Method for predicting sensitivity of anti-cancer drug based on depth factorization machine

By combining deep neural networks and factor decomposition machines, a deep factor decomposition machine model was constructed, which solved the problem of insufficient accuracy of drug feature extraction and insufficient connection between multivariate network models, and achieved higher accuracy drug sensitivity prediction.

CN118919096AActive Publication Date: 2024-11-08THE ACAD OF TIANJIN UNIV HEFEI
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202410924385.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-08
Estimated Expiration
2044-07-11

AI Technical Summary

Technical Problem

The prior art lacks accuracy in extracting drug characteristic structures, and the establishment of multivariate network models is insufficient to consider relevant data connections, resulting in insufficient accuracy in drug sensitivity analysis.

Method used

Combining deep neural networks with factor decomposition machines to build a deep factor decomposition machine model, predict the correlation between genes, anti-cancer drugs and cell gene expression, fuse low-order and high-order feature extraction, and improve drug sensitivity prediction accuracy.

Benefits of technology

It improves the accuracy and stability of drug sensitivity prediction, builds a more accurate multivariate network model, considers the complex relationship between drugs, cell lines and genes, and enhances the prediction effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118919096A_ABST
    Figure CN118919096A_ABST
Patent Text Reader

Abstract

The method for predicting the sensitivity of the anti-cancer drug based on the depth factorization machine is characterized by comprising the following steps: step 1, constructing a data set and preprocessing; 2, constructing a deep factorization machine DeepFM model, wherein the main thought of the deep factorization machine DeepFM model is to construct a factorization machine FM model, construct a deep neural network model DNN, and fuse the factorization machine FM and the deep neural network to construct the deep factorization machine DeepFM model; and step 3, training and evaluating model results of each stage, averagely dividing the anti-cancer drug sensitivity data obtained in the step 1 into five parts, executing five-fold cross validation, and training the model results of each stage in the stage. According to the method, the drug low-order feature combination is independently modeled, the high-order feature combination is fused, the model fusing the deep neural network model and the factorization machine model is constructed, the drug feature structure is better extracted, and the constructed model is more accurate and higher in precision.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of deep neural network (DNN) and factorization machine (FM) feature extraction, and in particular to a method for predicting anticancer drug sensitivity based on a deep factorization machine. Background Art

[0002] Cancer refers to a type of disease caused by the uncontrolled proliferation and spread of abnormal cells. These abnormal cells can form tumors and invade surrounding tissues or spread to other parts, causing physical dysfunction or even endangering life. Cancer is closely related to human DNA. It is a disease caused by mutations or abnormal changes in cell DNA. Studies have shown that different types of cancer are usually associated with specific DNA mutations or abnormalities. By analyzing the tumor tissues of cancer patients in terms of genomics, transcriptomics, and epigenomics, we can reveal the mechanism of cancer occurrence and understand the relationship between cancer and DNA. However, in the clinical treatment of cancer in real life, different cancer patients often show different sensitivities to the same anticancer drug. In other words, different individuals also need different treatment plans for the same drug. Therefore, cancer patients should be treated individually, that is, the most targeted and effective anticancer drugs should be selected for them using the DNA sequence of cancer patients, which will greatly promote the development of cancer treatment.

[0003] With the rapid development of artificial intelligence, artificial intelligence has been applied by researchers in the field of anti-cancer. Based on big data and machine learning, artificial intelligence can understand the pathogenesis, case characteristics and drug sensitivity of tumors based on tumor DNA omics, transcriptomics and other data, and take earlier preventive measures and personalized treatment plans for individuals. In addition, artificial intelligence technology can be carried out in a safer environment to ensure the safety of experiments. The knowledge related to machine learning and artificial intelligence technology is of great help to people in data prediction, so that humans can choose a simpler way to standardize and process complex data in the face of extremely large and messy data sets, so that it can be turned into usable data in the prescribed format required for research. At the same time, researchers have also applied data prediction to the field of biology, which has enabled the rapid development of the field of biological information. Some research organizations have established data analysis platforms with it, such as DNA sequence test data of several patients, including CCLE (Cancer Cell Line Encyclopedia), TCGA (Cancer Genome Atlas Project), etc. Many research organizations at home and abroad have explored various genomic databases and studied the pathogenic genes of certain cancers, such as the article "Breast cancer-specific copy number variation leads to DNA repair imbalance" published by Gao Manganese, Cheng Zhujun, and Zhang Cheng. There are also studies on the CCLE database, using baseline gene expression to establish a classification model for predicting drug sensitivity, and introducing algorithms such as random forests and SVM. Among the existing research results, the prediction of drug sensitivity also has a high accuracy rate. For example, in the article "Anticancerdrug sensitivity prediction in cell lines from baseline gene expression through recursive feature selection" published by Zuoli Dong et al., baseline gene expression data was used through recursive feature selection to predict the sensitivity of cell lines to anticancer drugs, and the prediction accuracy can reach 80%. In addition, in the article "Prediction of Anticancer Drug Sensitivity Based on Cell Line-Drug K-Nearest Neighbor" published by Wang Bo, Wei Dong, and Li Yushuang on February 23, 2020, the gene expression and gene mutation characteristics of the cell lines were comprehensively considered, a new definition of cell line similarity was given, and the drug similarity measurement method was combined to propose a "cell line-drug K-nearest neighbor" calculation model. The obtained anticancer drug sensitivity prediction results are significantly better than the classical model.

[0004] FM Factorization Machine is a machine learning model that combines the efficiency of linear models with the capabilities of factorization models. It can process large-scale sparse data and learn the interaction between features. The factorization machine model decomposes the features and represents the embedding of each feature as a low-dimensional vector to capture the interaction effect between features. By considering feature interactions, the factorization machine can better model the nonlinear relationship between features, thereby improving the accuracy of prediction. The factorization machine model can process high-dimensional sparse data, is suitable for large-scale data sets, has high prediction accuracy and efficiency, and can learn the interaction between features, including second-order and higher-order interactions. The deep neural network DNN is a neural network structure composed of multiple hidden layers, with strong learning and representation capabilities, which can effectively capture complex nonlinear relationships in large data sets. In the prediction of cancer drug sensitivity, the deep neural network DNN can be trained using large-scale biomedical data (such as genomic data, transcriptomic data, protein interaction data, etc.) and learn complex drug-gene or drug-disease interaction patterns. Through deep learning, the deep neural network DNN can automatically extract and learn high-level features in the data, thereby more accurately predicting the sensitivity of drugs to cancer.

[0005] The existing technology is not accurate enough in extracting drug feature structures, and does not fully consider the relevant data connections when establishing multivariate network models, which leads to inaccurate drug sensitivity analysis. Therefore, it is necessary to study a method based on deep factor decomposition machine to predict anticancer drug sensitivity. Summary of the invention

[0006] The purpose of the present invention is to combine deep neural networks and factor decomposition machines to better extract feature vectors and improve the accuracy of models for predicting anticancer drug sensitivity in cell lines by considering data such as the correlation between genes, anticancer drugs and cell gene expression.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The method for predicting anticancer drug sensitivity based on a deep factor decomposition machine is characterized by comprising the following steps:

[0009] Step 1: Build data set and preprocess;

[0010] 1-1, obtain gene expression data of cells to build a data set;

[0011] 1-2, obtaining drug chemical structure data, obtaining the chemical structures of all drugs in the data set described in step 1-1 from PubChem;

[0012] 1-3, using ActArea as an indicator of drug sensitivity, standardizing and normalizing the drug sensitivity data to determine the final data set;

[0013] Step 2: Construct a model of a deep factorization machine DeepFM: The main idea of ​​the model is to construct a factorization machine FM model, a deep neural network model, and integrate the factorization machine with the deep neural network to construct a deep factorization machine DeepFM model;

[0014] Step 3: Train and evaluate the model results at each stage. The anticancer drug sensitivity data obtained in step 1 are evenly divided into 5 parts, and a 5-fold cross validation is performed. In this stage, the Pearson correlation coefficient and epoch parameter are introduced for evaluation in order to train the model results at each stage.

[0015] 3-1, train the factorization machine FM result data and analyze it;

[0016] 3-2, train the deep neural network model result data and analyze it;

[0017] 3-3, train the DeepFM model result data and analyze it;

[0018] 3-4, Analysis and comparison of the results of three models.

[0019] Furthermore, in step 1-1, the data set comes from the Cancer Cell Line Encyclopedia (CCLE). This project provides researchers with genomic data and analysis based on multiple cell lines based on the genetic expression characteristics, pharmacological characterization and visualization of a large number of cancer genome composition models. Drug anti-cancer sensitivity data containing more than 11,000 data items of 24 drugs and 504 cell lines were extracted from the CCLE database.

[0020] Furthermore, in step 1-2, the chemical structures of the 24 drugs in the data set of step 1-1 are obtained from PubChem, which is an open database containing a large amount of chemical and physical properties, biological structures, toxicity, drug chemical structures and other data, maintained by the National Center for Biotechnology Information (NCBI) of the National Institutes of Health (NIH).

[0021] Furthermore, in the steps 1-3, the variable ActArea (Activity Area) describing the active area of ​​a certain drug on the cell line is used as an indicator of drug sensitivity, which can describe whether the cell line is sensitive to the drug. The initial data of Actarea is between 0 and 7.8, so that its value is evenly distributed around 0, and the minimum-maximum normalization processing is performed on it using the following formula so that its value is evenly distributed around 0:

[0022]

[0023] Among them, ActArea is the original data, ActArea * is the normalized data, min(ActArea) and max(ActArea) are the minimum and maximum values ​​of the original data respectively;

[0024] If drug-cell lines with ActArea greater than 0.8 are considered as drug-sensitive cell lines, and drug-cell lines with ActArea less than negative 0.8 are considered as drug-resistant cell lines, there are 2215 sensitive drug-cell line pairs, and the number of sensitive and resistant samples is basically the same;

[0025] In the examples of the present invention, 2215 cell lines of CCLE were obtained after screening.

[0026] Furthermore, the specific operation steps of step 2 are as follows:

[0027] 2-1: Construct a factor decomposition machine FM model. This model is mainly used to solve the problem of how to combine features when data is sparse. In order to express the correlation information between various feature quantities, a second-order polynomial model is used here. The final data set obtained in step 1 is used as the input layer. The low-order features are processed by the factor decomposition machine FM and the cross-relationships between the low-order features are extracted.

[0028] 2-2: Build a deep neural network DNN model, use feature conversion to one-hot form to process discrete features, divide the features into different regions to avoid a large number of redundant network parameters, then add two fully connected layers, combine dense vectors, and then extract high-order feature combinations;

[0029] 2-3: Using the fusion mechanism of the parallel structure, the model prediction results constructed in step 2-1 and step 2-2 are fused into the deep factorization machine DeepFM model.

[0030] Furthermore, in step 2-1, a factorization machine FM model is constructed by using a binomial polynomial to express the correlation information between various feature quantities. The second-order polynomial model expression is:

[0031]

[0032] Among them, y represents the output value, x i represents the feature variable, w0 represents the constant term of the model, and represents the input feature variable x i When both are 0, the output value predicted by the model, w i Represents the weight parameter of the model, x ix j Represents the feature variable x i and x j The combined variable, w ij Represents variable x i and x j The weight of the cross part, when i≠j, w ij Represents variable x i and x j The degree of interaction between them.

[0033] Since the factorization machine FM machine is a classification function model, after one-hot encoding, the sample data becomes very sparse, and there are very few cases where two variables x i and x j If it is not 0, it will result in the inability to obtain w through training. ij , in order to find w ij , for each eigenvector x i Introducing auxiliary vector U i =(U i1 , U i2 ,…,U ik ), and similarly for each eigenvector x j Introducing U j =(U j1 , U j2 ,…,U jk ), use U i U j T Solving for w ij .

[0034] w ij The matrix can be expressed as:

[0035]

[0036] in, represents the output value of the matrix, U represents the auxiliary vector of the eigenvector, and U T Indicates transposing the row vector in matrix U into a column vector, u n represents the nth column in the matrix U, represents the nth column vector u in the matrix U n The transpose of , that is, the row vector form.

[0037] Finally, we use stochastic gradient descent (SGD) to train the model:

[0038]

[0039] in, is a parameter vector, φ represents a parameter, y(x) represents the output value, xi represents the characteristic variable, w0 represents the constant term of the model, and w i Represents the weight parameter of the model, U i Represents the feature variable x i The auxiliary vector of the model will change according to the different values ​​of the parameters. The factor decomposition machine FM model framework is used to process low-order features and extract cross-relationships of low-order features.

[0040] Furthermore, in step 2-2, a deep neural network is introduced to extract high-order features. After dividing the input features of the deep neural network DNN into feature areas, two fully connected layers are added to combine dense vectors to extract high-order features.

[0041] Furthermore, in step 2-3, the parallel structure of the fusion mechanism is used to obtain the deep factorization machine DeepFM model.

[0042] The structural framework of the deep factorization machine DeepFM is divided into two parts, one is the structural layer of the factorization machine FM, and the other is the structural layer of the deep neural network. The two share the same feature input. The factorization machine FM and the deep neural network DNN are responsible for the extraction of low-order and high-order features respectively. The prediction result of the deep factorization machine DeepFM can be expressed as:

[0043] Y DeepFM =sigmoid(Y FM +Y DNN ) (4)

[0044] Among them, Y DeepFM is the output of the DeepFM model, Y FM is the output of the factorization machine FM, Y DNN It is the output result of the deep neural network. The outputs of the factorization machine FM part and the deep neural network DNN part are weighted summed, and then the result is mapped to (0,1) through the Sigmoid function.

[0045] Furthermore, step 3 uses the 5-fold cross-validation method to divide the training set and the validation set, and finally uses these data to obtain the sensitivity of anticancer drugs predicted by the DeepFM model. The 5-fold cross-validation is to divide the data into 5 parts, with the training set accounting for 80% and the validation set accounting for 20%. During the entire cross-validation process, the training set and the validation set are replaced in turn, so that each part is used as both a training set and a validation set. The final result is obtained by taking the average of each fold result of the 5-fold cross-validation.

[0046] After 5-fold cross validation, a series of training results were obtained, which were analyzed and evaluated in the following three ways:

[0047] (Ⅰ), epoch-pearson linear analysis diagram of three models: factorization machine FM, deep neural network DNN and deep factorization machine DeepFM;

[0048] (Ⅱ), linear analysis diagram of drug types and their Pearson correlation coefficients;

[0049] (III), epoch-Pearson linear comparison chart analysis and Drug-Pearson bar chart comparison chart analysis of the three models;

[0050] Among them, the horizontal axis of the linear analysis graph of (Ⅰ) is the number of training times of the epoch parameter 35, and the vertical axis is the Pearson correlation coefficient. Epoch means when a complete data set passes through the neural network once and returns once. This process is called an epoch. All training samples have undergone forward propagation and backward propagation in the neural network. An epoch is the process of training all training samples once.

[0051] The Pearson correlation coefficient refers to the product-moment correlation coefficient when two variables are normal continuous variables and there is a linear relationship between the two variables. The Pearson correlation coefficient is used to measure whether two data sets are on the same line and to measure the linear relationship between fixed-distance variables. The larger the Pearson correlation coefficient, the stronger the correlation. The closer the Pearson correlation coefficient is to 1, the stronger the correlation is. The closer the Pearson correlation coefficient is to 0, the weaker the correlation is. The calculation formula is as follows:

[0052]

[0053] Where r represents the Pearson correlation coefficient between two variables (x and y), x i and i are the i-th dimension features of the two nodes x and y respectively, and N is the number of feature dimensions.

[0054] Furthermore, in step 3-4, the epoch-Pearson linear comparison graph and drug-Pearson bar graph of the three models were analyzed. The results show that the anticancer drug sensitivity prediction calculation model constructed by the deep factorization machine DeepFM is more accurate and stable.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. The low-order feature combination of the drug in the present invention is modeled separately, and the high-order feature combination is integrated to construct a model integrating the deep neural network DNN and the factorization machine FM, so as to better extract the drug feature structure and make the constructed model more accurate and more precise.

[0057] 2. The present invention mainly studies the relationship between gene expression data and anticancer drug sensitivity. While establishing a multivariate network model, it considers the relationship between anticancer drugs, the relationship between cell lines, and the relationship between genes and anticancer drug sensitivity, and establishes a model with the highest possible accuracy based on these relationships. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flow chart of the present invention;

[0059] Figure 2 It is a schematic diagram of the full connection of the deep neural network DNN model in the embodiment;

[0060] Figure 3 Schematic diagram of input feature division of a deep neural network DNN model in an embodiment;

[0061] Figure 4 Schematic diagram of the overall structure of the deep neural network DNN model after adding a link layer in the embodiment;

[0062] Figure 5 This is a schematic diagram of a DeepFM model in an embodiment;

[0063] Figure 6 It is the epoch-pearson linear analysis diagram of the factorization machine FM model in the embodiment;

[0064] Figure 7 It is a drug-pearson linear analysis diagram of the factorization machine FM model in the embodiment;

[0065] Figure 8 It is the epoch-pearson linear analysis diagram of the deep neural network DNN model in the embodiment;

[0066] Fig. 9 It is a drug-Pearson linear analysis diagram of the deep neural network DNN model in the embodiment;

[0067] Fig.10 It is the epoch-pearson linear analysis diagram of the deep factorization machine DeepFM model in the embodiment;

[0068] Fig.11 It is a drug-Pearson linear analysis diagram of the deep factorization machine DeepFM model in the embodiment;

[0069] Fig.12 Epoch-Pearson linear comparison analysis diagram of the three models in the embodiment

[0070] Fig.13 It is a comparative analysis chart of drug-Pearson bar graphs of three models in the embodiment.

[0071] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0073] like Figure 1-6 As shown, the present application provides a method for predicting anticancer drug sensitivity based on a deep factor decomposition machine, comprising the following steps:

[0074] Step 1: Build and preprocess the dataset

[0075] Step 1-1 Obtain the gene expression dataset of cells: The dataset comes from the Cancer Cell Line Encyclopedia (CCLE). This project is based on the genetic expression characteristics, pharmacological characterization and visualization of a large number of cancer gene composition models, and provides researchers with genomic data and analysis based on multiple cell lines. It was downloaded from the website containing cell gene data (https: / / portals.broadinstitute.org / ccle / data), which contains drug anticancer sensitivity data of 24 drugs and 504 cell lines with a total of more than 11,000 data.

[0076] Step 1-2: Obtain drug chemical structure data: Obtain the chemical structures of the 24 drugs in the data set described in step 1-1 from PubChem.

[0077] PubChem is an open database containing a large amount of chemical and physical properties, biological structures, toxicity, drug chemical structures and other data, maintained by the National Center for Biotechnology Information (NCBI) of the National Institutes of Health (NIH). The chemical structures of the 24 drugs in step 1-1 were obtained in sdf (Structure Data File) format by downloading the two-dimensional structural information files of the relevant drug names from the PubChem website (https: / / pubchem.ncbi.nlm.nih.gov / ), and then the PaDEL-Descriptor software (an open source chemical information processing software designed to calculate the molecular descriptors of compounds, which can be downloaded from http: / / padel-descriptor.winsite.com / ) was used to convert the sdf file into the corresponding fingerprint file (a file format representing the molecular structure or characteristics).

[0078] Step 1-3 normalizes the data set obtained in step 1-1 to determine the final data set: the variable ActArea (Activity Area) describing the active area of ​​a certain drug on the cell line is used as an indicator of drug sensitivity, which can describe whether the cell line is sensitive to the drug. The initial data of Actarea is between 0 and 7.8, and the following formula is used to perform minimum-maximum normalization processing on it so that its value is evenly distributed around 0:

[0079]

[0080] Among them, ActArea is the original data, ActArea * is the normalized data, min(ActArea) and max(ActArea) are the minimum and maximum values ​​of the original data respectively;

[0081] Considering drug-cell lines with ActArea greater than 0.8 as drug-sensitive cell lines, and ActArea less than -0.8 as drug-resistant cell lines, there are 2215 sensitive drug-cell line pairs, and the number of sensitive and resistant samples is basically the same.

[0082] In the examples of the present invention, 2215 cell lines of CCLE were obtained after screening.

[0083] Step 2, construct a model of a deep factor decomposition machine DeepFM. The main idea of ​​the model is to construct a factor decomposition machine model, a deep neural network model, and integrate the factor decomposition machine and the deep neural network to construct a deep factor decomposition machine model.

[0084] Step 2-1, by using binomial polynomial to construct factorization machine FM model, and then express the correlation information between various feature quantities, the expression of this second-order polynomial model is:

[0085]

[0086] Among them, y represents the output value, w0 represents the constant term of the model, which means that when the input feature variable x i When both are 0, the output value predicted by the model, w i Represents the weight parameter of the model, x i x j Represents the feature variable x i and x j The combined variable, w ij Represents variable x i and x j The weight of the cross part, when i≠j, w ij Represents variable x i and x j The degree of interaction between them.

[0087] Since the factorization machine FM is a classification function model, after one-hot encoding, the sample data becomes very sparse, and there are very few cases where two variables x i and x j If it is not 0, it will result in the inability to obtain w through training. ij , in order to find w ij , for each eigenvector x i Introducing auxiliary vector U i =(U i1 , U i2 ,…,U ik ), and similarly for each eigenvector x j Introducing U j =(U j1 , U j2 ,…,U jk ), use U i U j T Solving for w ij .

[0088] w ij The matrix can be expressed as:

[0089]

[0090] in, represents the output value of the matrix, U represents the auxiliary vector of the eigenvector, and U T Indicates transposing the row vector in matrix U into a column vector, u nrepresents the nth column in the matrix U, represents the nth column vector u in the matrix U n The transpose of , that is, the row vector form.

[0091] Finally, we use stochastic gradient descent (SGD) to train the model:

[0092]

[0093] in, is a parameter vector, represents a parameter, y(x) represents the output value, x i represents the characteristic variable, w0 represents the constant term of the model, and w i Represents the weight parameter of the model, U i Represents the feature variable x i The auxiliary vector of the model will change according to the different values ​​of the parameters. The low-order features are processed and the cross-relationships of low-order features are extracted through the factorization machine FM model framework.

[0094] Step 2-2, build a deep neural network DNN model: Since the factor decomposition model built in step 2-1 cannot extract the high-order features of the drug, a deep neural network is introduced to extract high-order features. The full-link structure diagram of the deep neural network DNN is shown in the attached figure. Figure 2 As shown in the attached Figure 2 It can be seen that the number of full-link parameters is too large, which makes the model prone to overfitting. Therefore, in this experiment, the deep neural network DNN input feature is divided into regions, and the structure diagram is shown in the attached figure. Figure 3 shown.

[0095] After dividing the feature regions, two fully connected layers are added to combine dense vectors to extract high-order features, such as Figure 4 As shown. Figure 4 It can be seen that although the input variables are divided into feature regions and two fully connected layers are added, the low-order and high-order feature combinations are still implicitly reflected in the hidden layer. In order to better extract the characteristics of high-order features and low-order features, the deep factor decomposition machine DeepFM is introduced.

[0096] Step 2-3 uses the fusion mechanism of the parallel structure to fuse the prediction results of the two models constructed in step 2-1 and step 2-3 to obtain the DeepFM model. The model principle is as shown in the attached figure. Figure 5 shown.

[0097] The structural framework of the deep factorization machine DeepFM is divided into two parts, one is the structural layer of the factorization machine FM, and the other is the structural layer of the deep neural network. The two share the same feature input. The factorization machine FM and the deep neural network DNN are responsible for the extraction of low-order and high-order features respectively. The prediction result of the deep factorization machine DeepFM can be expressed as:

[0098] Y DeepFM =sigmoid(Y FM +Y DNN ) (4)

[0099] Among them, Y DeepFM is the output of the DeepFM model, Y FM is the output of the factorization machine FM, Y DNN It is the output result of the deep neural network. The outputs of the decomposition machine FM part and the deep neural network DNN part are weighted summed, and then the result is mapped to (0,1) through the Sigmoid function.

[0100] Step 3: Train the model results at each stage and analyze and evaluate

[0101] In this invention, a 5-fold cross-validation method is used to divide the training set and the validation set, and finally these data are used to obtain the sensitivity of anticancer drugs predicted by the deep factor decomposition machine DeepFM model. The 5-fold cross-validation is to divide the data into 5 parts equally, with the training set accounting for 80% and the validation set accounting for 20%. During the entire process of cross-validation, the training set and the validation set are replaced in turn, so that each part is used as both a training set and a validation set. The final result is obtained by taking the average of each fold result of the 5-fold cross-validation.

[0102] After 5-fold cross validation, a series of training results were obtained, which were analyzed and evaluated in the following three ways:

[0103] (I) Epoch-Pearson linear analysis of the three models: factorization machine FM, deep neural network DNN and deep factorization machine DeepFM

[0104] (II) Linear analysis of drug types and their Pearson correlation coefficients

[0105] (III) Epoch-Pearson linear comparison chart analysis and Drug-Pearson bar chart comparison chart analysis of the three models

[0106] Among them, the horizontal axis of the linear analysis graph of (Ⅰ) is the number of training times of the epoch parameter 35, and the vertical axis is the Pearson correlation coefficient. Epoch means when a complete data set passes through the neural network once and returns once, this process is called an epoch. All training samples have undergone a forward propagation and a backward propagation in the neural network. An epoch is the process of training all training samples once.

[0107] The Pearson correlation coefficient refers to the product-moment correlation coefficient when two variables are both normal continuous variables and there is a linear relationship between the two variables. The Pearson correlation coefficient is used to measure whether two data sets are on the same line and to measure the linear relationship between interval variables. The larger the Pearson correlation coefficient, the stronger the correlation. The closer the Pearson correlation coefficient is to 1, the stronger the correlation is, and the closer the Pearson correlation coefficient is to 0, the weaker the correlation is. The calculation formula is as follows:

[0108]

[0109] Where r represents the Pearson correlation coefficient between two variables (x and y), x i and i are the i-th dimension features of the two nodes x and y respectively, and N is the number of feature dimensions.

[0110] Step 3-1 Train the factorization machine FM model result data and analyze it:

[0111] 3-1.1、Epoch-Pearson linear analysis diagram is as attached Figure 6 shown.

[0112] By the attached Figure 6 It can be seen that the overall Pearson correlation coefficients are all above 0.8, among which the relatively high Pearson correlation coefficient can reach 0.9, and most of the Pearson correlation coefficients fluctuate around 0.88, indicating that the predicted value and the true value of the 5-fold data results of the factor decomposition machine FM model are still highly correlated, indicating that the accuracy of the model is relatively reliable. Although the Pearson correlation coefficient values ​​of the 34th group of experimental data results compared with the same group showed a sharp drop, it was because of certain data errors in the experimental data entry or certain randomness of the experiment. The data deviation in a very small range does not affect the overall presentation of the experimental data results.

[0113] 3-1.2. The linear analysis diagram of drug types and their Pearson correlation coefficients is shown in the attached figure. Figure 7 shown.

[0114] By the attached Figure 7It can be seen that the overall Pearson correlation coefficient of each drug is above 0.3, most of them can even reach above 0.7, and basically fluctuate around 0.6. The Pearson correlation coefficients of 4 drugs are even close to 0.9, and the Pearson correlation coefficients of 19 drugs are stable between 0.5 and 0.8. However, the Pearson correlation coefficient of one drug (LBW242) is lower than that of other drugs in the same group (less than 0.4). The prediction result for this drug is not very ideal, and the accuracy of this anticancer drug sensitivity prediction model needs to be further improved.

[0115] Step 3-2: Train the deep neural network model and analyze the result data:

[0116] 3-2.1、Epoch-Pearson linear analysis diagram is as attached Figure 8 shown.

[0117] By the attached Figure 8 It can be seen that the overall Pearson correlation coefficients of both the training set and the validation set are above 0.86. The relatively high Pearson correlation coefficient of the training set can reach 0.98, and the relatively high Pearson correlation coefficient of the validation set can reach 0.92. Most of the Pearson correlation coefficients of the training set fluctuate around 0.94, and most of the Pearson correlation coefficients of the validation set fluctuate around 0.9, indicating that the predicted value and the true value of the 5-fold data results of the deep neural network DNN model are still highly correlated, and the accuracy of the model is relatively reliable.

[0118] 3-2.2. The linear analysis diagram of drug types and their Pearson correlation coefficients is shown in the attached figure. Fig. 9 shown.

[0119] By the attached Fig. 9 It can be seen that the overall Pearson correlation coefficient of each drug is above 0.3, most of them can even reach above 0.7, and basically fluctuate around 0.6. The Pearson correlation coefficients of 4 drugs are even close to 0.9, and the Pearson correlation coefficients of 19 drugs are stable between 0.5 and 0.8. However, the Pearson correlation coefficient of one drug (LBW242) is lower than that of other drugs in the same group (less than 0.4). The prediction result for this drug is not very ideal, and the accuracy of this anticancer drug sensitivity prediction model needs to be further improved.

[0120] Step 3-3 Train the DeepFM model result data and analyze it:

[0121] 3-3.1、Epoch-Pearson linear analysis diagram is as attached Fig.10 shown.

[0122] By the attached Fig.10 It can be seen that the overall Pearson correlation coefficients of both the training set and the validation set are above 0.8. The relatively high Pearson correlation coefficient of the training set can reach 0.95, and the relatively high Pearson correlation coefficient of the validation set can reach 0.92. Most of the Pearson correlation coefficients of the training set show an upward trend and fluctuate around 0.92, and most of the Pearson correlation coefficients of the validation set fluctuate around 0.9 and show an upward trend, indicating that the predicted value and the true value of the 5-fold data results of the deep factor decomposition machine DeepFM model are still highly correlated. Although the experimental data results of around the 16th group show a small unstable fluctuation compared with the Pearson correlation coefficient values ​​of the same group, the overall trend of the experimental results is not affected.

[0123] 3-3.2. The linear analysis diagram of drug types and their Pearson correlation coefficients is shown in the attached figure. Fig.11 shown.

[0124] By the attached Fig.11 It can be seen that the overall Pearson correlation coefficient of each drug is above 0.3, and most of them can even reach above 0.7, and basically fluctuate around 0.7. The Pearson correlation coefficients of 4 drugs are even close to 0.9, and the Pearson correlation coefficients of 19 drugs are stable between 0.5 and 0.8. However, the Pearson correlation coefficient of one drug (LBW242) is lower than that of other drugs in the same group (0.3). The prediction result for this drug is not very ideal, and the accuracy of this anticancer drug sensitivity prediction model needs to be further improved.

[0125] Step 3-4: Analyze the result data of the three models:

[0126] 3-4.1. The factorization machine FM model and the deep neural network DNN model are combined through a parallel structure to form the deep factorization machine DeepFM model. The epoch-pearson linear comparison diagram of the three models is shown in the attached figure. Fig.12 As shown in the attached Fig.12 It can be seen that the results obtained by the deep factor decomposition machine DeepFM model have higher accuracy and reduce the "jump fluctuations" of individual data compared with the results obtained by the factor decomposition machine FM model, while the results obtained by the deep factor decomposition machine DeepFM model reduce the gap in the Pearson correlation coefficient between the training set and the validation set compared with the results obtained by the deep neural network DNN model, which improves the accuracy and stability of the prediction to a certain extent.

[0127] 3-4.2. The linear relationship between the predicted value and the true value of the drug obtained by the factorization machine FM, deep neural network DNN and deep factorization machine DeepFM model is divided according to each drug as shown in the attached figure. Fig.13 As shown. Fig.13 It can be seen that in the overall data, the linear curve of the Pearson correlation coefficient obtained by the deep factor decomposition machine DeepFM model is basically above the Y-axis value of each drug in the factor decomposition machine FM and the deep neural network DNN model. That is to say, the absolute value of the correlation between the predicted value and the true value of almost every drug in the deep factor decomposition machine DeepFM model is greater than the absolute value of the correlation obtained by the factor decomposition machine FM and the deep neural network DNN model. The anticancer drug sensitivity prediction calculation model constructed by the deep factor decomposition machine DeepFM is more accurate and stable.

[0128] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0129] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting anticancer drug sensitivity based on a deep factorization machine, characterized in that: The following steps are involved: Step 1: Build data set and preprocess; 1-1, obtain gene expression data of cells to build a data set; 1-2, obtaining drug chemical structure data, obtaining the chemical structures of all drugs in the data set described in step 1-1 from PubChem; 1-3, using ActArea as an indicator of drug sensitivity, standardizing and normalizing the drug sensitivity data to determine the final data set; Step 2: Construct a model of a deep factorization machine DeepFM: The main idea of ​​the deep factorization machine DeepFM is to construct a factorization machine FM model, a deep neural network DNN model, and integrate the factorization machine FM with the deep neural network DNN to construct a deep factorization machine DeepFM model; Step 3: Train and evaluate the model results at each stage. The anticancer drug sensitivity data obtained in step 1 are evenly divided into 5 parts, and a 5-fold cross validation is performed. In this stage, the Pearson correlation coefficient and epoch parameter are introduced for evaluation in order to train the model results at each stage. 3-1, train the factorization machine FM model result data and analyze it; 3-2, train the deep neural network model result data and analyze it; 3-3, train the DeepFM model result data and analyze it; 3-4, Analysis and comparison of the results of three models.

2. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 1, characterized in that: In step 1-1, the data set comes from the Cancer Cell Line Encyclopedia CCLE. This project provides researchers with genomic data and analysis based on multiple cell lines based on the genetic expression characteristics, pharmacological characterization and visualization of a large number of cancer genome composition models. Drug anticancer sensitivity data containing more than 11,000 data items of 24 drugs and 504 cell lines were extracted from the CCLE database.

3. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 2, characterized in that: In step 1-2, the chemical structures of the 24 drugs in the data set of step 1-1 are obtained from PubChem, which is an open database containing a large amount of chemical and physical properties, biological structures, toxicity, drug chemical structures and other data, and is maintained by the National Center for Biotechnology Information (NCBI) of the National Institutes of Health (NIH).

4. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 3, characterized in that: In the steps 1-3, the variable ActArea describing the active area of ​​a certain drug on the cell line is used as an indicator of drug sensitivity. It can describe whether the cell line is sensitive to the drug. The initial data of Actarea is between 0 and 7.

8. The following formula is used to perform minimum-maximum normalization processing on it so that its value is evenly distributed around 0: Among them, ActArea is the original data, ActArea * is the normalized data, min(ActArea) and max(ActArea) are the minimum and maximum values ​​of the original data respectively; If drug-cell lines with ActArea greater than 0.8 are considered as drug-sensitive cell lines, and drug-cell lines with ActArea less than negative 0.8 are considered as drug-resistant cell lines, there are 2215 sensitive drug-cell line pairs, and the number of sensitive and resistant samples is basically the same; In the example of the present invention, 2215 cell lines of CCLE were obtained after screening.

5. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 1, characterized in that: The specific operation steps of step 2 are as follows: 2-1: Construct a factorization machine FM model. This model is mainly used to solve the problem of how to combine features when data is sparse. In order to express the correlation information between various feature quantities, a second-order polynomial model is used here. The final data set obtained in step 1 is used as the input layer. The low-order features are processed through the deep neural network DNN model and the cross-relationships between low-order features are extracted. 2-2: Build a deep neural network DNN model, use feature conversion to one-hot form to process discrete features, divide the features into different regions to avoid a large number of redundant network parameters, then add two fully connected layers, combine dense vectors, and then extract high-order feature combinations; 2-3: Using the fusion mechanism of the parallel structure, the model prediction results constructed in step 2-1 and step 2-2 are fused into the deep factorization machine DeepFM model.

6. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 5, characterized in that: In step 2-1, a binomial polynomial is used to construct a factorization machine FM model, thereby expressing the correlation information between various feature quantities. The second-order polynomial model expression is: Among them, y represents the output value, x i represents the characteristic variable, w0 represents the constant term of the model, which means that when the input characteristic variable x i When both are 0, the output value predicted by the model, w i Represents the weight parameter of the model, x i x j Represents the feature variable x i and x j The combined variable, w ij Represents variable x i and x j The weight of the cross part, when i≠j, w ij Represents variable x i and x j The degree of influence of the interaction between them; Since the factorization machine FM is a classification function model, after one-hot encoding, the sample data becomes very sparse, and there are very few cases where two variables x i and x j If it is not 0, it will result in the inability to obtain w through training. ij , in order to find w ij , for each eigenvector x i Introducing auxiliary vector U i =(U i1 , U i2 ,…,U ik ), and similarly for each eigenvector x j Introducing U j =(U j1 , U j2 ,…,U jk ), use U i U j T Solving for w ij ; w ij The matrix can be expressed as: in, represents the output value of the matrix, U represents the auxiliary vector of the eigenvector, and U T Indicates transposing the row vector in matrix U into a column vector, u n represents the nth column in the matrix U, represents the nth column vector u in the matrix U n The transpose of , i.e. the row vector form; Finally, we use stochastic gradient descent (SGD) to train the model: in, is a parameter vector, φ represents a parameter, y(x) represents the output value, x i represents the characteristic variable, w0 represents the constant term of the model, and w i Represents the weight parameter of the model, U i Represents the feature variable x i Auxiliary vector; Depending on the different values ​​of the parameters, the final output value of the model will also change. The factor decomposition machine FM model framework is used to process low-order features and extract low-order feature cross-relationships.

7. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 5, characterized in that: In step 2-2, a deep neural network is introduced, and after the input features of the deep neural network DNN are divided into feature areas, two fully connected layers are added to extract high-order features.

8. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 5, characterized in that: In step 2-3, the parallel structure of the fusion mechanism is used to obtain the deep factorization machine DeepFM model; The structural framework of the deep factorization machine DeepFM model is divided into two parts, one is the structural layer of the factorization machine FM, and the other is the structural layer of the deep neural network. The two share the same feature input. The factorization machine model and the deep neural network DNN model are responsible for the extraction of low-order and high-order features respectively. The prediction result of the deep factorization machine DeepFM can be expressed as: AND DeepFM =sigmoid(Y FM +Y DNN ) (5) Among them, Y DeepFM is the output of the DeepFM model, Y FM is the output of the factorization machine FM, Y DNN It is the output result of the deep neural network. The outputs of the factorization machine FM part and the deep neural network DNN part are weighted summed, and then the result is mapped to (0,1) through the Sigmoid function.

9. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 1, characterized in that: Step 3 uses the 5-fold cross-validation method to divide the training set and the validation set, and finally uses these data to obtain the sensitivity of anticancer drugs predicted by the deep factorization machine DeepFM model. The 5-fold cross-validation is to divide the data into 5 parts, with the training set accounting for 80% and the validation set accounting for 20%. During the whole process of cross-validation, the training set and the validation set are replaced in turn, so that each part is used as both a training set and a validation set. The final result is obtained by taking the average of each fold result of the 5-fold cross-validation. After 5-fold cross validation, a series of training results were obtained, which were analyzed and evaluated in the following three ways: (Ⅰ), epoch-pearson linear analysis diagram of three models: factorization machine FM, deep neural network DNN and deep factorization machine DeepFM; (Ⅱ), linear analysis diagram of drug types and their Pearson correlation coefficients; (III), epoch-Pearson linear comparison chart analysis and Drug-Pearson bar chart comparison chart analysis of the three models; Among them, the horizontal axis of the linear analysis graph of (Ⅰ) is the number of training times of the epoch parameter 35, and the vertical axis is the Pearson correlation coefficient. Epoch means when a complete data set passes through the neural network once and returns once. This process is called an epoch. All training samples have undergone a forward propagation and a backward propagation in the neural network. An epoch is the process of training all training samples once. The Pearson correlation coefficient refers to the product-moment correlation coefficient when two variables are normal continuous variables and there is a linear relationship between the two variables. The Pearson correlation coefficient is used to measure whether two data sets are on the same line and to measure the linear relationship between fixed-distance variables. The larger the Pearson correlation coefficient, the stronger the correlation. The closer the Pearson correlation coefficient is to 1, the stronger the correlation is. The closer the Pearson correlation coefficient is to 0, the weaker the correlation is. The calculation formula is as follows: Where r represents the Pearson correlation coefficient between two variables (x and y), x i and i are the i-th dimension features of the two nodes x and y respectively, and N is the number of feature dimensions.

10. The method for predicting anticancer drug sensitivity based on deep factor decomposition machine according to claim 1, characterized in that: In step 3-4, the epoch-Pearson linear comparison graph and drug-Pearson bar graph of the three models are analyzed. The results show that the anticancer drug sensitivity prediction calculation model constructed by the deep factor decomposition machine DeepFM is more accurate and stable.

Citation Information

Patent Citations

  • Food-borne disease pathogenic factor prediction method and system based on BP neural network

    CN112216399A

  • Drug sensitivity prediction method based on depth factorization machine

    CN112435754A

  • Medical recommendation method and system based on multi-modal cardiovascular disease information and medium

    CN113010783A

  • Drug recommendation method based on twin neural network and deep factorization machine

    CN113889217A

  • Methylation data signatures of aging and methods of determining a methylation aging clock

    US20220005552A1