataset is input into a pre-trained generative model to generate a predicted sequence based on an arbitrary decoding order.

[0017] In some embodiments, the pre-trained generative model uses an order-agnostic autoregressive diffusion model.

[0018] In some embodiments, the predicted sequence is input into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity, comprising:

[0019] Performing physical and chemical property screening on the predicted sequence to obtain a screened sequence;

[0020] The screened sequence is input into a pre-trained antimicrobial peptide classifier, and a conclusion of whether the screened sequence has antimicrobial activity is output.

[0021] In some embodiments, the method further comprises:

[0022] Training a binary classification model based on the obtained antimicrobial peptide label set and the non-antimicrobial peptide sequence label set;

[0023] During the training process of the binary classification model, the parameters of the binary classification model are adjusted based on the ten-fold cross validation results to obtain a pre-trained antimicrobial peptide classifier.

[0024] In some embodiments, training a binary classification model based on the obtained antimicrobial peptide label set and the non-antimicrobial peptide sequence label set includes:

[0025] Performing feature extraction on the antimicrobial peptide tag set and the non-antimicrobial peptide sequence tag set to obtain antimicrobial peptide features and non-antimicrobial peptide features;

[0026] The antimicrobial peptide features and the non-antimicrobial peptide features are input into the binary classification model to train the binary classification model.

[0027] In a second aspect, this embodiment provides an antimicrobial peptide generation device based on a generative model, the device comprising:

[0028] An existing data acquisition module, used to acquire a collection of existing antimicrobial peptide sequences;

[0029] A new sequence prediction module, for obtaining a predicted sequence with a structure similar to the existing antimicrobial peptide sequence using a pre-trained generative model;

[0030] An activity verification module is used to input the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set;

[0031] The target antimicrobial peptide output module is used to output the predicted sequence with antimicrobial activity as the final target antimicrobial peptide sequence.

[0032] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the generative model-based antimicrobial peptide generation method described in the first aspect is implemented.

[0033] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the generative model-based antimicrobial peptide generation method described in the first aspect.

[0034] Compared with related technologies, the generative model-based antimicrobial peptide generation method, apparatus, and computer device provided in this embodiment obtain a set of existing antimicrobial peptide sequences; utilize a pre-trained generative model to obtain a predicted sequence; input the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set; and output the predicted sequence with antimicrobial activity as the final target antimicrobial peptide sequence. This solves the problem of low efficiency in antimicrobial peptide design and generation, and enables the de novo generation and determination of antimicrobial peptides with the help of a generative model and classifier, thereby improving generation efficiency and accuracy.

[0035] The details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0037] Figure 1FIG1 is a hardware structure block diagram of a terminal for a method for generating antimicrobial peptides based on a generative model in one embodiment;

[0038] Figure 2 1 is a schematic diagram of a process for generating antimicrobial peptides based on a generative model in one embodiment;

[0039] Figure 3 FIG1 is a schematic diagram of a performance curve of an antimicrobial peptide classifier in a ten-fold cross validation in one embodiment;

[0040] Figure 4 Schematic diagram of a process for generating antimicrobial peptides based on a generative model in a preferred embodiment;

[0041] Figure 5 4 is a structural block diagram of an antimicrobial peptide generation device based on a generative model in one embodiment. DETAILED DESCRIPTION

[0042] In order to more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0043] Unless otherwise defined, technical or scientific terms used in this application shall have the ordinary meanings as understood by persons of ordinary skill in the art to which this application belongs. The terms "a," "an," "the," "these," and similar expressions in this application do not denote limitations on quantity and may be singular or plural. The terms "comprise," "include," "have," and any variations thereof, as used in this application, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device comprising a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include unlisted steps or modules (units) or other steps or modules (units) inherent to the process, method, product, or device. The terms "connected," "connected," "coupled," and similar expressions used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. As used in this application, "plurality" means two or more. "And / or" describes an association between associated objects, indicating that three possible relationships exist. For example, "A and / or B" may mean: A exists alone; A and B exist simultaneously; or B exists alone. Generally, the character " / " indicates that the objects in the preceding and following relationship are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0044] The method embodiment provided in this embodiment can be executed in a terminal, a computer or a similar computing device. For example, running on a terminal, Figure 1 FIG. 1 is a block diagram of the hardware structure of the terminal of the antimicrobial peptide generation method based on the generative model of this embodiment. Figure 1 As shown, the terminal may include one or more ( Figure 1 The processor 102 (only one is shown) and a memory 104 for storing data, wherein the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA. The terminal may also include a transmission device 106 for communication functions and an input / output device 108. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above terminal. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0045] Memory 104 can be used to store computer programs, such as software programs and modules for application software, such as the computer program corresponding to the generative model-based antimicrobial peptide generation method in this embodiment. Processor 102 executes the computer program stored in memory 104 to perform various functional applications and data processing, thereby implementing the above-mentioned method. Memory 104 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 104 may further include memory remotely located relative to processor 102, and such remote memory may be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0046] Transmission device 106 is used to receive or transmit data via a network. This network may include a wireless network provided by the terminal's communications provider. In one embodiment, transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0047] In this embodiment, a method for generating antimicrobial peptides based on a generative model is provided. Figure 2 is a flow chart of the method for generating antimicrobial peptides based on the generative model of this embodiment, as shown in FIG. Figure 2 As shown, the process includes the following steps:

[0048] Step S210: Acquire a set of existing antimicrobial peptide sequences.

[0049] Specifically, in actual application, the methods for obtaining existing antimicrobial peptide sequences in the embodiments of the present application include, but are not limited to, obtaining them from a pre-stored database, or crawling antimicrobial peptide data with antimicrobial activity from literature reports on a network platform, and performing operations such as merging, deduplication, and filtering on the obtained antimicrobial peptide data sequences to obtain a set of existing antimicrobial peptide sequences. The embodiments of the present application do not limit the method for obtaining antimicrobial peptide sequences.

[0050] In step S220, the pre-trained generative model is used to predict new antimicrobial peptides based on existing antimicrobial peptide sequences to obtain a predicted sequence.

[0051] Specifically, based on multiple sequence alignment technology, the existing set of antimicrobial peptide sequences is expanded to obtain an antimicrobial peptide multiple sequence alignment dataset. This dataset is then input into a pre-trained diffusion model to generate a predicted sequence based on an arbitrary decoding order. In other embodiments, step S220 can also be implemented by training a large pre-trained language model (e.g., the GPT series) on a corpus containing existing antimicrobial peptide sequence information and annotations such as properties, functions, and similarity, thereby enabling the model to learn the syntax, semantics, and underlying features of antimicrobial peptide sequences and obtain a predicted sequence. In other embodiments, a long short-term memory (LSTM) network can be used to generate candidate antimicrobial peptide sequences. These candidate sequences are then input into a Transformer for decoding and optimization to obtain an expanded antimicrobial peptide sequence, i.e., the predicted sequence.

[0052] Step S230: input the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set.

[0053] Specifically, antimicrobial peptide classifiers include, but are not limited to, models such as support vector machines, neural networks, and decision trees to learn the characteristics of antimicrobial peptides. Antimicrobial peptide tag sets can be derived from a collection of existing antimicrobial peptide sequences, while non-antimicrobial peptide tag sets include various protein sequences that are not antimicrobial peptides.

[0054] In step S240 , the predicted sequence with antimicrobial activity is output as the final target antimicrobial peptide sequence.

[0055] Specifically, the target antimicrobial peptide sequence can be further processed or analyzed, and the results can be applied in different scenarios. For example, the tumor suppression properties of the target antimicrobial peptide sequence can be analyzed for medical applications, or the bactericidal effects of the target antimicrobial peptide sequence can be analyzed, and target antimicrobial peptide sequences that meet bactericidal standards can be used as preservatives for food safety applications. Applications in scientific research, agriculture, and other fields are also possible.

[0056] In this embodiment, a set of existing antimicrobial peptide sequences is obtained; a pre-trained generative model is used to obtain predicted sequences with similar structures to the existing antimicrobial peptide sequences; the predicted sequences are input into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequences have antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set; and the predicted sequences with antimicrobial activity are output as the final target antimicrobial peptide sequence. This solves the problem of low efficiency in antimicrobial peptide design and generation, enables the de novo generation and identification of antimicrobial peptides with the help of the generative model and classifier, and significantly improves the efficiency and success rate of antimicrobial peptide design.

[0057] In some embodiments, step S210, obtaining a set of existing antimicrobial peptide sequences, includes:

[0058] Step S211: extracting initial antimicrobial peptide data from a public database.

[0059] Step S212: Clean the initial antimicrobial peptide data to obtain a set of existing antimicrobial peptide sequences.

[0060] Specifically, the existing collection of antimicrobial peptide sequences, the AMP dataset, was compiled from the following six public databases: the Antimicrobial Peptide Database (APD); the Collection of Antimicrobial Peptides (DADP); the Database of Antimicrobial Activity and Structure of Peptides (DBAASP); the Database of Research on Antimicrobial Peptides (DRAMP); the Yet Another Database of Antimicrobial Peptides (YADAMP); and the Database of Antimicrobial Peptides (dbAMP). Data extracted from these databases were merged, deduplicated, and filtered to remove incomplete or meaningless data entries, thereby improving the quality of the AMP dataset (i.e., the existing collection of antimicrobial peptide sequences).

[0061] In some embodiments, based on step S220, a pre-trained generative model is used to obtain a predicted sequence that is structurally similar to an existing antimicrobial peptide sequence, including:

[0062] Step S221 : Based on the multiple sequence alignment technology, the existing set of antimicrobial peptide sequences is expanded to obtain an antimicrobial peptide multiple sequence alignment data set.

[0063] Specifically, a multiple sequence alignment (MSA) was performed on an existing collection of antimicrobial peptide sequences (i.e., the AMP dataset) to generate an AMP-MSA dataset containing evolutionary information. In one embodiment, AMP-positive (i.e., antimicrobial activity) MSA sequences were extracted using a search tool to generate the AMP-MSA dataset.

[0064] Step S222: Input the antimicrobial peptide multiple sequence alignment dataset into a pre-trained generative model to generate a predicted sequence based on an arbitrary decoding order.

[0065] Specifically, based on the antimicrobial peptide multiple sequence alignment dataset, the generation length is set (e.g., set to 15-35 amino acids), and the generative model is used to predict new sequences.

[0066] The pre-trained generative model uses the order-agnostic autoregressive diffusion model (OADM). This model is trained on the evolving multiple sequence alignment (MSA) data used for training and allows sequences to be generated in any order, expanding the generative capabilities of traditional autoregressive models.

[0067] The OADM diffusion model is trained using the log-likelihood of the generated sequence as the loss function. The log-likelihood of the generated sequence is expressed as the expectation of all possible decoding orders:

[0068] ;

[0069] Among them, x σ(t) represents the amino acid generated at position t according to the decoding order σ. σ(＜t) represents all the previous amino acids arranged in this order, where logp(x) represents the log-likelihood value of generating the x sequence, Eσ~U(S L ) represents the expected value of all possible decoding orders, L represents the length of the sequence, S L Represents all possible decoding orders.

[0070] During training, the model processes the MSA matrix through an alternating axial attention mechanism, reducing the row attention computation complexity to O(ML 2 ), the column attention complexity is reduced to O(LM 2 ), and combined with the tied row attention mechanism to share inter-sequence structure.

[0071] The model is pre-trained with the Masked Language Modeling (MLM) objective, and its loss function is expressed as:

[0072] ;

[0073] Among them, (m,i) represents the masking position, To correctly predict the probability of the masked amino acid at position (m,i), is the MSA sequence after masking, X represents the MSA sequence before masking, and θ represents the parameters of the model.

[0074] In some embodiments, based on S230, the predicted sequence is input into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity, including:

[0075] Based on S231, the predicted sequence was screened for physical and chemical properties to obtain the screened sequence.

[0076] Based on S232, the screened sequence is input into a pre-trained antimicrobial peptide classifier, and a conclusion on whether the screened sequence has antimicrobial activity is output.

[0077] Specifically, the screening conditions include isoelectric point, positive charge, and hydrophobicity. The purpose of the physicochemical property screening is to evaluate whether the sequence generated by the generator has AMP (antimicrobial peptide)-like properties.

[0078] In some embodiments, the method further comprises:

[0079] Step S250 : training a binary classification model based on the acquired antimicrobial peptide label set and non-antimicrobial peptide sequence label set.

[0080] Step S260 , during the training process of the binary classification model, the parameters of the binary classification model are adjusted based on the ten-fold cross validation results to obtain a pre-trained antimicrobial peptide classifier.

[0081] Specifically, ten-fold cross-validation divides all data into ten parts, each of which serves as a validation set, while the others serve as training sets for training and validation. During this process, hyperparameters are kept consistent, and the average training and validation losses of the ten models are taken to measure the quality of the hyperparameters. After achieving satisfactory hyperparameters, the entire data is used as the training set, and the final model is trained with those hyperparameters. The beneficial effect of cross-validation is that it reduces randomness and improves model generalization. Because single-partitioning of the training and validation sets can lead to randomness, multiple partitions of the existing dataset are fully utilized to avoid the accidental selection of hyperparameters and models that lack generalization ability due to special partitioning.

[0082] In some embodiments, based on step S250, a binary classification model is trained based on the acquired antimicrobial peptide label set and the non-antimicrobial peptide sequence label set, including:

[0083] Step S251 , feature extraction is performed on the antimicrobial peptide tag set and the non-antimicrobial peptide sequence tag set to obtain antimicrobial peptide features and non-antimicrobial peptide features.

[0084] Step S252: inputting the antimicrobial peptide features and the non-antimicrobial peptide features into a binary classification model to train the binary classification model.

[0085] Specifically, the feature extraction methods include the pseudo K-tuple reduced amino acid composition (PseKRAAC) encoding method and the quasi-sequence-order (QSOrder) encoding method.

[0086] After the features are extracted, the features are selected. Feature selection is a two-step process. First, the features are evaluated based on their Pearson correlation coefficient (PCC), which is calculated as follows:

[0087] ;

[0088] where y i represents the true target value, represents the predicted value, u y and where is the average of the true and predicted values, and N is the total number of samples. This step allows us to rank features based on their predictive power, thereby quantifying their effectiveness. Subsequently, the most effective features are selected and further evaluated using a multi-branch convolutional neural network with attention (MBC-Attention) model, resulting in the antimicrobial peptide and non-antimicrobial peptide features used to train the binary classification model.

[0089] After the feature engineering process, the feature data is used to train a binary classification model using the Extreme Gradient Boosting (XGBoost) decision tree algorithm. Antimicrobial peptide feature sequences are labeled as 1, and non-antimicrobial peptide feature sequences are labeled as 0. The XGBoost model optimizes a regularized objective function to balance the accuracy and complexity of the model to prevent overfitting. Objective function Defined as:

[0090] ;

[0091] Where n represents the number of samples in the training process, l represents the loss function, and k represents the index of each tree in the XGBoost ensemble model. is x i The prediction formula is: ;Ω(f) is the regularization term: ; where T represents the number of leaves in the tree, w j represents the weight of the (j)th1 leaf, γ controls the number of leaves, and λ controls the L2 norm of the leaf weight.

[0092] During training, the XGBoost model is built in an additive manner, optimizing the following objective L at each iteration t:

[0093] ;

[0094] Among them, n represents the number of samples in the training process, l represents the loss function, Ω(f t ) represents the regularization term at time t, f t (x i ) means that in the tth iteration of the training process, the new tree f t For sample x i Predictions made.

[0095] In the objective function L (t) middle, Calculated when adding new tree f t The overall loss of all samples in the dataset after improving the previous round of predictions. The regularization term Ω(f t ) for the new tree f t The complexity of the model is penalized to prevent overfitting.

[0096] Use the second-order Taylor expansion to approximate L:

[0097] ;

[0098] Among them, g i and h iare the first-order and second-order gradients, respectively.

[0099] XGBoost model tuning is based on the F1 score and AUC index, using 10-fold cross validation (k times 10) to prevent overfitting. The F1 score is a metric that considers both precision and recall to calculate a balanced measure of model accuracy, especially when dealing with imbalanced classes. It is defined as the harmonic mean of precision and recall:

[0100] ;

[0101] Among them, F1 Score represents the F1 score, Precision represents accuracy, and Recall (also known as sensitivity or true positive rate) is the proportion of actual positive cases correctly identified by the model, which is calculated as follows:

[0102] ;

[0103] Among them, TP stands for true positive and FN stands for false negative.

[0104] The best split for each node during training is determined by maximizing the gain:

[0105] ;

[0106] Among them, I L and I R Represents the left and right child nodes after segmentation, g i and h i are the first-order and second-order gradients, respectively. λ controls the L2 norm of the leaf node weights to prevent overfitting, and γ penalizes splits that increase the number of leaf nodes in the tree.

[0107] Apply shrinkage (learning rate η) to each tree’s predictions to scale them:

[0108] ;

[0109] in, Represents the updated sample after adding the t-th tree The predicted value of Represents the sample x before joining the t-th tree i The predicted value, f t (x i ) means that in the tth iteration of the training process, the new tree f t For sample x i η represents the learning rate, which prevents overfitting by reducing the influence of each tree.

[0110] The best performing model determined by cross-validation was used for subsequent analyses.

[0111] Figure 3 Figure 2 shows the receiver operating characteristic (ROC) performance curves of the XGBoost-based antimicrobial peptide classifier model in ten-fold cross-validation. Each curve represents a single-fold ROC curve, depicting the trade-off between true positive rate and false positive rate at different threshold settings. The average ROC curve, indicated by the bold line, summarizes the overall performance of the model, and the area under the curve (AUC) provides a measure of classification accuracy.

[0112] The present embodiment is described and illustrated below through preferred embodiments.

[0113] Figure 4 FIG. 1 is a flow chart of the method for generating antimicrobial peptides based on a generative model according to the preferred embodiment of the present invention. Figure 4 As shown, the method for generating antimicrobial peptides based on a generative model provided in this preferred embodiment includes:

[0114] S1. Construct an antimicrobial peptide dataset, i.e., a collection of existing antimicrobial peptide sequences. Antimicrobial peptide data were extracted from public antimicrobial peptide databases and merged, deduplicated, and filtered to remove incomplete or meaningless data entries. This dataset was then generated.

[0115] S2, constructing an antimicrobial peptide multiple sequence alignment dataset. Based on multiple sequence alignment technology, the existing collection of antimicrobial peptide sequences is expanded to obtain an antimicrobial peptide multiple sequence alignment dataset.

[0116] S3, the antimicrobial peptide multiple sequence alignment dataset is input into a pre-trained order-agnostic autoregressive diffusion model to generate predicted sequences based on arbitrary decoding order.

[0117] S4, screening the predicted sequence for physical and chemical properties to obtain a screened sequence.

[0118] S5, training a binary classification model based on the obtained antimicrobial peptide label set and non-antimicrobial peptide sequence label set; during the training process of the binary classification model, adjusting the parameters of the binary classification model based on the ten-fold cross validation results to obtain an antimicrobial peptide classifier.

[0119] S6, inputting the screened sequence into the antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity.

[0120] S7, outputting the sequence with antimicrobial activity as the final target antimicrobial peptide sequence.

[0121] This preferred embodiment proposes a generation and screening strategy that combines diffusion models, multiple sequence alignment data, and machine learning optimization tools to significantly improve the efficiency and success rate of antimicrobial peptide design, which is an important innovation in the fields of bioinformatics and protein engineering.

[0122] It should be noted that the steps shown in the above process or the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0123] This embodiment also provides a generative model-based antimicrobial peptide production device for implementing the aforementioned embodiments and preferred embodiments. Details already described will not be repeated. The terms "module," "unit," "subunit," etc., used below, may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented using software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0124] Figure 5 This is a structural block diagram of the antimicrobial peptide generation device based on the generative model of this embodiment, as shown in FIG. Figure 5 As shown, the device includes: an existing data acquisition module 51 , a new sequence prediction module 52 , an activity verification module 53 and a target antimicrobial peptide output module 54 .

[0125] An existing data acquisition module 51 is used to acquire a set of existing antimicrobial peptide sequences;

[0126] A new sequence prediction module 52 is used to predict new antimicrobial peptides based on existing antimicrobial peptide sequences using a pre-trained generative model to obtain a predicted sequence;

[0127] Activity verification module 53, used to input the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set;

[0128] The target antimicrobial peptide output module 54 is configured to output the predicted sequence having antimicrobial activity as the final target antimicrobial peptide sequence.

[0129] In some embodiments, obtaining the existing set of antimicrobial peptide sequences includes: extracting initial antimicrobial peptide data from a public database; and cleaning the initial antimicrobial peptide data to obtain the existing set of antimicrobial peptide sequences.

[0130] In some embodiments, a pre-trained generative model is used to obtain a predicted sequence with a structure similar to an existing antimicrobial peptide sequence, including: expanding the existing set of antimicrobial peptide sequences based on multiple sequence alignment technology to obtain an antimicrobial peptide multiple sequence alignment dataset; and inputting the antimicrobial peptide multiple sequence alignment dataset into the pre-trained generative model to generate a predicted sequence based on an arbitrary decoding order.

[0131] In some embodiments, the pre-trained generative model uses a sequence-agnostic autoregressive diffusion model.

[0132] In some embodiments, the predicted sequence is input into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity, including: screening the predicted sequence for physical and chemical properties to obtain a screened sequence; inputting the screened sequence into the pre-trained antimicrobial peptide classifier, and outputting a conclusion on whether the screened sequence has antimicrobial activity.

[0133] In some embodiments, the method further includes: training a binary classification model based on the obtained antimicrobial peptide label set and non-antimicrobial peptide sequence label set; during the training of the binary classification model, adjusting the parameters of the binary classification model based on the ten-fold cross-validation results to obtain a pre-trained antimicrobial peptide classifier.

[0134] In some embodiments, a binary classification model is trained based on the obtained antimicrobial peptide label set and non-antimicrobial peptide sequence label set, including: extracting features from the antimicrobial peptide label set and the non-antimicrobial peptide sequence label set to obtain antimicrobial peptide features and non-antimicrobial peptide features; and inputting the antimicrobial peptide features and the non-antimicrobial peptide features into the binary classification model to train the binary classification model.

[0135] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0136] This embodiment further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0137] Optionally, the computer device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0138] It should be noted that, for specific examples in this embodiment, reference may be made to the examples described in the above embodiments and optional implementation modes, and will not be repeated in this embodiment.

[0139] In addition, in conjunction with the generative model-based antimicrobial peptide generation methods provided in the above embodiments, this embodiment may also provide a storage medium for implementation. The storage medium stores a computer program; when executed by a processor, the computer program implements any of the generative model-based antimicrobial peptide generation methods described in the above embodiments.

[0140] It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit it. Based on the embodiments provided in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0141] Obviously, the accompanying drawings are merely examples or embodiments of the present application. A person skilled in the art can also apply the present application to other similar situations based on these drawings without inventive effort. Furthermore, it is understandable that, although the work involved in this development process may be complex and lengthy, certain design, manufacturing, or production changes based on the technical content disclosed in this application are merely routine technical means for a person skilled in the art and should not be considered to constitute a deficiency in the disclosure of the present application.

[0142] The term "embodiment" as used in this application refers to specific features, structures, or characteristics described in conjunction with the embodiment that can be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily mean that the embodiment is the same, nor does it mean that it is mutually exclusive with other embodiments and is independent or optional. It is understood, either explicitly or implicitly, by those skilled in the art that the embodiments described in this application can be combined with other embodiments when there is no conflict.

[0143] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for generating antimicrobial peptides based on a generative model, characterized in that: The method comprises: Based on the public database of antimicrobial peptides, a collection of existing antimicrobial peptide sequences is obtained; Using a pre-trained generative model, predicting a novel antimicrobial peptide based on the existing antimicrobial peptide sequence to obtain a predicted sequence; Inputting the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set; Outputting the predicted sequence with antibacterial activity as the final target antimicrobial peptide sequence; Wherein, using the pre-trained generative model, novel antimicrobial peptides are predicted based on the existing antimicrobial peptide sequences to obtain predicted sequences, including: Based on the multiple sequence alignment technology, the existing set of antimicrobial peptide sequences is expanded to obtain an antimicrobial peptide multiple sequence alignment dataset; Inputting the antimicrobial peptide multiple sequence alignment dataset into a pre-trained generative model to generate a predicted sequence based on an arbitrary decoding order; Among them, the pre-trained generative model adopts an order-agnostic autoregressive diffusion model, which is trained with the log-likelihood of the generated sequence as the loss function, and the log-likelihood of the generated sequence is expressed as the expectation of all possible decoding orders; during training, the order-agnostic autoregressive diffusion model processes the multi-sequence alignment matrix through an alternating axial attention mechanism, and combines the bound row attention mechanism to share the inter-sequence structure.

2. The method for generating antimicrobial peptides based on a generative model according to claim 1, wherein: Access a collection of existing antimicrobial peptide sequences, including: Initial antimicrobial peptide data were extracted from public databases; The initial antimicrobial peptide data are cleaned to obtain a set of existing antimicrobial peptide sequences.

3. The method for generating antimicrobial peptides based on a generative model according to claim 1, wherein: Inputting the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity comprises: Performing physical and chemical property screening on the predicted sequence to obtain a screened sequence; The screened sequence is input into a pre-trained antimicrobial peptide classifier, and a conclusion of whether the screened sequence has antimicrobial activity is output.

4. The method for generating antimicrobial peptides based on a generative model according to claim 1, wherein The method further comprises: Training a binary classification model based on the obtained antimicrobial peptide label set and the non-antimicrobial peptide sequence label set; During the training process of the binary classification model, the parameters of the binary classification model are adjusted based on the ten-fold cross validation results to obtain a pre-trained antimicrobial peptide classifier.

5. The method for generating antimicrobial peptides based on a generative model according to claim 3, wherein: Based on the obtained antimicrobial peptide label set and the non-antimicrobial peptide sequence label set, a binary classification model is trained, including: Performing feature extraction on the antimicrobial peptide tag set and the non-antimicrobial peptide sequence tag set to obtain antimicrobial peptide features and non-antimicrobial peptide features; The antimicrobial peptide features and the non-antimicrobial peptide features are input into the binary classification model to train the binary classification model.

6. A device for generating antimicrobial peptides based on a generative model, characterized in that: The device comprises: An existing data acquisition module is used to acquire a set of existing antimicrobial peptide sequences based on a public database of antimicrobial peptides; a new sequence prediction module, configured to predict a new antimicrobial peptide based on the existing antimicrobial peptide sequence using a pre-trained generative model to obtain a predicted sequence; Wherein, using the pre-trained generative model, novel antimicrobial peptides are predicted based on the existing antimicrobial peptide sequences to obtain predicted sequences, including: Based on multiple sequence alignment technology, the existing set of antimicrobial peptide sequences is expanded to obtain an antimicrobial peptide multiple sequence alignment dataset; the antimicrobial peptide multiple sequence alignment dataset is input into a pre-trained generative model to generate a predicted sequence based on an arbitrary decoding order; wherein the pre-trained generative model adopts an order-agnostic autoregressive diffusion model, which is trained using the log-likelihood of the generated sequence as a loss function, and the log-likelihood of the generated sequence is expressed as the expectation of all possible decoding orders; during training, the order-agnostic autoregressive diffusion model processes the multiple sequence alignment matrix through an alternating axial attention mechanism, and combines it with a bound row attention mechanism to share inter-sequence structure; An activity verification module is used to input the predicted sequence into a pre-trained antimicrobial peptide classifier to determine whether the predicted sequence has antimicrobial activity; the antimicrobial peptide classifier is trained based on an antimicrobial peptide label set and a non-antimicrobial peptide sequence label set; The target antimicrobial peptide output module is used to output the predicted sequence with antimicrobial activity as the final target antimicrobial peptide sequence.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images