A method and device for processing a charge carrier mobility prediction model

Abstract

The embodiment of the present invention relates to a method and device for processing a charge carrier mobility prediction model, the method comprising: constructing a prediction model for predicting charge carrier mobility based on two input molecular sequences, recorded as a corresponding first prediction model; training the first prediction model based on a preset first data set; after the model training is completed, receiving a list of molecular pair sequences input by a user; and performing batch prediction processing on the list of molecular pair sequences based on the first prediction model to obtain a corresponding batch prediction report and feedback to the user. Applying the first prediction model provided by the present invention to the design process of organic semiconductor material molecules can shorten the design cycle and improve the design efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and device for processing a charge carrier mobility prediction model. Background Art

[0002] In the field of (opto)electronic applications, organic semiconductor materials have broad application prospects. At present, the conventional design steps of organic semiconductor materials are simply divided into the following steps: a) First, multiple new material molecular structures are constructed according to the artificial design method or model generation method, and a property index for verification (such as charge carrier mobility index) is given in advance; b) Then, based on experimental means or molecular motion simulation + quantum chemical calculation means, the dimer properties (such as charge carrier mobility) composed of each type of new material molecules are measured or calculated to obtain the corresponding measured property parameters; c) Finally, the measured property parameters of various new material molecules are compared based on the preset property indicators, and the molecular structure is screened according to the comparison results. It can be seen from practice that step b in the above design process is very time-consuming, which makes the time period of the entire design process longer and the design efficiency lower. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, electronic device and computer-readable storage medium for processing a charge carrier mobility prediction model in view of the defects of the prior art. The present invention constructs an end-to-end prediction model for predicting charge carrier mobility based on two input molecular sequences, recorded as a first prediction model, and trains the model based on a preset first data set; and after the model training is completed, the first prediction model is used to perform batch predictions on the list of molecular pairs input by the user. If the first prediction model provided by the present invention is applied to the design process of organic semiconductor material molecules, it is bound to achieve the purpose of shortening the design cycle and improving the design efficiency.

[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present invention provides a method for processing a charge carrier mobility prediction model, the method comprising:

[0005] Constructing a prediction model for predicting charge carrier mobility according to two input molecular sequences, which is recorded as a corresponding first prediction model; the first prediction model predicts the charge carrier mobility between molecules according to the input molecular sequences A and B and outputs a corresponding mobility prediction value C;

[0006] Performing model training on the first prediction model based on a preset first data set;

[0007] After the model training is completed, a list of molecule pair sequences input by a user is received; and batch prediction processing is performed on the list of molecule pair sequences based on the first prediction model to obtain a corresponding batch prediction report and feed it back to the user.

[0008] Preferably, the model input end of the first prediction model is used to receive the corresponding molecular sequences A, B, and the model output end is used to output the corresponding mobility prediction value C;

[0009] The molecular sequence A and the molecular sequence B of the molecular sequences A and B are each a SMILES sequence of one molecule;

[0010] The first prediction model includes a preprocessing module, a molecular conformation encoding module, an electronic coupling prediction model, a reorganization energy prediction model, a vector splicing module and a mobility prediction model;

[0011] The input end of the preprocessing module is connected to the input end of the model, and the first and second output ends are respectively connected to the input ends of the molecular conformation encoding module and the reorganization energy prediction model; the output end of the molecular conformation encoding module is connected to the input end of the electron coupling prediction model; the output ends of the electron coupling prediction model and the reorganization energy prediction model are respectively connected to the first and second input ends of the vector splicing module; the output end of the vector splicing module is connected to the input end of the mobility prediction model; the output end of the mobility prediction model is connected to the output end of the model;

[0012] The preprocessing module is used to input the input molecular sequence A and the molecular sequence B into the first processing interface provided by the preset chemical informatics tool respectively to perform molecular sequence standardization processing to obtain the corresponding first and second standard sequences; and input the first and second standard sequences into the second processing interface provided by the chemical informatics tool respectively to perform molecular sequence two-dimensional topological map conversion processing to obtain the corresponding first and second conversion topological maps; and input the first and second standard sequences into the third processing interface provided by the chemical informatics tool respectively to perform molecular sequence three-dimensional conformation conversion processing to obtain the corresponding first and second conversion conformations; and perform conformation merging on the first and second conversion conformations to obtain the corresponding dimer conformation X1; and perform topological map merging on the first and second conversion topological maps to obtain the corresponding dimer topological map X2; and send the dimer conformation X1 to the molecular conformation encoding module; and send the dimer topological map X2 to the reorganization energy prediction model;

[0013] The chemical informatics tools include at least OpenBabel and RDKit;

[0014] The dimer conformation X1 is composed of a first atomic vector set and a first chemical bond vector set; the first atomic vector set is composed of multiple first atomic vectors; the first atomic vector includes a first atomic identifier, a first molecule identifier, a first atomic type and a first atomic coordinate; the first atomic identifier is a unique identifier of the corresponding atom; the first molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the first atomic type is the element type of the corresponding atom; the first atomic coordinate is the three-dimensional coordinate of the corresponding atom; the first chemical bond vector set is composed of multiple first chemical bond vectors; the first chemical bond vector includes a first bond identifier, a first bond length, a first bond type and a first atomic identifier group; the first bond identifier is a unique identifier of the corresponding chemical bond; the first bond length is the length of the corresponding chemical bond; the first bond type is the chemical bond type of the corresponding chemical bond; the first atomic identifier group is composed of the atomic identifiers of the head and tail atoms of the corresponding chemical bond;

[0015] The dimer topology map X2 is composed of a second atomic vector set and a second chemical bond vector set; the second atomic vector set is composed of multiple second atomic vectors; the second atomic vector includes a second atomic identifier, a second molecule identifier, a second atomic type and a second atomic coordinate; the second atomic identifier is a unique identifier of the corresponding atom; the second molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the second atomic type is the element type of the corresponding atom; the second atomic coordinate is the two-dimensional coordinate of the corresponding atom; the second chemical bond vector set is composed of multiple second chemical bond vectors; the second chemical bond vector includes a second bond identifier, a second bond length, a second bond type and a second atomic identifier group; the second bond identifier is a unique identifier of the corresponding chemical bond; the second bond length is the length of the corresponding chemical bond; the second bond type is the chemical bond type of the corresponding chemical bond; the second atomic identifier group is composed of the atomic identifiers of the head and tail atoms of the corresponding chemical bond;

[0016] The molecular conformation encoding module is composed of a preset first number N E The Uni-Mol models are sequentially connected; each Uni-Mol model is used to perform conformation optimization processing on the currently input three-dimensional conformation and output the corresponding optimized conformation; each Uni-Mol model has completed model training in advance; the molecular conformation encoding module is used to use N E The Uni-Mol model optimizes the dimer conformation X1 received this time step by step and uses the optimized conformation output by the last Uni-Mol model as the corresponding coded output conformation X E Sending to the electronic coupling prediction model;

[0017] The encoding outputs the conformation X EIt is composed of a third atomic vector set and a third chemical bond vector set; the third atomic vector set is composed of multiple third atomic vectors; the third atomic vector includes a third atomic identifier, a third molecule identifier, a third atomic type and a third atomic coordinate; the third atomic identifier is a unique identifier of the corresponding atom; the third molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the third atomic type is the element type of the corresponding atom; the third atomic coordinate is the three-dimensional coordinate of the corresponding atom; the third chemical bond vector set is composed of multiple third chemical bond vectors; the third chemical bond vector includes a third bond identifier, a third bond length, a third bond type and a third atomic identifier group; the third bond identifier is a unique identifier of the corresponding chemical bond; the third bond length is the length of the corresponding chemical bond; the third bond type is the chemical bond type of the corresponding chemical bond; the third atomic identifier group is composed of the atomic identifiers of the bond head and bond tail atoms of the corresponding chemical bond;

[0018] The electronic coupling prediction model is formed by sequentially connecting a 3D-MPNN model, a first fully connected network and a feature fusion module; the electronic coupling prediction model is used to convert the encoded output conformation X E Input the 3D-MPNN model to perform feature extraction and obtain the corresponding feature vector Y 11 ; and the feature vector Y 11 Input the first fully connected network for regression calculation to obtain the corresponding prediction vector Y 12 ; and the feature vector Y 11 and the prediction vector Y 12 Input the feature fusion module to perform feature fusion processing to obtain a corresponding first vector Y1; and send the first vector Y1 to the vector splicing module;

[0019] The 3D-MPNN model is implemented based on the SphereNet model; the 3D-MPNN model is used to output the encoding conformation X of the input E The three-dimensional spherical features, two-dimensional plane angle features and one-dimensional radial features between two atoms of each group of atomic pairs between two molecules are extracted to obtain the corresponding atomic pair space feature vector; and each of the atomic pair space feature vectors and its corresponding two third atomic types are vector-joined and the obtained joint vector is used as the corresponding atomic pair feature vector; and all the obtained atomic pair feature vectors are sequentially joined to form the corresponding feature vector Y 11 ; Wherein, each group of atomic pairs between the two molecules is composed of two atoms belonging to different molecules;

[0020] The first fully connected network is formed by sequentially connecting two or more fully connected layers; the first fully connected network is used to input the feature vector Y 11The corresponding atom pair electron coupling parameters are regressed and calculated for each of the atom pair feature vectors to obtain the corresponding atom pair electron coupling parameters; and the corresponding prediction vector Y is formed by sequentially sorting all the obtained atom pair electron coupling parameters. 12 ;

[0021] The feature fusion module is used to align the feature vector Y according to the atomic pair alignment method. 11 and the prediction vector Y 12 The atomic pair feature vectors and the atomic pair electron coupling parameters corresponding to the same group of atomic pairs are extracted to form a corresponding fused feature vector; and all the obtained fused feature vectors are sequentially sorted to form the corresponding first vector Y1;

[0022] The reorganization energy prediction model is formed by sequentially connecting the 2D-MPNN model and the second fully connected network; the reorganization energy prediction model is used to input the dimer topology X2 into the 2D-MPNN model for feature extraction to obtain the corresponding feature vector Y 21 ; and the feature vector Y 21 Input the second fully connected network to perform regression calculation to obtain a corresponding second vector Y2; and send the second vector Y2 to the vector splicing module;

[0023] The 2D-MPNN model is implemented based on a type of GNN model; the 2D-MPNN model is used to extract atomic features from the input dimer topology X2 to obtain a corresponding atomic feature vector; and all the obtained atomic feature vectors are sequentially spliced ​​to form the corresponding feature vector Y 21 ;

[0024] The second fully connected network is formed by sequentially connecting two or more fully connected layers; the second fully connected network is used to input the feature vector Y 21 Performing corresponding dimer reorganization energy parameter regression calculation to obtain corresponding dimer reorganization energy parameters; and forming the corresponding second vector Y1 from the obtained dimer reorganization energy parameters;

[0025] The vector splicing module is used to perform vector splicing processing on the first vector Y1 and the second vector Y2 to obtain a corresponding splicing vector Z and send it to the mobility prediction model;

[0026] The mobility prediction model is formed by sequentially connecting two or more fully connected layers; the mobility prediction model is used to perform corresponding charge carrier mobility prediction processing according to the input splicing vector Z to obtain the corresponding mobility prediction value C and output it.

[0027] Preferably, the first data set includes a plurality of first data records; the first data record includes a first training molecule sequence, a second training molecule sequence, a first electronic coupling tag V aver,T , the second electronic coupling tag V sum,T , the first reorganization energy label λ T and the first mobility label μ T The first and second training molecular sequences are each a SMILES sequence of one molecule; the first electronic coupling label V aver,T is the average value of the electronic coupling parameters of each heterogeneous atomic pair in the dimer corresponding to the first and second training molecular sequences, and the second electronic coupling label V sum,T is the sum of the electronic coupling parameters of all the heterogeneous atom pairs in the dimer corresponding to the first and second training molecular sequences, wherein the heterogeneous atom pairs are composed of two atoms belonging to different molecules; the first reorganization energy label λ T is the reorganization energy label value of the dimer corresponding to the first and second training molecular sequences; the first mobility label μ T is the charge carrier mobility label value of the dimer corresponding to the first and second training molecule sequences.

[0028] Preferably, the performing model training on the first prediction model based on a preset first data set specifically includes:

[0029] Step 4-1, randomly dividing the first data set into two sub-data sets as a corresponding first training set and a first evaluation set based on a preset first segmentation ratio;

[0030] Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first segmentation ratio;

[0031] Step 4-2, extracting the first first data record of the first training set as the corresponding current training record;

[0032] Step 4-3, the first and second training molecular sequences of the current training record are used as the current molecular sequences A and B; and the first prediction model predicts the charge carrier mobility between molecules according to the current molecular sequences A and B; and the prediction vector Y generated by this prediction process is 12 , the second vector Y1 and the mobility prediction value C as the corresponding first training vector, the first training reorganization energy λ P and the first training mobility μ P ; and performing mean calculation on all the atom-pair electron coupling parameters of the first training vector to obtain the corresponding first training mean Vaver,P , and summing up all the atomic-pair electron coupling parameters of the first training vector to obtain the corresponding first training sum V sum,P ;

[0033] Step 4-4: The first training mean V aver,P , the first training sum V sum,P and the first electronic coupling tag V of the current training record aver,T , the second electronic coupling tag V sum,T Bring in the preset first model loss function L M1 Calculate and obtain the corresponding first loss value;

[0034] The first model loss function L M1 for:

[0035] L M1 =a1×L1(V aver,P ,V aver,T )+a2×L2(V sum,P ,V sum,T );

[0036] Wherein, a1 and a2 are two preset weight parameters; the first loss function L1 and the second loss function L2 are implemented based on the same type of loss function, which includes at least L1 loss function and L2 loss function;

[0037] Step 4-5, the first training reorganization energy λ P and the first reorganization energy label λ of the current training record T Bring in the preset second model loss function L M2 Calculate and obtain a corresponding second loss value;

[0038] Among them, the second model loss function L M2 At least including L1 loss function and L2 loss function;

[0039] Step 4-6, identifying whether the first and second loss values ​​satisfy the preset first and second loss value ranges; if the first and second loss values ​​do not satisfy the respective corresponding first and second loss value ranges, then based on the preset first model optimizer, the first loss function L1, the second loss function L2 and the first model loss function L M1 The model parameters of the 3D-MPNN model of the first prediction model and the first fully connected network are modulated in a direction in which both of them reach the minimum value, and based on the preset second model optimizer, the second model loss function L is adjusted. M2The 2D-MPNN model of the first prediction model and the model parameters of the second fully connected network are modulated in a round in the direction of reaching the minimum value, and when the two rounds of parameter modulation are completed, return to step 4-3; if the first loss value does not satisfy the corresponding first loss value range but the second loss value satisfies the corresponding second loss value range, only based on the first model optimizer, the first loss function L1, the second loss function L2 and the first model loss function L M1 The model parameters of the 3D-MPNN model and the first fully connected network are modulated in a direction in which both of them reach the minimum value, and return to step 4-3 at the end of this round of parameter modulation; if the second loss value does not satisfy the corresponding second loss value range but the first loss value satisfies the corresponding first loss value range, then only based on the second model optimizer, the second model loss function L is adjusted in a direction in which both of them reach the minimum value. M2 Perform a round of modulation on the model parameters of the 2D-MPNN model and the second fully connected network in the direction of reaching the minimum value, and return to step 4-3 when this round of parameter modulation ends; if the first and second loss values ​​both satisfy the first and second loss value ranges corresponding to each other, go to step 4-7;

[0040] Wherein, the first and second model optimizers both include at least an SGD optimizer and an ADAM optimizer;

[0041] Step 4-7, the first training mobility μ P and the first mobility label μ of the current training record T Bring in the preset third model loss function L M3 Calculate and obtain the corresponding third loss value;

[0042] Among them, the third model loss function L M3 At least including L1 loss function and L2 loss function;

[0043] Step 4-8, identifying whether the third loss value satisfies a preset third loss value range; if the third loss value does not satisfy the third loss value range, then based on a preset third model optimizer, moving toward making the third model loss function L M3 Perform a round of modulation on the model parameters of the mobility prediction model of the first prediction model in the direction of reaching the minimum value, and return to step 4-3 when this round of parameter modulation ends; if the third loss value satisfies the third loss value range, go to step 4-9;

[0044] Wherein, the third model optimizer includes at least an SGD optimizer and an ADAM optimizer;

[0045] Step 4-9, identifying whether the current training record is the last first data record of the first training set; if so, going to step 4-10; if not, extracting the next first data record of the first training set as the new current training record and returning to step 4-3;

[0046] Step 4-10, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and use the first and second training molecular sequences of the current evaluation record as the current molecular sequences A and B; and use the first prediction model to predict the charge carrier mobility between molecules according to the current molecular sequences A and B to obtain the mobility prediction value C; and use the mobility prediction value C obtained this time and the first mobility label μ of the current evaluation record to obtain the mobility prediction value C. T A corresponding prediction-evaluation data pair is formed; and at the end of this round of traversal, all the obtained prediction-evaluation data pairs are brought into a preset first evaluation function to calculate and obtain a corresponding first evaluation value;

[0047] Wherein, the first evaluation function at least includes a root mean square error function, a mean absolute error function and a standard deviation function;

[0048] Step 4-11, identifying whether the first evaluation value satisfies the preset first evaluation value range; if the first evaluation value does not satisfy the first evaluation value range, returning to step 4-1 to continue training; if the first evaluation value satisfies the first evaluation value range, stopping training and confirming that the model training of the first prediction model is completed.

[0049] Preferably, the molecule pair sequence list includes a plurality of molecule pair sequence records; the molecule pair sequence records include a first molecule sequence and a second molecule sequence;

[0050] The batch prediction report includes a plurality of molecule pair prediction records; the molecule pair prediction records correspond to the molecule pair sequence records one by one; and the molecule pair prediction records are composed of the corresponding molecule pair sequence records and predicted mobility.

[0051] Preferably, performing batch prediction processing on the molecular pair sequence list based on the first prediction model to obtain a corresponding batch prediction report and feeding it back to the user specifically includes:

[0052] A round of traversal is performed on all the molecule pair sequence records in the molecule pair sequence list; and in this round of traversal, the molecule pair sequence record currently traversed is used as the corresponding current record; and the first and second training molecule sequences of the current record are used as the corresponding molecule sequences A and B and input into the first prediction model for prediction processing to obtain the corresponding mobility prediction value C as a corresponding predicted mobility; and the current record and the corresponding predicted mobility form a corresponding molecule pair prediction record; and at the end of this round of traversal, all the obtained molecule pair prediction records form the corresponding batch prediction report to feedback to the user.

[0053] A second aspect of an embodiment of the present invention provides a device for implementing the processing method of the charge carrier mobility prediction model described in the first aspect, the device comprising: a model building module, a model training module and a model application module;

[0054] The model building module is used to build a prediction model for predicting charge carrier mobility according to two input molecular sequences, which is recorded as the corresponding first prediction model; the first prediction model predicts the charge carrier mobility between molecules according to the input molecular sequences A and B and outputs the corresponding mobility prediction value C;

[0055] The model training module is used to perform model training on the first prediction model based on a preset first data set;

[0056] The model application module is used to receive a list of molecule pair sequences input by a user after model training is completed; and perform batch prediction processing on the list of molecule pair sequences based on the first prediction model to obtain a corresponding batch prediction report and provide feedback to the user.

[0057] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0058] The processor is used to be coupled to the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above;

[0059] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0060] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions. When the computer instructions are executed by a computer, the computer executes the instructions of the method described in the first aspect above.

[0061] The embodiment of the present invention provides a method, device, electronic device and computer-readable storage medium for processing a charge carrier mobility prediction model. As can be seen from the above content, the embodiment of the present invention constructs an end-to-end prediction model for predicting charge carrier mobility based on two input molecular sequences, which is recorded as a first prediction model, and trains the model based on a preset first data set; and after the model training is completed, the first prediction model is used to batch predict the sequence list of molecular pairs input by the user. Applying the first prediction model provided by the embodiment of the present invention to the design process of organic semiconductor material molecules not only shortens the design cycle but also improves the design efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a method for processing a charge carrier mobility prediction model provided in Embodiment 1 of the present invention;

[0063] Figure 2 A schematic diagram of a module of a first prediction model provided in Embodiment 1 of the present invention;

[0064] Figure 3 A module structure diagram of a processing device for a charge carrier mobility prediction model provided in Embodiment 2 of the present invention;

[0065] Figure 4 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] Embodiment 1 of the present invention provides a method for processing a charge carrier mobility prediction model, such as Figure 1 As shown in a schematic diagram of a processing method of a charge carrier mobility prediction model provided in the first embodiment of the present invention, the method mainly comprises the following steps:

[0068] Step 1: construct a prediction model for predicting charge carrier mobility based on two input molecular sequences, which is recorded as the corresponding first prediction model.

[0069] Here, the first prediction model of the embodiment of the present invention is used to predict the charge carrier mobility between molecules according to the input molecular sequences A and B and output the corresponding mobility prediction value C.

[0070] like Figure 2 As shown in the module schematic diagram of the first prediction model provided in the first embodiment of the present invention, the model input end of the first prediction model of the embodiment of the present invention is used to receive the corresponding molecular sequences A and B, and the model output end is used to output the corresponding mobility prediction value C; the molecular sequence A and the molecular sequence B of the molecular sequences A and B here are each the SMILES sequence of a molecule.

[0071] like Figure 2 As shown, the first prediction model of the embodiment of the present invention includes a preprocessing module, a molecular conformation encoding module, an electron coupling prediction model, a recombinant energy prediction model, a vector splicing module and a mobility prediction model; wherein, the connection mode of each internal component is: the input end of the preprocessing module is connected to the model input end, and the first and second output ends are respectively connected to the input ends of the molecular conformation encoding module and the recombinant energy prediction model; the output end of the molecular conformation encoding module is connected to the input end of the electron coupling prediction model; the output ends of the electron coupling prediction model and the recombinant energy prediction model are respectively connected to the first and second input ends of the vector splicing module; the output end of the vector splicing module is connected to the input end of the mobility prediction model; the output end of the mobility prediction model is connected to the output end of the model.

[0072] The functions of the various internal components of the first prediction model according to the embodiment of the present invention are briefly described below.

[0073] 1) Preprocessing module:

[0074] The preprocessing module is used to input the input molecular sequence A and molecular sequence B into the first processing interface provided by the preset chemical informatics tool respectively to perform molecular sequence standardization processing to obtain the corresponding first and second standard sequences; and input the first and second standard sequences into the second processing interface provided by the chemical informatics tool respectively to perform molecular sequence two-dimensional topological map conversion processing to obtain the corresponding first and second conversion topological maps; and input the first and second standard sequences into the third processing interface provided by the chemical informatics tool respectively to perform molecular sequence three-dimensional conformation conversion processing to obtain the corresponding first and second conversion conformations; and perform conformation merging on the first and second conversion conformations to obtain the corresponding dimer conformation X1; and perform topological map merging on the first and second conversion topological maps to obtain the corresponding dimer topological map X2; and send the dimer conformation X1 to the molecular conformation encoding module; and send the dimer topological map X2 to the reorganization energy prediction model.

[0075] Here, the chemical informatics tools include at least OpenBabel and RDKit.

[0076] The first, second and third processing interfaces are three processing interfaces provided by the chemical informatics tool; the first processing interface is used to perform standardized sequence conversion on the molecular sequence input into the interface according to a preset standardized molecular sequence arrangement rule and output the obtained standardized sequence as the interface processing result; the standardized sequence output by the first processing interface for different SMIELS sequence expressions of the same molecule is the same; the second processing interface is used to generate a corresponding two-dimensional molecular topology map according to the molecular sequence input into the interface and output the obtained two-dimensional molecular topology map as the interface processing result; the third processing interface is used to generate a corresponding three-dimensional molecular conformation according to the molecular sequence input into the interface and output the obtained three-dimensional molecular conformation as the interface processing result.

[0077] The dimer conformation X1 is composed of a first atomic vector set and a first chemical bond vector set; the first atomic vector set is composed of multiple first atomic vectors; the first atomic vector includes a first atomic identifier, a first molecule identifier, a first atomic type and a first atomic coordinate; the first atomic identifier is a unique identifier of the corresponding atom; the first molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the first atomic type is the element type of the corresponding atom; the first atomic coordinate is the three-dimensional coordinate of the corresponding atom; the first chemical bond vector set is composed of multiple first chemical bond vectors; the first chemical bond vector includes a first bond identifier, a first bond length, a first bond type and a first atomic identifier group; the first bond identifier is a unique identifier of the corresponding chemical bond; the first bond length is the length of the corresponding chemical bond; the first bond type is the chemical bond type of the corresponding chemical bond; the first atomic identifier group is composed of the atomic identifiers of the bond head and bond tail atoms of the corresponding chemical bond.

[0078] The dimer topology diagram X2 is composed of a second atomic vector set and a second chemical bond vector set; the second atomic vector set is composed of multiple second atomic vectors; the second atomic vector includes a second atomic identifier, a second molecule identifier, a second atomic type and a second atomic coordinate; the second atomic identifier is a unique identifier of the corresponding atom; the second molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the second atomic type is the element type of the corresponding atom; the second atomic coordinate is the two-dimensional coordinate of the corresponding atom; the second chemical bond vector set is composed of multiple second chemical bond vectors; the second chemical bond vector includes a second bond identifier, a second bond length, a second bond type and a second atomic identifier group; the second bond identifier is a unique identifier of the corresponding chemical bond; the second bond length is the length of the corresponding chemical bond; the second bond type is the chemical bond type of the corresponding chemical bond; the second atomic identifier group is composed of the atomic identifiers of the bond head and bond tail atoms of the corresponding chemical bond.

[0079] 2) Molecular conformation encoding module:

[0080] The molecular conformation encoding module is composed of a preset first number N EThe Uni-Mol models are connected in sequence; each Uni-Mol model is used to optimize the three-dimensional conformation of the current input and output the corresponding optimized conformation; each Uni-Mol model has completed model training in advance; the molecular conformation encoding module is used to use N E The Uni-Mol model optimizes the dimer conformation X1 received this time step by step and uses the optimized conformation output by the last Uni-Mol model as the corresponding coded output conformation X E Send to the electron coupling prediction model;

[0081] Here, the first number N E A preset positive integer.

[0082] The Uni-Mol model used in the embodiment of the present invention is used to perform molecular conformation optimization processing on the input three-dimensional molecular conformation and output the corresponding optimized molecular conformation; the model is an end-to-end three-dimensional molecular conformation optimization model implemented based on the transformer model, and its model design details can be understood through the public technical document "UNI-MOL: A UNIVERSAL 3DMOLECULAR REPRESENTATION LEARNING FRAMEWORK". It can be known from the technical document that based on the Uni-Mol model, a new three-dimensional molecular conformation can be generated according to the input three-dimensional molecular conformation; the pre-training and fine-tuning training scheme of the model can also be completed by referring to the training scheme given in the above technical document.

[0083] The encoded output conformation X output by the molecular conformation encoding module E It is composed of a third atomic vector set and a third chemical bond vector set; the third atomic vector set is composed of multiple third atomic vectors; the third atomic vector includes a third atomic identifier, a third molecule identifier, a third atomic type and a third atomic coordinate; the third atomic identifier is a unique identifier of the corresponding atom; the third molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the third atomic type is the element type of the corresponding atom; the third atomic coordinate is the three-dimensional coordinate of the corresponding atom; the third chemical bond vector set is composed of multiple third chemical bond vectors; the third chemical bond vector includes a third bond identifier, a third bond length, a third bond type and a third atomic identifier group; the third bond identifier is a unique identifier of the corresponding chemical bond; the third bond length is the length of the corresponding chemical bond; the third bond type is the chemical bond type of the corresponding chemical bond; the third atomic identifier group is composed of the atomic identifiers of the head and tail atoms of the corresponding chemical bond.

[0084] 3) Electron coupling prediction model:

[0085] The electron coupling prediction model is composed of a 3D-MPNN model, a first fully connected network, and a feature fusion module connected in sequence; the electron coupling prediction model is used to encode the output conformation XE Input the 3D-MPNN model to extract features and obtain the corresponding feature vector Y 11 ; and the feature vector Y 11 Input the first fully connected network for regression calculation to obtain the corresponding prediction vector Y 12 ; and the feature vector Y 11 and the prediction vector Y 12 The input feature fusion module performs feature fusion processing to obtain the corresponding first vector Y1; and the first vector Y1 is sent to the vector splicing module.

[0086] The 3D-MPNN model is implemented based on the SphereNet model; the 3D-MPNN model is used to encode the input output conformation X E The three-dimensional spherical features, two-dimensional plane angle features and one-dimensional radial features between the two atoms of each group of atomic pairs between the two molecules are extracted to obtain the corresponding atomic pair space feature vector; and each atomic pair space feature vector is vector-joined with its corresponding two third atomic types and the obtained spliced ​​vector is used as the corresponding atomic pair feature vector; and all the obtained atomic pair feature vectors are sequentially joined to form the corresponding feature vector Y 11 ; Among them, each group of atomic pairs between two molecules is composed of two atoms belonging to different molecules.

[0087] The first fully connected network is composed of two or more fully connected layers connected in sequence; the first fully connected network is used to 11 The corresponding atomic pair electron coupling parameters are regressed and calculated for each atomic pair feature vector to obtain the corresponding atomic pair electron coupling parameters; and the corresponding prediction vector Y is formed by sorting all the obtained atomic pair electron coupling parameters in order. 12 .

[0088] The feature fusion module is used to align the feature vector Y 11 and the prediction vector Y 12 The atomic pair feature vectors and atomic pair electronic coupling parameters corresponding to the same group of atomic pairs are extracted to form a corresponding fused feature vector; and all the obtained fused feature vectors are sorted in sequence to form the corresponding first vector Y1.

[0089] It should be noted that the electronic coupling prediction model constructed in the embodiment of the present invention is implemented by referring to the prediction model for predicting the electronic coupling properties of dimers in the technical document "Rapid Estimation of the Intermolecular Electronic Couplings and Charge-Carrier Mobilities of Crystalline Molecular Organic Semiconductors through a Machine Learning Pipeline"; the 3D-MPNN model and the first fully connected network used in the electronic coupling prediction model are also implemented by referring to the SphereNet model and FFN network mentioned in the above technical document.

[0090] 4) Reorganization energy prediction model:

[0091] The recombinant energy prediction model is composed of a 2D-MPNN model and a second fully connected network connected in sequence; the recombinant energy prediction model is used to input the dimer topology X2 into the 2D-MPNN model for feature extraction to obtain the corresponding feature vector Y 21 ; and the feature vector Y 21 The second fully connected network is input for regression calculation processing to obtain the corresponding second vector Y2; and the second vector Y2 is sent to the vector splicing module.

[0092] The 2D-MPNN model is implemented based on a type of GNN model; the 2D-MPNN model is used to extract atomic features from the input dimer topology X2 to obtain the corresponding atomic feature vector; and all the obtained atomic feature vectors are sequentially concatenated to form the corresponding feature vector Y 21 .

[0093] The second fully connected network is composed of two or more fully connected layers connected in sequence; the second fully connected network is used to 21 The corresponding dimer reorganization energy parameter regression calculation is performed to obtain the corresponding dimer reorganization energy parameter; and the obtained dimer reorganization energy parameter is used to form the corresponding second vector Y1.

[0094] It should be noted that the recombinant energy prediction model constructed in the embodiment of the present invention is implemented by referring to the prediction model for predicting the recombinant energy properties of dimers in the technical document "Electronic, redox, and optical property prediction of organic p-conjugated molecules through a hierarchy of machine learning approache"; the 2D-MPNN model and the second fully connected network used in the recombinant energy prediction model are also implemented by referring to the MPNN model and FFN network mentioned in the above technical document.

[0095] 5) Vector splicing module:

[0096] The vector splicing module is used to perform vector splicing processing on the first vector Y1 and the second vector Y2 to obtain a corresponding splicing vector and send it to the Z-direction mobility prediction model.

[0097] 6) Mobility prediction model:

[0098] The mobility prediction model is composed of two or more fully connected layers connected in sequence; the mobility prediction model is used to perform corresponding charge carrier mobility prediction processing according to the input splicing vector Z to obtain the corresponding mobility prediction value C and output it.

[0099] Here, it is known from the well-known Marcus theory that when the spatial structure / electronic coupling / reorganization energy characteristics of a dimer are known, the corresponding Marcus hopping rate can be calculated by a fixed calculation method, and after the Marcus hopping rate is obtained, the corresponding charge carrier mobility can be calculated by a fixed calculation method. The mobility prediction model of the embodiment of the present invention is actually a regression calculation model that simulates the above calculation process.

[0100] Step 2: Performing model training on a first prediction model based on a preset first data set;

[0101] The first data set includes a plurality of first data records; the first data record includes a first training molecule sequence, a second training molecule sequence, a first electronic coupling tag V aver,T , the second electronic coupling tag V sum,T , the first reorganization energy label λ T and the first mobility label μ T ; The first and second training molecular sequences are each a SMILES sequence of one molecule; the first electronic coupling label V aver,T is the average value of the electronic coupling parameters of each heterogeneous atomic pair in the dimer corresponding to the first and second training molecular sequences. The second electronic coupling label Vsum,T is the sum of the electronic coupling parameters of all heterogeneous atomic pairs in the dimer corresponding to the first and second training molecular sequences. The heterogeneous atomic pairs consist of two atoms belonging to different molecules. The first reorganization energy label λ T is the reorganization energy label value of the dimer corresponding to the first and second training molecular sequences; the first mobility label μ T is the charge carrier mobility label value of the dimer corresponding to the first and second training molecule sequences;

[0102] Specifically comprising: step 2-1, randomly dividing the first data set into two sub-data sets as a corresponding first training set and a first evaluation set based on a preset first segmentation ratio;

[0103] The first split ratio is a preset ratio, such as 8:2; the first training set and the first evaluation set are both composed of a plurality of first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first split ratio;

[0104] Step 2-2, extracting the first first data record of the first training set as the corresponding current training record;

[0105] Step 2-3, the first and second training molecular sequences of the current training record are used as the current molecular sequences A and B; and the first prediction model predicts the charge carrier mobility between molecules according to the current molecular sequences A and B; and the prediction vector Y generated by this prediction process is 12 , the second vector Y1 and the mobility prediction value C as the corresponding first training vector, the first training reorganization energy λ P and the first training mobility μ P ; and calculate the mean value of the electron coupling parameters of all atoms of the first training vector to obtain the corresponding first training mean value V aver,P , and the sum of all the atomic pairs of electron coupling parameters of the first training vector is calculated to obtain the corresponding first training sum V sum,P ;

[0106] Step 2-4: Set the first training mean V aver,P , the first training sum V sum,P and the first electronic coupling tag V of the current training record aver,T , the second electronic coupling tag V sum,T Bring in the preset first model loss function L M1 Calculate and obtain the corresponding first loss value;

[0107] Here, the first model loss function L M1 for:

[0108] L M1=a1×L1(V aver,P ,V aver,T )+a2×L2(V sum,P ,V sum,T );

[0109] Wherein, a1 and a2 are two preset weight parameters; the first loss function L1 and the second loss function L2 are implemented based on the same type of loss function, which includes at least L1 loss function and L2 loss function;

[0110] Step 2-5, the first training reorganization energy λ P and the first reorganization energy label λ of the current training record T Bring in the preset second model loss function L M2 Calculate and obtain a corresponding second loss value;

[0111] Among them, the second model loss function L M2 At least including L1 loss function and L2 loss function;

[0112] Step 2-6, identifying whether the first and second loss values ​​satisfy the preset first and second loss value ranges; if the first and second loss values ​​do not satisfy the respective corresponding first and second loss value ranges, then based on the preset first model optimizer, the first loss function L1, the second loss function L2 and the first model loss function L M1 The model parameters of the 3D-MPNN model of the first prediction model and the first fully connected network are modulated in a direction in which both reach the minimum value, and based on the preset second model optimizer, the second model loss function L is adjusted. M2 The 2D-MPNN model of the first prediction model and the model parameters of the second fully connected network are modulated in a round in the direction of reaching the minimum value, and when the two rounds of parameter modulation are completed, return to step 2-3; if the first loss value does not satisfy the corresponding first loss value range but the second loss value satisfies the corresponding second loss value range, then only based on the first model optimizer, the first loss function L1, the second loss function L2 and the first model loss function L M1 The model parameters of the 3D-MPNN model and the first fully connected network are modulated in a direction in which both of them reach the minimum value, and when this round of parameter modulation ends, the steps 2-3 are returned; if the second loss value does not satisfy the corresponding second loss value range but the first loss value satisfies the corresponding first loss value range, then only the second model optimizer is directed towards making the second model loss function L M2 Perform a round of modulation on the model parameters of the 2D-MPNN model and the second fully connected network in the direction of reaching the minimum value, and return to step 2-3 at the end of this round of parameter modulation; if the first and second loss values ​​both satisfy the corresponding first and second loss value ranges, go to step 2-7;

[0113] The first and second loss value ranges are two preset loss value ranges; the first and second model optimizers both include at least an SGD optimizer and an ADAM optimizer;

[0114] Step 2-7, the first training mobility μ P and the first migration label μ of the current training record T Bring in the preset third model loss function L M3 Calculate and obtain the corresponding third loss value;

[0115] Among them, the third model loss function L M3 At least including L1 loss function and L2 loss function;

[0116] Step 2-8, identifying whether the third loss value satisfies a preset third loss value range; if the third loss value does not satisfy the third loss value range, then based on the preset third model optimizer, the third model loss function L is adjusted to the value of the third model loss function L. M3 A round of modulation is performed on the model parameters of the mobility prediction model of the first prediction model in the direction of reaching the minimum value, and when this round of parameter modulation is finished, the step 2-3 is returned; if the third loss value satisfies the third loss value range, the step 2-9 is turned to;

[0117] The third loss value range is a preset loss value range; the third model optimizer includes at least an SGD optimizer and an ADAM optimizer;

[0118] Step 2-9, identifying whether the current training record is the last first data record of the first training set; if so, go to step 2-10; if not, extract the next first data record of the first training set as the new current training record and return to step 2-3;

[0119] Step 2-10, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and use the first and second training molecular sequences of the current evaluation record as the current molecular sequences A and B; and use the first prediction model to predict the charge carrier mobility between molecules according to the current molecular sequences A and B to obtain a mobility prediction value C; and use the mobility prediction value C obtained this time and the first mobility label μ of the current evaluation record T A corresponding prediction-evaluation data pair is formed; and at the end of this round of traversal, all the obtained prediction-evaluation data pairs are brought into a preset first evaluation function to calculate and obtain a corresponding first evaluation value;

[0120] Wherein, the first evaluation function at least includes a root mean square error function, a mean absolute error function and a standard deviation function;

[0121] Step 2-11, identify whether the first evaluation value meets the preset first evaluation value range; if the first evaluation value does not meet the first evaluation value range, return to step 2-1 to continue training; if the first evaluation value meets the first evaluation value range, stop training and confirm that the model training of the first prediction model is completed.

[0122] Here, the first evaluation value range is a preset evaluation value range.

[0123] Step 3: After the model training is completed, the list of molecule pair sequences input by the user is received; and batch prediction processing is performed on the list of molecule pair sequences based on the first prediction model to obtain a corresponding batch prediction report and provide feedback to the user;

[0124] Specifically, it includes: step 31, after the model training is completed, receiving a list of molecular pair sequences input by a user;

[0125] Step 32, batch prediction processing is performed on the molecular pair sequence list based on the first prediction model to obtain a corresponding batch prediction report and provide feedback to the user;

[0126] Wherein, the molecule pair sequence list includes a plurality of molecule pair sequence records; the molecule pair sequence record includes a first molecule sequence and a second molecule sequence;

[0127] The batch prediction report includes multiple molecular pair prediction records; the molecular pair prediction records correspond to the molecular pair sequence records one by one; the molecular pair prediction records consist of the corresponding molecular pair sequence records and the predicted mobility;

[0128] The current step 32 specifically includes: performing a round of traversal on all the molecule pair sequence records in the molecule pair sequence list; and in this round of traversal, taking the currently traversed molecule pair sequence record as the corresponding current record; and inputting the first and second training molecule sequences of the current record as the corresponding molecule sequences A and B into the first prediction model for prediction processing to obtain the corresponding mobility prediction value C as a corresponding predicted mobility; and forming a corresponding molecule pair prediction record by the current record and the corresponding predicted mobility; and at the end of this round of traversal, forming a corresponding batch prediction report by all the obtained molecule pair prediction records to feedback to the user.

[0129] Figure 3 A module structure diagram of a charge carrier mobility prediction model processing device provided in the second embodiment of the present invention, the device is a terminal device or server that implements the aforementioned method embodiment, and can also be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiment, for example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 3As shown, the device includes: a model building module 201, a model training module 202 and a model application module 203.

[0130] The model building module 201 is used to build a prediction model for predicting charge carrier mobility based on two input molecular sequences, which is recorded as the corresponding first prediction model; the first prediction model predicts the charge carrier mobility between molecules based on the input molecular sequences A and B and outputs the corresponding mobility prediction value C.

[0131] The model training module 202 is used to perform model training on the first prediction model based on a preset first data set.

[0132] The model application module 203 is used to receive the molecule pair sequence list input by the user after the model training is completed; and perform batch prediction processing on the molecule pair sequence list based on the first prediction model to obtain a corresponding batch prediction report and feedback it to the user.

[0133] An embodiment of the present invention provides a processing device for a charge carrier mobility prediction model, which can execute the method steps in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0134] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the model building module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above-mentioned module is determined. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0135] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital signal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0136] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the above method embodiments are generated. The above computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The above-mentioned computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above-mentioned computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) methods. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The above-mentioned available medium can be a magnetic medium (such as a floppy disk, a hard disk, a tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0137] Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. The electronic device may be a terminal device or a server for implementing the method of the aforementioned embodiment, or may be a terminal device or a server for implementing the method of the aforementioned embodiment connected to the aforementioned terminal device or server. Figure 4As shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver 303. Various instructions may be stored in the memory 302 to complete various processing functions and implement the processing steps described in the aforementioned embodiment method. Preferably, the electronic device involved in the embodiment of the present invention also includes: a power supply 304, a system bus 305 and a communication port 306. The system bus 305 is used to realize the communication connection between components. The above-mentioned communication port 306 is used for connecting and communicating between the electronic device and other peripherals.

[0138] exist Figure 4 The system bus 305 mentioned in the figure can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize the communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM), and may also include non-volatile memory (Non-Volatile Memory), such as at least one disk storage.

[0139] The above-mentioned processor can be a general-purpose processor, including a central processing unit CPU, a network processor (Network Processor, NP), a graphics processing unit (Graphics Processing Unit, GPU), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0140] It should be noted that an embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the method and processing process provided in the above embodiments.

[0141] The embodiment of the present invention provides a method, device, electronic device and computer-readable storage medium for processing a charge carrier mobility prediction model. As can be seen from the above content, the embodiment of the present invention constructs an end-to-end prediction model for predicting charge carrier mobility based on two input molecular sequences, which is recorded as a first prediction model, and trains the model based on a preset first data set; and after the model training is completed, the first prediction model is used to batch predict the sequence list of molecular pairs input by the user. Applying the first prediction model provided by the embodiment of the present invention to the design process of organic semiconductor material molecules not only shortens the design cycle but also improves the design efficiency.

[0142] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0143] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for processing a charge carrier mobility prediction model, characterized in that: The method comprises: Constructing a prediction model for predicting charge carrier mobility according to two input molecular sequences, which is recorded as a corresponding first prediction model; the first prediction model predicts the charge carrier mobility between molecules according to the input molecular sequences A and B and outputs a corresponding mobility prediction value C; Performing model training on the first prediction model based on a preset first data set; After the model training is completed, a list of molecule pair sequences input by a user is received; and batch prediction processing is performed on the list of molecule pair sequences based on the first prediction model to obtain a corresponding batch prediction report and provide feedback to the user; Wherein, the model input end of the first prediction model is used to receive the corresponding molecular sequences A and B, and the model output end is used to output the corresponding mobility prediction value C; The molecular sequence A and the molecular sequence B of the molecular sequences A and B are each a SMILES sequence of one molecule; The first prediction model includes a preprocessing module, a molecular conformation encoding module, an electronic coupling prediction model, a reorganization energy prediction model, a vector splicing module and a mobility prediction model; The input end of the preprocessing module is connected to the input end of the model, and the first and second output ends are respectively connected to the input ends of the molecular conformation encoding module and the reorganization energy prediction model; the output end of the molecular conformation encoding module is connected to the input end of the electron coupling prediction model; the output ends of the electron coupling prediction model and the reorganization energy prediction model are respectively connected to the first and second input ends of the vector splicing module; the output end of the vector splicing module is connected to the input end of the mobility prediction model; the output end of the mobility prediction model is connected to the output end of the model; The preprocessing module is used to input the input molecular sequence A and the molecular sequence B into the first processing interface provided by the preset chemical informatics tool respectively to perform molecular sequence standardization processing to obtain the corresponding first and second standard sequences; and input the first and second standard sequences into the second processing interface provided by the chemical informatics tool respectively to perform molecular sequence two-dimensional topological map conversion processing to obtain the corresponding first and second conversion topological maps; and input the first and second standard sequences into the third processing interface provided by the chemical informatics tool respectively to perform molecular sequence three-dimensional conformation conversion processing to obtain the corresponding first and second conversion conformations; and perform conformation merging on the first and second conversion conformations to obtain the corresponding dimer conformation X1; and perform topological map merging on the first and second conversion topological maps to obtain the corresponding dimer topological map X2; and send the dimer conformation X1 to the molecular conformation encoding module; and send the dimer topological map X2 to the reorganization energy prediction model; The chemical informatics tools include at least OpenBabel and RDKit; The dimer conformation X1 is composed of a first atomic vector set and a first chemical bond vector set; the first atomic vector set is composed of multiple first atomic vectors; the first atomic vector includes a first atomic identifier, a first molecule identifier, a first atomic type and a first atomic coordinate; the first atomic identifier is a unique identifier of the corresponding atom; the first molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the first atomic type is the element type of the corresponding atom; the first atomic coordinate is the three-dimensional coordinate of the corresponding atom; the first chemical bond vector set is composed of multiple first chemical bond vectors; the first chemical bond vector includes a first bond identifier, a first bond length, a first bond type and a first atomic identifier group; the first bond identifier is a unique identifier of the corresponding chemical bond; the first bond length is the length of the corresponding chemical bond; the first bond type is the chemical bond type of the corresponding chemical bond; the first atomic identifier group is composed of the atomic identifiers of the head and tail atoms of the corresponding chemical bond; The dimer topology map X2 is composed of a second atomic vector set and a second chemical bond vector set; the second atomic vector set is composed of multiple second atomic vectors; the second atomic vector includes a second atomic identifier, a second molecule identifier, a second atomic type and a second atomic coordinate; the second atomic identifier is a unique identifier of the corresponding atom; the second molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the second atomic type is the element type of the corresponding atom; the second atomic coordinate is the two-dimensional coordinate of the corresponding atom; the second chemical bond vector set is composed of multiple second chemical bond vectors; the second chemical bond vector includes a second bond identifier, a second bond length, a second bond type and a second atomic identifier group; the second bond identifier is a unique identifier of the corresponding chemical bond; the second bond length is the length of the corresponding chemical bond; the second bond type is the chemical bond type of the corresponding chemical bond; the second atomic identifier group is composed of the atomic identifiers of the head and tail atoms of the corresponding chemical bond; The molecular conformation encoding module is composed of a preset first number N E The Uni-Mol models are sequentially connected; each Uni-Mol model is used to perform conformation optimization processing on the currently input three-dimensional conformation and output the corresponding optimized conformation; each Uni-Mol model has completed model training in advance; the molecular conformation encoding module is used to use N E The Uni-Mol model optimizes the dimer conformation X1 received this time step by step and uses the optimized conformation output by the last Uni-Mol model as the corresponding coded output conformation X E Sending to the electronic coupling prediction model; The encoding outputs the conformation X E It is composed of a third atomic vector set and a third chemical bond vector set; the third atomic vector set is composed of multiple third atomic vectors; the third atomic vector includes a third atomic identifier, a third molecule identifier, a third atomic type and a third atomic coordinate; the third atomic identifier is a unique identifier of the corresponding atom; the third molecule identifier is a unique identifier of the molecule to which the corresponding atom belongs; the third atomic type is the element type of the corresponding atom; the third atomic coordinate is the three-dimensional coordinate of the corresponding atom; the third chemical bond vector set is composed of multiple third chemical bond vectors; the third chemical bond vector includes a third bond identifier, a third bond length, a third bond type and a third atomic identifier group; the third bond identifier is a unique identifier of the corresponding chemical bond; the third bond length is the length of the corresponding chemical bond; the third bond type is the chemical bond type of the corresponding chemical bond; the third atomic identifier group is composed of the atomic identifiers of the bond head and bond tail atoms of the corresponding chemical bond; The electronic coupling prediction model is formed by sequentially connecting a 3D-MPNN model, a first fully connected network and a feature fusion module; the electronic coupling prediction model is used to convert the encoded output conformation X E Input the 3D-MPNN model to perform feature extraction and obtain the corresponding feature vector Y 11 ; and the feature vector Y 11 Input the first fully connected network for regression calculation to obtain the corresponding prediction vector Y 12 ; and the feature vector Y 11 and the prediction vector Y 12 Input the feature fusion module to perform feature fusion processing to obtain a corresponding first vector Y1; and send the first vector Y1 to the vector splicing module; The 3D-MPNN model is implemented based on the SphereNet model; the 3D-MPNN model is used to output the encoding conformation X of the input E The three-dimensional spherical features, two-dimensional plane angle features and one-dimensional radial features between two atoms of each group of atomic pairs between two molecules are extracted to obtain the corresponding atomic pair space feature vector; and each of the atomic pair space feature vectors and its corresponding two third atomic types are vector-joined and the obtained joint vector is used as the corresponding atomic pair feature vector; and all the obtained atomic pair feature vectors are sequentially joined to form the corresponding feature vector Y 11 ; Wherein, each group of atomic pairs between the two molecules is composed of two atoms belonging to different molecules; The first fully connected network is formed by sequentially connecting two or more fully connected layers; the first fully connected network is used to input the feature vector Y 11 The corresponding atom pair electron coupling parameters are regressed and calculated for each of the atom pair feature vectors to obtain the corresponding atom pair electron coupling parameters; and the corresponding prediction vector Y is formed by sequentially sorting all the obtained atom pair electron coupling parameters. 12 ; The feature fusion module is used to align the feature vector Y according to the atomic pair alignment method. 11 and the prediction vector Y 12 The atomic pair feature vectors and the atomic pair electron coupling parameters corresponding to the same group of atomic pairs are extracted to form a corresponding fused feature vector; and all the obtained fused feature vectors are sequentially sorted to form the corresponding first vector Y1; The reorganization energy prediction model is formed by sequentially connecting the 2D-MPNN model and the second fully connected network; the reorganization energy prediction model is used to input the dimer topology X2 into the 2D-MPNN model for feature extraction to obtain the corresponding feature vector Y 21 ; and the feature vector Y 21 Input the second fully connected network to perform regression calculation to obtain a corresponding second vector Y2; and send the second vector Y2 to the vector splicing module; The 2D-MPNN model is implemented based on a type of GNN model; the 2D-MPNN model is used to extract atomic features from the input dimer topology X2 to obtain a corresponding atomic feature vector; and all the obtained atomic feature vectors are sequentially spliced ​​to form the corresponding feature vector Y 21 ; The second fully connected network is formed by sequentially connecting two or more fully connected layers; the second fully connected network is used to input the feature vector Y 21 Performing corresponding dimer reorganization energy parameter regression calculation to obtain corresponding dimer reorganization energy parameters; and forming the corresponding second vector Y1 from the obtained dimer reorganization energy parameters; The vector splicing module is used to perform vector splicing processing on the first vector Y1 and the second vector Y2 to obtain a corresponding splicing vector Z and send it to the mobility prediction model; The mobility prediction model is formed by sequentially connecting two or more fully connected layers; the mobility prediction model is used to perform corresponding charge carrier mobility prediction processing according to the input splicing vector Z to obtain the corresponding mobility prediction value C and output it.

2. The method for processing the charge carrier mobility prediction model according to claim 1, characterized in that: The first data set includes a plurality of first data records; the first data record includes a first training molecule sequence, a second training molecule sequence, a first electronic coupling tag V aver,T , the second electronic coupling tag V sum,T , the first reorganization energy label λ T and the first mobility label μ T The first and second training molecular sequences are each a SMILES sequence of one molecule; the first electronic coupling label V aver,T is the average value of the electronic coupling parameters of each heterogeneous atomic pair in the dimer corresponding to the first and second training molecular sequences, and the second electronic coupling label V sum,T is the sum of the electronic coupling parameters of all the heterogeneous atom pairs in the dimer corresponding to the first and second training molecular sequences, wherein the heterogeneous atom pairs are composed of two atoms belonging to different molecules; the first reorganization energy label λ T is the reorganization energy label value of the dimer corresponding to the first and second training molecular sequences; the first mobility label μ T is the charge carrier mobility label value of the dimer corresponding to the first and second training molecule sequences.

3. The method for processing the charge carrier mobility prediction model according to claim 2, characterized in that: The performing model training on the first prediction model based on the preset first data set specifically includes: Step 4-1, randomly dividing the first data set into two sub-data sets as a corresponding first training set and a first evaluation set based on a preset first segmentation ratio; Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set to the total number of records in the first evaluation set satisfies the first segmentation ratio; Step 4-2, extracting the first first data record of the first training set as the corresponding current training record; Step 4-3, the first and second training molecular sequences of the current training record are used as the current molecular sequences A and B; and the first prediction model predicts the charge carrier mobility between molecules according to the current molecular sequences A and B; and the prediction vector Y generated by this prediction process is 12 , the second vector Y1 and the mobility prediction value C as the corresponding first training vector, the first training reorganization energy λ P and the first training mobility μ P ; and performing mean calculation on all the atom-pair electron coupling parameters of the first training vector to obtain the corresponding first training mean V aver,P , and summing up all the atomic-pair electron coupling parameters of the first training vector to obtain the corresponding first training sum V sum,P ; Step 4-4: The first training mean V aver,P , the first training sum V sum,P and the first electronic coupling tag V of the current training record aver,T , the second electronic coupling tag V sum,T Bring in the preset first model loss function L M1 Calculate and obtain the corresponding first loss value; The first model loss function L M1 for: L M1 =a1×L1(V aver,P ,V aver,T )+a2×L2(V sum,P ,V sum,T ); Wherein, a1 and a2 are two preset weight parameters; the first loss function L1 and the second loss function L2 are implemented based on the same type of loss function, which includes at least L1 loss function and L2 loss function; Step 4-5, the first training reorganization energy λ P and the first reorganization energy label λ of the current training record T Bring in the preset second model loss function L M2 Calculate and obtain a corresponding second loss value; Among them, the second model loss function L M2 At least including L1 loss function and L2 loss function; Step 4-6, identifying whether the first and second loss values ​​satisfy the preset first and second loss value ranges; if the first and second loss values ​​do not satisfy the respective corresponding first and second loss value ranges, then based on the preset first model optimizer, the first loss function L1, the second loss function L2 and the first model loss function L M1 The model parameters of the 3D-MPNN model of the first prediction model and the first fully connected network are modulated in a direction in which both of them reach the minimum value, and based on the preset second model optimizer, the second model loss function L is adjusted. M2 The 2D-MPNN model of the first prediction model and the model parameters of the second fully connected network are modulated in a round in the direction of reaching the minimum value, and when the two rounds of parameter modulation are completed, return to step 4-3; if the first loss value does not satisfy the corresponding first loss value range but the second loss value satisfies the corresponding second loss value range, only based on the first model optimizer, the first loss function L1, the second loss function L2 and the first model loss function L M1 The model parameters of the 3D-MPNN model and the first fully connected network are modulated in a direction in which both of them reach the minimum value, and return to step 4-3 at the end of this round of parameter modulation; if the second loss value does not satisfy the corresponding second loss value range but the first loss value satisfies the corresponding first loss value range, then only based on the second model optimizer, the second model loss function L is adjusted in a direction in which both of them reach the minimum value. M2 Perform a round of modulation on the model parameters of the 2D-MPNN model and the second fully connected network in the direction of reaching the minimum value, and return to step 4-3 when this round of parameter modulation ends; if the first and second loss values ​​both satisfy the first and second loss value ranges corresponding to each other, go to step 4-7; Wherein, the first and second model optimizers both include at least an SGD optimizer and an ADAM optimizer; Step 4-7, the first training mobility μ P and the first mobility label μ of the current training record T Bring in the preset third model loss function L M3 Calculate and obtain the corresponding third loss value; Among them, the third model loss function L M3 At least including L1 loss function and L2 loss function; Step 4-8, identifying whether the third loss value satisfies a preset third loss value range; if the third loss value does not satisfy the third loss value range, then based on a preset third model optimizer, moving toward making the third model loss function L M3 Perform a round of modulation on the model parameters of the mobility prediction model of the first prediction model in the direction of reaching the minimum value, and return to step 4-3 when this round of parameter modulation ends; if the third loss value satisfies the third loss value range, go to step 4-9; Wherein, the third model optimizer includes at least an SGD optimizer and an ADAM optimizer; Step 4-9, identifying whether the current training record is the last first data record of the first training set; if so, going to step 4-10; if not, extracting the next first data record of the first training set as the new current training record and returning to step 4-3; Step 4-10, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and use the first and second training molecular sequences of the current evaluation record as the current molecular sequences A and B; and use the first prediction model to predict the charge carrier mobility between molecules according to the current molecular sequences A and B to obtain the mobility prediction value C; and use the mobility prediction value C obtained this time and the first mobility label μ of the current evaluation record to obtain the mobility prediction value C. T A corresponding prediction-evaluation data pair is formed; and at the end of this round of traversal, all the obtained prediction-evaluation data pairs are brought into a preset first evaluation function to calculate and obtain a corresponding first evaluation value; Wherein, the first evaluation function at least includes a root mean square error function, a mean absolute error function and a standard deviation function; Step 4-11, identifying whether the first evaluation value satisfies the preset first evaluation value range; if the first evaluation value does not satisfy the first evaluation value range, returning to step 4-1 to continue training; if the first evaluation value satisfies the first evaluation value range, stopping training and confirming that the model training of the first prediction model is completed.

4. The method for processing a charge carrier mobility prediction model according to claim 1, characterized in that: The molecule pair sequence list includes a plurality of molecule pair sequence records; the molecule pair sequence records include a first molecule sequence and a second molecule sequence; The batch prediction report includes a plurality of molecule pair prediction records; the molecule pair prediction records correspond to the molecule pair sequence records one by one; and the molecule pair prediction records are composed of the corresponding molecule pair sequence records and predicted mobility.

5. The method for processing the charge carrier mobility prediction model according to claim 4, characterized in that: The step of performing batch prediction processing on the molecular pair sequence list based on the first prediction model to obtain a corresponding batch prediction report and feeding it back to the user specifically includes: A round of traversal is performed on all the molecule pair sequence records in the molecule pair sequence list; and in this round of traversal, the molecule pair sequence record currently traversed is used as the corresponding current record; and the first and second training molecule sequences of the current record are used as the corresponding molecule sequences A and B and input into the first prediction model for prediction processing to obtain the corresponding mobility prediction value C as a corresponding predicted mobility; and the current record and the corresponding predicted mobility form a corresponding molecule pair prediction record; and at the end of this round of traversal, all the obtained molecule pair prediction records form the corresponding batch prediction report to feedback to the user.

6. A device for executing the method for processing the charge carrier mobility prediction model according to any one of claims 1 to 5, characterized in that: The device comprises: a model building module, a model training module and a model application module; The model building module is used to build a prediction model for predicting charge carrier mobility according to two input molecular sequences, which is recorded as the corresponding first prediction model; the first prediction model predicts the charge carrier mobility between molecules according to the input molecular sequences A and B and outputs the corresponding mobility prediction value C; The model training module is used to perform model training on the first prediction model based on a preset first data set; The model application module is used to receive a list of molecule pair sequences input by a user after model training is completed; and perform batch prediction processing on the list of molecule pair sequences based on the first prediction model to obtain a corresponding batch prediction report and provide feedback to the user.

7. An electronic device, characterized in that: include: memory, processors, and transceivers; The processor is used to couple with the memory, read and execute instructions in the memory, so as to implement the method according to any one of claims 1 to 5; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer is enabled to execute the method according to any one of claims 1 to 5.

Images