Llps protein recognition method, apparatus, device, and storage medium
Patent Information
- Application Number
- CN202410013841.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-02
- Publication Date
- 2026-10-20
- Estimated Expiration
- 2044-01-02
AI Technical Summary
[0004]但是，目前基于生物化学的传统方法识别RNA依赖型的LLPS蛋白质既耗时又昂贵
[0023]第四方面，本申请提供一种计算机程序产品，当该计算机程序产品在电子设备中运行时，使得电子设备执行上述第一方面所述相关方法的步骤，以实现上述第一方面所述的方法。
Smart Images
Figure CN117688431B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, device and storage medium for LLPS protein recognition. Background Technology
[0002] RNA-dependent liquid-liquid phase separation (LLPS) proteins play crucial roles in cellular processes such as stress granule formation, deoxyribonucleic acid (DNA) repair, germ cell development, and protein translation regulation.
[0003] Abnormal behavior of RNA-dependent LLPS proteins is also associated with a variety of diseases, especially neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) and frontotemporal dementia.
[0004] However, current traditional biochemical methods for identifying RNA-dependent LLPS proteins are both time-consuming and expensive. Summary of the Invention
[0005] To address the aforementioned technical issues, this application provides a method, apparatus, device, and storage medium for identifying LLPS proteins, which can train a machine learning model capable of identifying RNA-dependent LLPS proteins, thereby reducing the time and cost of identification.
[0006] In a first aspect, this application provides an LLPS protein identification method, the method comprising: obtaining an initial sample set; the initial sample set including multiple liquid-liquid phase separated LLPS protein sequences; extracting a preset type of feature vector from each LLPS protein sequence; obtaining the dependency type corresponding to each LLPS protein sequence; the dependency type including RNA-dependent and non-RNA-dependent types; obtaining multiple training samples based on the preset type of feature vector; using the dependency type corresponding to each LLPS protein sequence as the label of the multiple training samples to obtain a training sample set; training a preset initial model based on the training sample set to obtain an LLPS protein identification model; the LLPS protein identification model being used to predict the dependency type corresponding to an LLPS protein sequence.
[0007] Optionally, the preset feature vectors include: amino acid composition ACC feature vector, K-spaced amino acid pair composition CKSAAP feature vector, Geary autocorrelation feature vector, and conjugated triplet CTriad feature vector.
[0008] Optionally, the preset categories include multiple categories; based on the feature vectors of the preset categories extracted from each LLPS protein sequence, multiple training samples are obtained, including: calculating the mutual information between the feature vector of each category and the target value; sorting the feature vectors of multiple categories according to the first order of mutual information to obtain a sorting result; the first order includes ascending or descending order; selecting the feature vector of the category with the largest mutual information with the target value from the sorting result to complete one selection; repeatedly selecting the feature vector with the largest mutual information with the target value from the feature vectors of multiple categories until the number of feature vector categories reaches a preset threshold to obtain the feature vectors of the target categories selected multiple times; using the feature vector of the target category of each LLPS protein sequence as a training sample to obtain multiple training samples.
[0009] Optionally, before calculating the mutual information between the feature vector of each of the multiple categories and the target value, the method further includes: calculating the correlation coefficient between the feature vectors of the first category and the feature vector of the second category among the feature vectors of the multiple categories; the first category and the second category are different; if the correlation coefficient between the feature vectors of the first category and the feature vector of the second category is greater than a preset correlation coefficient threshold, then the feature vectors of the first category and the feature vector of the second category are deleted.
[0010] Optionally, the correlation coefficient is determined by the following formula:
[0011]
[0012] Where Cr represents the correlation coefficient; x i This represents the value of the eigenvector of the first category; y represents the average value of the eigenvectors of the first class; i This represents the value of the feature vector of the second type; This represents the average value of the eigenvectors of the second type.
[0013] Optionally, the method further includes: obtaining the protein sequence to be identified; extracting the feature vector of the target type from the protein sequence; and inputting the feature vector of the target type into the LLPS protein identification model to obtain the dependency type corresponding to the protein sequence to be identified.
[0014] The LLPS protein identification method provided in this application can obtain an initial sample set, which includes multiple LLPS protein sequences; extract feature vectors of a preset category from each LLPS protein sequence; obtain the dependency type corresponding to each LLPS protein sequence; the dependency type includes RNA-dependent and non-RNA-dependent types; obtain multiple training samples based on the feature vectors of the preset categories; use the dependency type corresponding to each LLPS protein sequence as the label of the multiple training samples to obtain a training sample set; train a preset initial model based on the training sample set to obtain an LLPS protein identification model; the LLPS protein identification model is used to predict the dependency type corresponding to LLPS protein sequences. Compared with the current traditional biochemical methods for identifying RNA-dependent LLPS proteins, this application can train a machine learning model capable of identifying RNA-dependent LLPS proteins, reducing the time and cost of identification.
[0015] Secondly, this application provides an LLPS protein recognition device, which includes: an acquisition module and a processing module; the acquisition module is used to acquire an initial sample set; the initial sample set includes multiple liquid-liquid phase separated LLPS protein sequences; the processing module is used to extract a feature vector of a preset category from each LLPS protein sequence; the acquisition module is also used to acquire the dependency type corresponding to each LLPS protein sequence; the dependency type includes RNA-dependent and non-RNA-dependent types; the processing module is also used to obtain multiple training samples based on the feature vector of the preset category; the dependency type corresponding to each LLPS protein sequence is used as the label of the multiple training samples to obtain a training sample set; a preset initial model is trained based on the training sample set to obtain an LLPS protein recognition model; the LLPS protein recognition model is used to predict the dependency type corresponding to the LLPS protein sequence.
[0016] Optionally, the preset feature vectors include: amino acid composition ACC feature vector, K-spaced amino acid pair composition CKSAAP feature vector, Geary autocorrelation feature vector, and conjugated triplet CTriad feature vector.
[0017] Optionally, the preset categories include multiple categories; the processing module is specifically used to calculate the mutual information between the feature vector of each category and the target value; sort the feature vectors of multiple categories according to the first order of mutual information to obtain the sorting result; the first order includes ascending or descending order; from the sorting result, select the feature vector of the category with the largest mutual information with the target value to complete one selection; repeatedly select the feature vector with the largest mutual information with the target value from the feature vectors of multiple categories until the number of feature vector categories reaches a preset threshold to obtain the feature vectors of the target categories selected multiple times; use the feature vector of the target category of each LLPS protein sequence as a training sample to obtain multiple training samples.
[0018] Optionally, before calculating the mutual information between the feature vectors of each of the multiple categories and the target value, the processing module is also used to calculate the correlation coefficient between the feature vectors of the first category and the feature vectors of the second category among the feature vectors of the multiple categories; the first category and the second category are different; if the correlation coefficient between the feature vectors of the first category and the feature vectors of the second category is greater than a preset correlation coefficient threshold, then the feature vectors of the first category and the feature vectors of the second category are deleted.
[0019] Optionally, the correlation coefficient is determined by the following formula:
[0020]
[0021] Where Cr represents the correlation coefficient; x i This represents the value of the eigenvector of the first category; y represents the average value of the eigenvectors of the first class; i This represents the value of the feature vector of the second type; This represents the average value of the eigenvectors of the second type.
[0022] Thirdly, this application provides an electronic device including a processor and a memory; the memory stores processor-executable instructions; when the processor is configured to execute the instructions, the electronic device performs the method described in the first aspect above.
[0023] Fourthly, this application provides a computer program product that, when run in an electronic device, causes the electronic device to perform the steps of the related method described in the first aspect, so as to implement the method described in the first aspect.
[0024] Fifthly, this application provides a readable storage medium comprising: software instructions; when the software instructions are executed in an electronic device, they cause the electronic device to perform the method described in the first aspect above.
[0025] The beneficial effects of the second to fifth aspects mentioned above can be referred to the first aspect, and will not be repeated here. Attached Figure Description
[0026] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0027] Figure 1 A schematic diagram illustrating the composition of the electronic device provided in the embodiments of this application;
[0028] Figure 2 A schematic flowchart of the LLPS protein recognition method provided in the embodiments of this application;
[0029] Figure 3 Another flowchart illustrating the LLPS protein recognition method provided in this application embodiment;
[0030] Figure 4 This is another schematic flowchart of the LLPS protein recognition method provided in the embodiments of this application;
[0031] Figure 5 This is a schematic diagram of the composition of the LLPS protein recognition device provided in the embodiments of this application. Detailed Implementation
[0032] Hereinafter, the terms "first," "second," and "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first," "second," or "third," etc., may explicitly or implicitly include one or more of that feature.
[0033] RNA-dependent LLPS proteins play crucial roles in cellular processes such as stress granule formation, DNA repair, germ cell development, and protein translation regulation.
[0034] Abnormal behavior of RNA-dependent LLPS proteins is also associated with a variety of diseases, especially neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) and frontotemporal dementia.
[0035] However, traditional biochemical-based methods for identifying RNA-dependent LLPS proteins are both time-consuming and expensive.
[0036] Based on this, embodiments of this application provide a method, apparatus, device, and storage medium for identifying LLPS proteins, which can train a machine learning model capable of identifying RNA-dependent LLPS proteins, reducing the time and cost of identification.
[0037] The LLPS protein recognition method provided in this application embodiment is executed by an LLPS protein recognition device. This device can be an electronic device with computing processing capabilities, such as a computer or server. Optionally, the device can also be an application (APP) with LLPS protein recognition function installed in the aforementioned electronic device; or, the device can be a processor (e.g., a central processing unit (CPU)) in the aforementioned electronic device; or, the device can be a software system or platform deployed in the aforementioned electronic device; or, the device can be a functional module in the aforementioned electronic device used to execute the LLPS protein recognition method, etc. This application embodiment does not impose any limitations on these aspects.
[0038] For simplicity, the following description will use the LLPS protein recognition device as an example of an electronic device.
[0039] Figure 1 This is a schematic diagram illustrating the composition of an electronic device provided in an embodiment of this application. For example... Figure 1 As shown, the electronic device may include: a processor 10, a memory 20, a communication line 30, a communication interface 40, and an input / output interface 50.
[0040] The processor 10, memory 20, communication interface 40, and input / output interface 50 can be connected via communication line 30.
[0041] Processor 10 is used to execute instructions stored in memory 20 to implement the LLPS protein recognition method provided in the following embodiments of this application. Processor 10 can be a CPU, a network processor (NP), a digital signal processor (DSP), a microprocessor, a microcontroller (MCU) / single-chip microcomputer / microcontroller, a programmable logic device (PLD), or any combination thereof. Processor 10 can also be any other device with processing capabilities, such as a circuit, device, or software module; this application embodiment does not limit this. In one example, processor 10 may include one or more CPUs, for example... Figure 1CPU0 and CPU1 are mentioned. As an optional implementation, the electronic device may include multiple processors; for example, in addition to processor 10, it may also include processor 60. Figure 1 (The example shown is a dashed line).
[0042] The memory 20 is used to store instructions. For example, the instructions may be computer programs. Optionally, the memory 20 may be a read-only memory (ROM) or other types of static storage devices that can store static information and / or instructions; it may also be a random access memory (RAM) or other types of dynamic storage devices that can store information and / or instructions; it may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, etc. The embodiments of this application do not limit this.
[0043] It should be noted that the memory 20 can exist independently of the processor 10 or it can be integrated with the processor 10. The memory 20 can be located inside or outside the electronic device, and this application embodiment does not impose any restrictions on this.
[0044] Communication line 30 is used to transmit information between the components included in the electronic device.
[0045] Communication interface 40 is used to communicate with other devices or other communication networks. These other communication networks can be Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc. Communication interface 40 can be a module, circuit, transceiver, or any device capable of enabling communication.
[0046] Input / output interface 50 is used to enable human-computer interaction between users and electronic devices. For example, it enables action interaction or information exchange between users and electronic devices.
[0047] For example, the input / output interface 50 can be a mouse, keyboard, display screen, or touch screen. Action or information interaction between the user and the electronic device can be achieved through a mouse, keyboard, display screen, or touch screen.
[0048] It should be noted that, Figure 1 The structures shown do not constitute a limitation on electronic devices, except... Figure 1 In addition to the components shown, the electronic device may include more or fewer components than illustrated (e.g., only processor 10 and memory 20), or combinations of certain components, or different component arrangements.
[0049] The following describes the LLPS protein recognition method provided in the embodiments of this application.
[0050] Figure 2 This is a schematic flowchart of the LLPS protein recognition method provided in an embodiment of this application. Figure 2 As shown, the method includes S101 to S106.
[0051] S101. The electronic device acquires the initial sample set.
[0052] The initial sample set included multiple LLPS protein sequences.
[0053] Alternatively, the electronic device can obtain LLPS protein sequences from a database storing LLPS protein sequences by downloading or transferring them to an intermediate storage medium.
[0054] For example, databases storing LLPS protein sequences may include LLPSDB, PhaSePro, and PhaSepDB.
[0055] Optionally, since redundant protein sequences may lead to overfitting and add unnecessary complexity to the training process, the electronic device can also remove protein sequences with similarity exceeding a similarity threshold after obtaining the initial sample set.
[0056] For example, the initial sample set may include multiple LLPS protein sequences, including positive samples (RNA-dependent LLPS protein sequences) and negative samples (non-RNA-dependent LLPS protein sequences). For positive samples, the similarity threshold can be set to 80%, and for negative samples, the similarity threshold can be set to 40%.
[0057] Alternatively, the electronic device may specifically utilize the CD-HIT tool to remove protein sequences whose similarity exceeds a similarity threshold.
[0058] S102, The electronic device extracts a feature vector of a preset type from each LLPS protein sequence.
[0059] Optionally, the feature vectors of the preset types may include amino acid composition (ACC) feature vectors, K-spaced amino acid pair composition (CKSAAP) feature vectors, Geary autocorrelation feature vectors, and conjugate triad (CTriad) feature vectors.
[0060] Alternatively, the AAC feature vector can be calculated using the following formula (1):
[0061]
[0062] In formula (1), f(t) represents the ACC eigenvector, which is the frequency of amino acid t in the LLPS protein sequence. N represents the length of the LLPS protein sequence. N(t) represents the number of times amino acid t appears in the LLPS protein sequence.
[0063] Alternatively, the CKSAAP feature vector can be calculated using the following formula (2):
[0064]
[0065] In formula (2), F represents the CKSAAP feature vector. AA This represents the frequency of amino acid AA at (K=0). K represents the frequency at which an amino acid is separated from any other amino acid. N t This indicates the total number of amino acids with zero spacers.
[0066] Alternatively, the Geary autocorrelation eigenvector can be calculated using the following formula (3):
[0067]
[0068] Where C(d) represents the Geary autocorrelation eigenvector, and nlag represents the maximum lag value. P i This represents the characteristic of the amino acid located at the i-th amino acid. i+d This represents the amino acid characteristics of the (i+d)th amino acid. This represents the average amino acid signature of an LLPS protein of length N.
[0069] Optionally, electronic devices can treat three consecutive amino acids as a unit and use the CTriad feature vector to evaluate the characteristics of an amino acid and its surrounding amino acids. Using this feature encoding method, the LLPS protein sequence first needs to be represented by a binary space (V,E), where V represents the vector control of the feature sequence, Vi represents a specific type of triplet, E represents the number of vectors V, and fi represents the value of the i-th dimension of F. The number of various types of Vi can be found in the LLPS protein sequence. Taking seven groups of amino acids as an example, the size of V is 7×7×7=343. The larger the value of Fi, the more difficult it is to compare protein sequences of different lengths. Therefore, fi can be normalized by the following formula (4) to obtain the CTriad feature vector:
[0070]
[0071] Where, d i This represents the CTriad feature vector.
[0072] S103. Electronic devices acquire the dependency type corresponding to each LLPS protein sequence.
[0073] Among them, the dependency types include RNA-dependent and non-RNA-dependent types.
[0074] In one possible implementation, as described above, the electronic device can obtain LLPS protein sequences from a database that stores LLPS protein sequences. This database may store the dependency types corresponding to each LLPS protein sequence, and the electronic device can directly obtain the dependency types corresponding to each LLPS protein sequence from this database.
[0075] In another possible implementation, the electronic device can receive the dependency type corresponding to each LLPS protein sequence input by the operator. As mentioned above, the electronic device may include an input / output interface 50, which may be a mouse, keyboard, or touch screen, etc. The electronic device can receive the dependency type corresponding to each LLPS protein sequence input by the operator through the mouse, keyboard, or touch screen, etc.
[0076] S104. The electronic device obtains multiple training samples based on the feature vectors of preset categories.
[0077] In some possible embodiments, the preset category may include multiple categories (e.g., the four categories described in S102 above), and the electronic device may select feature vectors of the preset categories to obtain feature vectors of the target category as training samples. In this case, Figure 3 This is another schematic flowchart illustrating the LLPS protein recognition method provided in an embodiment of this application. Figure 3 As shown, the above S104 may specifically include S1041 to S1045.
[0078] S1041. The electronic device calculates the mutual information between the feature vector of each of the multiple categories and the target value.
[0079] Alternatively, the electronic device can calculate mutual information using the following formula (5):
[0080] Mi(x,y)=∑∑P(X=x,Y=y)×log(P(X=x,Y=y)) / (P(X=x)×P(Y=y) Formula (5)
[0081] In formula (5), Mi(x,y) represents the mutual information between the eigenvector x and the target value y. P(X=x,Y=y) represents the joint probability function. P(X=x) and P(Y=y) represent the marginal probability functions.
[0082] S1042. The electronic device sorts the feature vectors of multiple types according to the first order of mutual information to obtain the sorting result.
[0083] The first order includes either from smallest to largest or from largest to smallest.
[0084] S1043. The electronic device selects the feature vector of the category with the largest mutual information with the target value from the sorting results, thus completing one selection.
[0085] S1044. The electronic device repeatedly selects the feature vector with the largest mutual information with the target value from multiple types of feature vectors until the number of feature vector types reaches a preset threshold, thus obtaining the feature vector of the target type selected multiple times.
[0086] S1045. The electronic device uses the feature vector of each target type of each LLPS protein sequence as a training sample to obtain multiple training samples.
[0087] In some possible embodiments, the electronic device may also remove feature vectors of highly correlated classes before calculating mutual information. In this case, Figure 4 This is another schematic flowchart illustrating the LLPS protein recognition method provided in this application. Figure 4 As shown, before S1041 above, the method may also include S201 to S202.
[0088] S201. The electronic device calculates the correlation coefficient between the first and second type of feature vectors among multiple types of feature vectors.
[0089] The first type is different from the second type.
[0090] Alternatively, the correlation coefficient can be determined using the following formula (6):
[0091]
[0092] In formula (6), Cr represents the correlation coefficient; x i This represents the value of the eigenvector of the first category; y represents the average value of the eigenvectors of the first class; i This represents the value of the feature vector of the second type; This represents the average value of the eigenvectors of the second type.
[0093] S202. If the correlation coefficient between the feature vectors of the first type and the second type is greater than the preset correlation coefficient threshold, the electronic device deletes the feature vectors of the first type and the second type.
[0094] The correlation coefficient threshold can be preset in the electronic device by the staff. For example, the correlation coefficient threshold can be set to 0.7 or 0.8, etc. The specific value of the correlation coefficient threshold is not limited in this application embodiment.
[0095] S105. The electronic device uses the dependency type corresponding to each LLPS protein sequence as the label of multiple training samples to obtain a training sample set.
[0096] Optionally, the electronic device can use the dependency type corresponding to each LLPS protein sequence as the label of the training sample corresponding to that LLPS protein sequence.
[0097] For example, taking multiple LLPS protein sequences including sequence 1, sequence 2, and sequence 3 as an example, assuming that training sample 1 is obtained based on the feature vector of the target type of sequence 1, training sample 2 is obtained based on the feature vector of the target type of sequence 2, and training sample 3 is obtained based on the feature vector of the target type of sequence 3, then the electronic device can use the dependency type corresponding to sequence 1 as the label of training sample 1, the dependency type corresponding to sequence 2 as the label of training sample 2, and the dependency type corresponding to sequence 3 as the label of training sample 3.
[0098] S106. The electronic device trains the preset initial model based on the training sample set to obtain the LLPS protein recognition model.
[0099] Among them, the LLPS protein recognition model is used to predict the dependency type corresponding to the LLPS protein sequence.
[0100] Alternatively, the initial model can be a random forest (RF) model, an extreme gradient boosting (XG-Boost) model, a K-nearest-neighbors (KNN) model, an artificial neural network (ANN) model, a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, etc.
[0101] Random forest classifier is a powerful machine learning algorithm widely used for prediction in bioinformatics. To make predictions, this classifier uses an ensemble of multiple decision trees. The target variable is predicted independently by each decision tree in the random forest, and then the predictions from each tree are summed to obtain a comprehensive prediction. Random forest can be defined using the following formula (7):
[0102]
[0103] In formula (7), x represents a vector. The attributes representing the x-vector. Represents a decision tree.
[0104] XG-boost uses gradient boosting trees for prediction. The final prediction is the sum of all predictions made by each tree. XG-boost can be defined as follows (8):
[0105]
[0106] In formula (8), xi and fk(xi) are the scores of the i-th training sample and the k-th tree, respectively. The objective function of XG-boost can be defined as:
[0107]
[0108] In formula (9), n represents the total number of samples, and y t , and y p Let l(y) represent the target value and predicted value of data sample i, respectively. t ,y p ) is the loss function, K and f k Let represent the total number of trees and the k-th tree, respectively. It is a regularization term that reduces overfitting.
[0109] The K-Nearest Neighbors (KNN) algorithm makes predictions by calculating the distance between sample points using Euclidean distance. The formula for Euclidean distance can be expressed as follows (10):
[0110]
[0111] In formula (10), r and s represent two samples, n represents the total feature dimension, and r i and s i These are the i-th dimensions of features r and s, respectively.
[0112] An artificial neural network (ANN) consists of interconnected neurons arranged in a hierarchical manner that process input data and produce outputs. Mathematically, an artificial neural network can be represented by the following formula (11):
[0113]
[0114] In formula (11), a (0) z represents the output of the input layer. (i) This represents a weighted sum, which is achieved through the weight matrix W. i Input to activate a i-1 and deviation b i The weighted sum is then used with the activation function f. (i) Calculate the output activation a (i) and used as input a for the next layer (i+1) This process is repeated until the output layer.
[0115] A CNN is a feedforward neural network consisting of convolutional layers, pooling layers, and fully connected layers. It learns from input data through convolution and pooling operations. A convolutional neural network can be represented by the following formula (12):
[0116]
[0117] In formula (12), Y represents the output, α represents the activation function, N represents the number of convolutional filters, and W... i This represents the i-th weighted convolutional filter, * represents the convolution operation, X represents the input vector, and b represents the bias term.
[0118] RNNs are designed for sequential data processing. Recurrent neural networks can maintain a hidden state, allowing them to retain and utilize information from the previous step. The basic mathematical expression for a recurrent neural network is shown in the following formula (13):
[0119] h(t)=f(Wh(t-1)+Ux(t)+b) Formula (13)
[0120] In formula (13), h(t) is the hidden state at step t, f and W are the weight matrices of the activation function and the recurrent connection, respectively. h(t-1) is the hidden state at the previous step, U is the weight matrix of the input data, xt is the input at step t, and b is the bias vector.
[0121] Optionally, the electronic device can input one or more labeled training samples into the initial model each time to obtain the prediction results of the initial model, calculate the loss function based on the error between the prediction results of the initial model and the label, and adjust the parameters in the initial model according to the loss function until the initial model converges, thus obtaining the trained initial model.
[0122] Optionally, the conditions for the initial model to converge (end training) may include: the number of times the training samples are input into the initial model reaches a preset threshold, or the error value of the initial model is less than a preset error threshold, etc. This application embodiment does not impose such limitations.
[0123] Optionally, after extracting the feature vectors of a preset category from each LLPS protein sequence, the electronic device can also extract a portion of the feature vectors to obtain a validation sample set to validate the trained initial model. If the trained initial model passes the validation, the electronic device can use the trained initial model as the LLPS protein recognition model.
[0124] Optionally, the conditions for the initial model to pass validation may include: the accuracy and / or recall of the initial model meeting a preset threshold.
[0125] Optionally, if the trained initial model fails to pass validation, the electronic device can retrain the initial model.
[0126] Optionally, to evaluate the performance of different types of initial models, electronic devices can also use some performance evaluation metrics, such as sensitivity (Sn), specificity (Sp), accuracy (ACC), and Matthews correlation coefficient (MCC). These evaluation metrics can be specifically expressed as the following formulas (14) to (17):
[0127]
[0128]
[0129]
[0130]
[0131] In the above formulas (14) to (17), TP represents a LLPS protein that is predicted to be RNA-dependent but is actually RNA-dependent, TN represents a LLPS protein that is predicted to be non-RNA-dependent but is actually non-RNA-dependent, FP represents a LLPS protein that is predicted to be RNA-dependent but is actually non-RNA-dependent, and FN represents a LLPS protein that is predicted to be non-RNA-dependent but is actually RNA-dependent.
[0132] For example, the performance of various types of models can be shown in Table 1 below:
[0133] Table 1
[0134]
[0135] As shown in Table 1, taking the three evaluation indicators Sn, Sp, and ACC as examples, the RF model has the highest Sn, ranks third in Sp, and has the highest ACC. Therefore, the overall performance of RF is better than other models.
[0136] In the LLPS protein identification method provided in this application, an electronic device can acquire an initial sample set; the initial sample set includes multiple LLPS protein sequences; a preset type of feature vector is extracted from each LLPS protein sequence; the dependency type corresponding to each LLPS protein sequence is obtained; the dependency type includes RNA-dependent and non-RNA-dependent types; multiple training samples are obtained according to the preset type of feature vector; the dependency type corresponding to each LLPS protein sequence is used as the label of the multiple training samples to obtain a training sample set; a preset initial model is trained based on the training sample set to obtain an LLPS protein identification model; the LLPS protein identification model is used to predict the dependency type corresponding to the LLPS protein sequence. Compared with the current traditional biochemical methods for identifying RNA-dependent LLPS proteins, this application can train a machine learning model capable of identifying RNA-dependent LLPS proteins, reducing the time and cost of identification.
[0137] The foregoing primarily describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the aforementioned functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0138] In an exemplary embodiment, this application provides an LLPS protein recognition device. Figure 5 This is a schematic diagram illustrating the composition of the LLPS protein recognition device provided in an embodiment of this application. Figure 5 As shown, the device may include an acquisition module 501 and a processing module 502.
[0139] The acquisition module 501 is used to acquire an initial sample set, which includes multiple liquid-liquid phase separated LLPS protein sequences. The processing module 502 is used to extract a preset type of feature vector from each LLPS protein sequence. The acquisition module 501 is also used to acquire the dependency type corresponding to each LLPS protein sequence. The dependency type includes RNA-dependent and non-RNA-dependent types. The processing module 502 is also used to obtain multiple training samples based on the preset type of feature vector. The dependency type corresponding to each LLPS protein sequence is used as the label of the multiple training samples to obtain a training sample set. The preset initial model is trained based on the training sample set to obtain an LLPS protein recognition model. The LLPS protein recognition model is used to predict the dependency type corresponding to the LLPS protein sequence.
[0140] Optionally, the preset feature vectors include: amino acid composition ACC feature vector, K-spaced amino acid pair composition CKSAAP feature vector, Geary autocorrelation feature vector, and conjugated triplet CTriad feature vector.
[0141] Optionally, the preset categories include multiple categories; the processing module 502 is specifically used to calculate the mutual information between the feature vector of each category and the target value; sort the feature vectors of multiple categories according to the first order of mutual information to obtain the sorting result; the first order includes ascending or descending order; select the feature vector of the category with the largest mutual information with the target value from the sorting result to complete one selection; repeatedly select the feature vector with the largest mutual information with the target value from the feature vectors of multiple categories until the number of feature vector categories reaches a preset threshold to obtain the feature vector of the target category selected multiple times; use the feature vector of the target category of each LLPS protein sequence as a training sample to obtain multiple training samples.
[0142] Optionally, before calculating the mutual information between the feature vectors of each of the multiple categories and the target value, the processing module 502 is further configured to calculate the correlation coefficient between the feature vectors of the first category and the feature vectors of the second category among the feature vectors of the multiple categories; the first category and the second category are different; if the correlation coefficient between the feature vectors of the first category and the second category is greater than a preset correlation coefficient threshold, then the feature vectors of the first category and the second category are deleted.
[0143] Optionally, the correlation coefficient is determined by the following formula:
[0144]
[0145] Where Cr represents the correlation coefficient; x i This represents the value of the eigenvector of the first category; y represents the average value of the eigenvectors of the first class; i This represents the value of the feature vector of the second type; This represents the average value of the eigenvectors of the second type.
[0146] It should be noted that, Figure 5 The module division shown is illustrative and represents only one logical functional division; in actual implementation, other division methods are possible. For example, two or more functions can be integrated into a single processing module. These integrated modules can be implemented in hardware or as software functional units.
[0147] In an exemplary embodiment, this application also provides a readable storage medium including software instructions that, when run on an electronic device, cause the electronic device to perform any of the methods provided in the above embodiments.
[0148] In an exemplary embodiment, this application also provides a computer program product containing computer execution instructions, which, when run on an electronic device, causes the electronic device to perform any of the methods provided in the above embodiments.
[0149] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer-executable instructions. When these computer-executable instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer-executable instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer-executable instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks, SSDs).
[0150] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple components. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.
[0151] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.
[0152] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for identifying LLPS proteins, characterized in that, The method includes: Obtain an initial sample set; the initial sample set includes multiple liquid-liquid phase separated LLPS protein sequences; Feature vectors of a predetermined type are extracted from each LLPS protein sequence. The feature vectors of the predetermined type include: amino acid composition ACC feature vector, K-spacer amino acid pair composition CKSAAP feature vector, Geary autocorrelation feature vector, and conjugated triplet CTriad feature vector. Obtain the dependency type corresponding to each LLPS protein sequence; the dependency type includes RNA-dependent and non-RNA-dependent types. The preset categories include multiple categories. The correlation coefficient between the feature vectors of the first category and the feature vectors of the second category is calculated among the feature vectors of the multiple categories. The first category and the second category are different. If the correlation coefficient between the feature vectors of the first type and the second type is greater than a preset correlation coefficient threshold, then the feature vectors of the first type and the second type are deleted. Calculate the mutual information between the feature vector and the target value for each of the multiple categories; The feature vectors of the multiple categories are sorted according to a first order of mutual information to obtain a sorting result; the first order includes ascending or descending order. From the sorting results, select the feature vector of the category with the largest mutual information with the target value to complete one selection; The feature vector with the largest mutual information with the target value is selected from the feature vectors of the multiple types multiple times until the number of feature vector types reaches a preset threshold, thus obtaining the feature vector of the target type selected multiple times. The feature vector of each target category of each LLPS protein sequence is used as a training sample to obtain the multiple training samples; The dependency type corresponding to each LLPS protein sequence is used as the label of the multiple training samples to obtain a training sample set; The preset initial model is trained based on the training sample set to obtain the LLPS protein recognition model; the LLPS protein recognition model is used to predict the dependency type corresponding to the LLPS protein sequence.
2. The method according to claim 1, characterized in that, The correlation coefficient is determined by the following formula: ； in, This represents the correlation coefficient; This represents the value of the eigenvector of the first category; This represents the average value of the eigenvectors of the first category; This represents the value of the feature vector of the second type; This represents the average value of the eigenvectors of the second type.
3. The method according to claim 1, characterized in that, The method further includes: Obtain the sequence of the protein to be identified; Extract the feature vector of the target species from the protein sequence; The feature vector of the target type is input into the LLPS protein recognition model to obtain the dependency type corresponding to the protein sequence to be identified.
4. An LLPS protein recognition device, characterized in that, The device includes: an acquisition module and a processing module; The acquisition module is used to acquire an initial sample set; the initial sample set includes multiple liquid-liquid phase separated LLPS protein sequences; The processing module is used to extract feature vectors of a preset type from each LLPS protein sequence. The feature vectors of the preset type include: amino acid composition ACC feature vector, K-spacer amino acid pair composition CKSAAP feature vector, Geary autocorrelation feature vector, and conjugated triplet CTriad feature vector. The acquisition module is also used to acquire the dependency type corresponding to each LLPS protein sequence; the dependency type includes RNA-dependent and non-RNA-dependent types. The preset categories include multiple categories. The processing module is further configured to calculate the correlation coefficient between the feature vectors of the first category and the feature vectors of the second category among the feature vectors of the multiple categories. The first category and the second category are different. If the correlation coefficient between the feature vectors of the first category and the second category is greater than the preset correlation coefficient threshold, the feature vectors of the first category and the second category are deleted. The processing module is specifically used to calculate the mutual information between the feature vector of each of the multiple categories and the target value; sort the feature vectors of the multiple categories according to a first order of mutual information to obtain a sorting result; the first order includes ascending or descending order; select the feature vector of the category with the largest mutual information with the target value from the sorting result to complete one selection; repeatedly select the feature vector with the largest mutual information with the target value from the feature vectors of the multiple categories until the number of feature vector categories reaches a preset threshold to obtain the feature vectors of the target categories selected multiple times; and use the feature vector of the target category of each LLPS protein sequence as a training sample to obtain the multiple training samples. The processing module is further configured to obtain multiple training samples based on the feature vectors of the preset categories; use the dependency type corresponding to each LLPS protein sequence as the label of the multiple training samples to obtain a training sample set; train a preset initial model based on the training sample set to obtain an LLPS protein recognition model; the LLPS protein recognition model is used to predict the dependency type corresponding to the LLPS protein sequence.
5. The apparatus according to claim 4, characterized in that, The correlation coefficient is determined by the following formula: ； in, This represents the correlation coefficient; This represents the value of the eigenvector of the first category; This represents the average value of the eigenvectors of the first category; This represents the value of the feature vector of the second type; This represents the average value of the eigenvectors of the second type.
6. An electronic device, characterized in that, The electronic device includes: a processor and a memory; The memory stores instructions that the processor can execute; When the processor is configured to execute the instructions, the electronic device performs the method as described in any one of claims 1-3.
7. A readable storage medium, characterized in that, The readable storage medium includes: software instructions; When the software instructions are executed in an electronic device, the electronic device causes the electronic device to perform the method as described in any one of claims 1-3.