Processing method and device for tRNA (transfer ribonucleic acid) adaptive index weight prediction model

Abstract

The embodiment of the invention relates to a processing method and device for a tRNA (transfer ribonucleic acid) adaptation index weight prediction model. The method comprises the following steps: constructing the adaptation index weight prediction model and three downstream task models; training the four models based on a target species data set; after training is finished, experimental conditions, genome sequences and target genes input by a user are preprocessed, and corresponding experimental condition coding vectors, genome sequence coding vectors, tRNA gene copy number vectors and target gene codon index sets are obtained; predicting by an adaptive index weight prediction model according to the genome sequence coding vector, the experimental condition coding vector and the tRNA gene copy number vector to obtain an adaptive index weight vector; and the transcription efficiency, the translation efficiency and the gene abundance of the target gene are predicted by the three downstream task models according to the adaptation index weight vector and the target gene codon index set respectively. According to the invention, the species suitability and prediction stability can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a processing method and device for a tRNA adaptation index weight prediction model. Background Art

[0002] The tRNA Adaptation Index (tAI) weight is used to reflect the degree of match between each codon and its corresponding tRNA abundance in a given species. The tRNA Adaptation Index weight has important applications in studying gene expression regulation, protein synthesis efficiency, and evolutionary biology. However, the derivation of the tRNA Adaptation Index weight for each codon has always been a difficult problem in bioinformatics research. Most of the existing derivation methods rely on experimental data or are based on simplified theoretical models, which often show great limitations and instability when facing different species. Summary of the invention

[0003] The purpose of the present invention is to provide a processing method, device, electronic device and computer-readable storage medium for a tRNA adaptation index weight prediction model in view of the defects of the prior art. The present invention constructs an adaptation index weight prediction model and three downstream task models (transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) for processing tRNA adaptation index weight prediction tasks; and collects data on the genome sequence of a certain type of target species under different conditions and constructs a data set based on the collected data to train the adaptation index weight prediction model and the three downstream task models; after the training, the adaptation index weight prediction model is used to predict the tRNA adaptation index weight of the current target species under the current given conditions, and the gene expression level (transcription efficiency, translation efficiency, gene abundance) of a given target gene (DNA gene, synthetic gene) is further predicted according to the tRNA adaptation index weight through the three downstream task models. The four types of models provided by the present invention (adaptation index weight prediction model, transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) are all end-to-end prediction models. Prediction of tRNA adaptation index weight and gene expression level based on these four types of models can reduce processing complexity and improve model versatility. The four types of models of the present invention do not need to change the model structure when switching target species, and only need to use the data set of the current species to perform one round of training. This can not only improve the species adaptability of the model, but also further improve the prediction stability of each type of species through training.

[0004] To achieve the above object, a first aspect of an embodiment of the present invention provides a method for processing a tRNA adaptation index weight prediction model, the method comprising:

[0005] Construct an adaptation index weight prediction model for tRNA adaptation index weight prediction tasks; and design three downstream task models: transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model;

[0006] Collecting data on the genome sequences of the target species under different conditions and constructing a first data set based on the collected data; and training the adaptation index weight prediction model and three downstream task models based on the first data set;

[0007] After the model training is completed, the current experimental conditions, current genome sequence and current target gene input by the user are received; and the current experimental conditions, the current genome sequence and the current target gene are preprocessed to obtain the corresponding current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set; the current genome sequence is a genome sequence of a type of cell of the target species obtained by an observer through manual or machine observation / experimental means under the current experimental conditions; the current target gene is a DNA gene or synthetic gene in the current genome sequence;

[0008] The adaptation index weight prediction model performs tRNA adaptation index weight prediction processing according to the current genome sequence encoding vector, the current experimental condition encoding vector and the current tRNA gene copy number vector to obtain the corresponding current adaptation index weight vector; and three downstream task models respectively perform corresponding transcription efficiency, translation efficiency and gene abundance prediction processing according to the current adaptation index weight vector and the current target gene codon index set to obtain the corresponding current transcription efficiency, current translation efficiency and current gene abundance to form a corresponding task report, which is fed back to the user.

[0009] Preferably, the adaptation index weight prediction model is used to perform tRNA adaptation index weight prediction processing according to the input genome sequence encoding vector A, experimental condition encoding vector B and tRNA gene copy number vector C and output the corresponding adaptation index weight vector W;

[0010] The genome sequence encoding vector A is composed of the codon encoding vector A C and anticodon encoding vector A AC spliced; the codon encoding vector A C The codons of all codon types in the corresponding genomic sequence encode c i The anti-codon encoding vector A is arranged in sequence; AC The anticodons of all anticodon types in the corresponding genomic sequence encode ac j The codons are arranged in sequence; the codons encode c iThe total number is recorded as N C , the anticodon encodes ac j The total number is recorded as N AC , 1≤i≤N C , 1≤j≤N AC ; The codon encodes c i and the anticodon encodes ac j Each is a triplet base code;

[0011] The experimental condition coding vector B at least includes temperature coding, humidity coding, light coding, nutritional condition coding, and physiological / culture stage coding;

[0012] The tRNA gene copy number vector C is composed of N AC Gene copy number tGCN j Composition; the gene copy number tGCN j The anticodon encodes ac j One to one correspondence;

[0013] The adaptation index weight vector W is composed of N C The adaptation index weight w i Composition; the adaptation index weight w i The codon encoding c i One to one correspondence;

[0014] The first model input end, the second model input end and the third model input end of the adaptation index weight prediction model are respectively used to receive the corresponding genome sequence encoding vector A, the experimental condition encoding vector B and the tRNA gene copy number vector C, and the model output end is used to output the corresponding adaptation index weight vector W;

[0015] The adaptation index weight prediction model includes a Uni-RNA model, a first MLP model, a second MLP model, a feature fusion module, a third MLP model and an adaptation index weight estimation module;

[0016] The input end of the Uni-RNA model is connected to the input end of the first model, and the output end is connected to the input ends of the first and second MLP models respectively; the output ends of the first and second MLP models are connected to the second and third input ends of the feature fusion module respectively; the first input end of the feature fusion module is connected to the input end of the second model, and the output end is connected to the input end of the third MLP model; the output end of the third MLP model is connected to the second input end of the adaptation index weight estimation module; the first input end of the adaptation index weight estimation module is connected to the input end of the third model, and the output end is connected to the output end of the model;

[0017] The Uni-RNA model has completed pre-training; the Uni-RNA model is used to perform feature encoding on the genome sequence encoding vector A to obtain corresponding encoding features X and send them to the first and second MLP models;

[0018] The first MLP model is used to predict the intracellular concentration of the tRNA gene corresponding to each anticodon type according to the coding feature X to obtain the corresponding concentration prediction vector P and send it to the feature fusion module; the concentration prediction vector P is composed of N AC The predicted concentration p j composition, the predicted concentration p j The anticodon encodes ac j One to one correspondence;

[0019] The second MLP model is used to predict the pairing probability of each group of codon-anticodon type pairs according to the coding feature X to obtain the corresponding pairing probability matrix M and send it to the feature fusion module; the shape of the pairing probability matrix M is N C ×N AC , by N C ×N AC The pairing probability m i,j Composition; each of the pairing probability m i,j With a set of codons encoding c i and the anticodon encodes ac j correspond;

[0020] The feature fusion module is used to perform dimensionality reduction processing on the pairing probability matrix M to obtain the corresponding pairing probability vector V M ; and the pairing probability vector V M , performing vector concatenation of the concentration prediction vector P and the experimental condition encoding vector B to obtain a corresponding concatenated vector R and sending it to the third MLP model;

[0021] The third MLP model is used to predict the coupling coefficient of each group of codon-anticodon type pairs according to the splicing vector R to obtain the corresponding coupling coefficient tensor S and send it to the adaptation index weight estimation module; the shape of the coupling coefficient tensor S is N C ×N AC , by N C ×N AC Coupling coefficient s i,j Composition; each of the coupling coefficients s i,j With a set of codons encoding c i and the anticodon encodes ac j correspond;

[0022] The adaptation index weight estimation module is used to estimate the adaptation index weight of each codon type according to the tRNA gene copy number vector C and the coupling coefficient tensor S to obtain the corresponding adaptation index weight vector W; the adaptation index weight vector W is composed of N C The adaptation index weight w i Composition; the adaptation index weight w i The codon encoding c i One-to-one correspondence; the adaptation index weight w i The estimation formula is:

[0023] Preferably, the transcription efficiency prediction model is used to predict the transcription efficiency of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding transcription efficiency TPE;

[0024] The target gene codon index set G consists of codon indexes corresponding to all codon types of the specified target gene;

[0025] The transcription efficiency prediction model includes a first sub-screening module and a fourth MLP model; the first sub-screening module input end is connected to the input end of the transcription efficiency prediction model, and the output end is connected to the input end of the fourth MLP model; the output end of the fourth MLP model is connected to the output end of the transcription efficiency prediction model;

[0026] The first screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G. i The first weight vector corresponding to the extracted value is sent to the fourth MLP model; the fourth MLP model is used to predict the transcription efficiency according to the first weight vector to obtain the corresponding transcription efficiency TPE and output it;

[0027] The translation efficiency prediction model is used to predict the translation efficiency of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding translation efficiency TLE;

[0028] The translation efficiency prediction model includes a second sub-screening module and a fifth MLP model; the second sub-screening module input end is connected to the input end of the translation efficiency prediction model, and the output end is connected to the input end of the fifth MLP model; the output end of the fifth MLP model is connected to the output end of the translation efficiency prediction model;

[0029] The second screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G. i The extracted second weight vector is sent to the fifth MLP model; the fifth MLP model is used to predict the translation efficiency according to the second weight vector to obtain the corresponding gene translation efficiency TLE and output it;

[0030] The gene abundance prediction model is used to predict the gene abundance of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding gene abundance TGA;

[0031] The gene abundance prediction model includes a third screening module and a sixth MLP model; the input end of the third screening module is connected to the input end of the gene abundance prediction model, and the output end is connected to the input end of the sixth MLP model; the output end of the sixth MLP model is connected to the output end of the gene abundance prediction model;

[0032] The third screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G. i The extracted corresponding third weight vector is sent to the sixth MLP model; the sixth MLP model is used to predict gene abundance according to the third weight vector to obtain the corresponding gene abundance TGA and output it.

[0033] Preferably, the first data set includes a plurality of first data records;

[0034] Each of the first data records corresponds to a genome sequence of a type of cells of the target species observed manually or by a machine under set conditions by an observer;

[0035] The first data record includes a first genome sequence, a first set condition, a first codon data table, a first anticodon data table, and a first gene expression data table;

[0036] The data format of the first genome sequence is a universal genome sequence, genome map or gene expression spectrum;

[0037] The first set condition is the observation / experimental set condition corresponding to the current genome sequence, including at least temperature, humidity, light, nutritional conditions, and physiological / cultivation stage;

[0038] The first codon data table, the first anticodon data table and the first gene expression data table are all analysis and calculation data obtained by an observer using a preset bioinformatics tool to perform targeted analysis and calculation on the first genome sequence;

[0039] Wherein, the bioinformatics tools at least include tRNAscan-SE software, Galaxy software, SAMtools software, and GATK software;

[0040] The first codon data table includes a plurality of first codon data; each of the first codon data corresponds to a codon type in the first genome sequence; the first codon data includes a codon identifier, a codon triplet base encoding and a codon usage deviation; the codon identifier is a unique identifier of the current codon type; the codon triplet base encoding is a triplet base encoding of the current codon type; the codon usage deviation is a usage deviation scalar of the current codon type in the first genome sequence;

[0041] The first anticodon data table includes a plurality of first anticodon data; each of the first anticodon data corresponds to an anticodon type in the first genome sequence; the first anticodon data includes an anticodon identifier, an anticodon triplet base code and a tRNA gene copy number; the anticodon identifier is a unique identifier of the current anticodon type; the anticodon triplet base code is a triplet base code of the current anticodon type; the tRNA gene copy number is the total number of tRNA genes corresponding to the current anticodon type in the first genome sequence;

[0042] The first gene expression data table includes multiple first gene expression data; each of the first gene expression data corresponds to a DNA gene or a synthetic gene in the first genome sequence; the first gene expression data includes a gene identifier, a gene type, a gene base coding sequence, a gene codon sequence, a gene transcription efficiency, a gene translation efficiency and a gene abundance; the gene identifier is a unique identifier of the current gene; the gene type includes at least a DNA gene and a synthetic gene; the gene base coding sequence is a base coding sequence of the current gene; the gene codon sequence is composed of the codon identifiers of all codon types of the current gene; the gene transcription efficiency, the gene translation efficiency and the gene abundance are the transcription efficiency scalar, the translation efficiency scalar and the abundance scalar of the current gene under the first set conditions.

[0043] Preferably, the performing model training on the adaptation index weight prediction model and three downstream task models based on the first data set specifically includes:

[0044] Step 501, preparing training data based on the first data set to obtain a corresponding second data set; extracting the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector and the first codon usage deviation vector of each second data record in the second data set to form a corresponding third data record; deduplicating all the obtained third data records; and forming a corresponding third data set from all the deduplicated third data records;

[0045] Wherein, the second data set includes a plurality of second data records; the second data record includes a first training genome sequence encoding vector, a first training experimental condition encoding vector, a first training gene copy number vector, a first codon usage deviation vector, a first training codon index set, a first tag transcription efficiency, a first tag translation efficiency and a first tag gene abundance; the third data set is composed of a plurality of the third data records, each of which is composed of the corresponding first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector and the first codon usage deviation vector;

[0046] Step 502: randomly split the third data set based on a preset first split ratio to obtain a corresponding first training set and a first evaluation set;

[0047] Wherein, both the first training set and the first evaluation set are composed of a plurality of the third 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;

[0048] Step 503, taking the first third data record of the first training set as the corresponding current training record;

[0049] Step 504, inputting the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current training record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain a corresponding first predicted adaptation index weight vector;

[0050] Step 505, bringing the first predicted adaptation index weight vector and the first codon usage deviation vector of the current training record into a preset first correlation function; and performing a round of parameter modulation on the adaptation index weight prediction model in a direction that makes the first correlation function reach a maximum positive correlation based on a preset first model optimization algorithm; and at the end of this round of parameter modulation, identifying whether the current training record is the last third data record of the first training set; if so, go to step 506; if not, extracting the next third data record of the first training set as the new current training record and returning to step 504 to continue training;

[0051] Wherein, the first correlation function includes at least Pearson correlation function, Spearman correlation function, and Kendall correlation function; the first model optimization algorithm includes at least hill climbing algorithm and genetic algorithm;

[0052] Step 506, perform a round of traversal on all the third data records of the first evaluation set; and in this round of traversal, use the third data record currently traversed as the corresponding current evaluation record; and input the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current evaluation record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain the corresponding second predicted adaptation index weight vector; and bring the second predicted adaptation index weight vector and the first codon usage deviation vector of the current training record into the first correlation function to calculate and obtain the corresponding first correlation; and at the end of this round of traversal, bring all the obtained first correlations into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value;

[0053] Wherein, the first model evaluation function at least includes a MAE function, an MSE function, and a RMSE function;

[0054] Step 507, identifying whether the first evaluation value satisfies a preset first evaluation value range; if the first evaluation value satisfies the first evaluation value range, identifying whether all the first correlations obtained satisfy the preset first correlation range, if not, returning to step 501 to continue training, if yes, solidifying the model parameters of the adaptive index weight prediction model and going to step 508; if the first evaluation value does not satisfy the first evaluation value range, returning to step 501 to continue training;

[0055] Step 508, randomly splitting the second data set based on a preset second split ratio to obtain a corresponding second training set and a second evaluation set;

[0056] Wherein, both the second training set and the second evaluation set are composed of a plurality of the second data records; the ratio of the total number of records in the second training set to the total number of records in the second evaluation set satisfies the second segmentation ratio;

[0057] Step 509, taking the first second data record of the second training set as the corresponding current training record;

[0058] Step 510, inputting the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current training record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain a corresponding third predicted adaptation index weight vector;

[0059] Step 511, inputting the third predicted adaptation index weight vector and the first training codon index set of the current training record into the transcription efficiency prediction model, the translation efficiency prediction model and the gene abundance prediction model respectively to predict and obtain the corresponding first predicted transcription efficiency, first predicted translation efficiency and first predicted gene abundance;

[0060] Step 512: bring the first predicted transcription efficiency and the first label transcription efficiency of the current training record into a preset first transcription loss function to calculate a corresponding first loss value; bring the first predicted translation efficiency and the first label translation efficiency of the current training record into a preset first translation loss function to calculate a corresponding second loss value; bring the first predicted gene abundance and the first label gene abundance of the current training record into a preset first abundance loss function to calculate a corresponding third loss value;

[0061] Wherein, the first transcription loss function, the first translation loss function and the first abundance loss function all include at least an L1 loss function and an L2 loss function;

[0062] Step 513, identifying whether the first, second and third loss values ​​satisfy the preset first, second and third loss value ranges; if the first, second and third loss values ​​all satisfy the corresponding first, second and third loss value ranges, go to step 514; if the first loss value does not satisfy the first loss value range, a round of parameter modulation is performed on the transcription efficiency prediction model in a direction in which the function value of the first transcription loss function reaches a minimum value based on the preset first model optimizer, and the method returns to step 511 at the end of this round of parameter modulation; if the second loss value does not satisfy the second loss value range, a round of parameter modulation is performed on the translation efficiency prediction model in a direction in which the function value of the first translation loss function reaches a minimum value based on the preset second model optimizer, and the method returns to step 511 at the end of this round of parameter modulation; if the third loss value does not satisfy the third loss value range, a round of parameter modulation is performed on the gene abundance prediction model in a direction in which the function value of the first abundance loss function reaches a minimum value based on the preset third model optimizer, and the method returns to step 511 at the end of this round of parameter modulation;

[0063] Wherein, the first, second and third model optimizers all include at least an SGD series optimizer and an ADAM series optimizer;

[0064] Step 514, perform a round of traversal on all the second data records of the second evaluation set; and in this round of traversal, use the second data record currently traversed as the corresponding current evaluation record; and input the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current evaluation record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain the corresponding fourth predicted adaptation index weight vector; and input the fourth predicted adaptation index weight vector and the first training codon index set of the current evaluation record into the transcription efficiency prediction model, the translation efficiency prediction model and the The gene abundance prediction model is used to predict the corresponding second predicted transcription efficiency, the second predicted translation efficiency and the second predicted gene abundance; and the second predicted transcription efficiency, the second predicted translation efficiency and the second predicted gene abundance form a corresponding first prediction vector, and the first label transcription efficiency, the first label translation efficiency and the first label gene abundance of the current evaluation record form a corresponding first label vector, and the first prediction vector and the first label vector form a corresponding first prediction-label pair; and at the end of this round of traversal, all the first prediction-label pairs obtained are brought into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value;

[0065] Wherein, the second model evaluation function at least includes a MAE function and a RMSE function;

[0066] Step 515, identifying whether the second evaluation value meets the preset second evaluation value range; if the second evaluation value meets the second evaluation value range, solidifying the model parameters of the transcription efficiency prediction model, the translation efficiency prediction model and the gene abundance prediction model and going to step 516; if the second evaluation value does not meet the second evaluation value range, returning to step 508 to continue training;

[0067] Step 516, stop model training and confirm that the model training of the adaptive index weight prediction model and the three downstream task models is completed.

[0068] Furthermore, the preparing training data based on the first data set to obtain a corresponding second data set specifically includes:

[0069] Step 61, taking each of the first data records of the first data set as a corresponding current data record;

[0070] Step 62, using the first set condition, the first codon data table, the first anti-codon data table and the first gene expression data table of the current data record as the corresponding current set condition, current codon data table, current anti-codon data table and current gene expression data table;

[0071] Step 63, extract all the codon triplet base codes in the current codon data table and sort them in sequence to form a corresponding first codon coding vector; extract all the anticodon triplet base codes in the current anticodon data table and sort them in sequence to form a corresponding first anticodon coding vector; and perform vector splicing on the first codon coding vector and the first anticodon coding vector to obtain a corresponding first training genome sequence coding vector;

[0072] Step 64, based on a preset condition coding rule, the temperature, humidity, light, nutritional condition, and physiological / cultivation stage in the currently set conditions are respectively encoded to obtain corresponding temperature codes, humidity codes, light codes, nutritional condition codes, and physiological / cultivation stage codes to form a corresponding first training experiment condition coding vector;

[0073] Step 65, extracting all the tRNA gene copy numbers in the current anticodon data table and sequentially sorting them to form a corresponding first training gene copy number vector;

[0074] Step 66, extracting all the codon usage deviations in the current codon data table and sequentially sorting them to form a corresponding first codon usage deviation vector;

[0075] Step 67, perform a round of traversal on all the first gene expression data in the current gene expression data table; and in this round of traversal, use the first gene expression data currently traversed as the corresponding current gene expression data; and use the gene codon sequence, gene transcription efficiency, gene translation efficiency and gene abundance of the current gene expression data as the corresponding first training codon index set, the first label transcription efficiency, the first label translation efficiency and the first label gene abundance; and form a corresponding second data record with the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector, the first codon usage deviation vector and the first training codon index set, the first label transcription efficiency, the first label translation efficiency and the first label gene abundance corresponding to the current gene expression data; and at the end of this round of traversal, form a first record set corresponding to the current data record with all the second data records obtained in this round of traversal;

[0076] Step 68: Merge the first record sets corresponding to all the first data records to obtain the corresponding second data sets.

[0077] Preferably, the preprocessing of the current experimental condition, the current genome sequence and the current target gene to obtain the corresponding current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set specifically includes:

[0078] Step 71, based on a preset condition coding rule, the temperature, humidity, light, nutritional condition, and physiological / cultivation stage in the current experimental condition are respectively encoded to obtain corresponding temperature codes, humidity codes, light codes, nutritional condition codes, and physiological / cultivation stage codes to form a corresponding current experimental condition coding vector;

[0079] Step 72, using a preset bioinformatics tool to analyze and count the codon types of all mRNA genes in the current genome sequence, and sequentially sorting all the codon types counted to obtain a corresponding first type sequence; and sequentially sorting the triplet base codes of all codon types in the first type sequence to obtain a corresponding current codon code vector;

[0080] Step 73, using the bioinformatics tool to analyze and count the anticodon types of all tRNA genes in the current genome sequence, and sequentially sorting all the statistical anticodon types to obtain a corresponding second type sequence; and sequentially sorting the triplet base codes of all anticodon types in the second type sequence to obtain a corresponding current anticodon coding vector;

[0081] Step 74, performing vector concatenation on the current codon encoding vector and the current anti-codon encoding vector to obtain the corresponding current genome sequence encoding vector;

[0082] Step 75, using the bioinformatics tool to perform statistical calculations on the total number of gene copies of each tRNA gene in the current genome sequence and using the calculation result as the corresponding tRNA gene copy number; and sequentially sorting all the obtained tRNA gene copy numbers to obtain the corresponding current tRNA gene copy number vector;

[0083] Step 76, using the bioinformatics tool to analyze and count all codon types of the current target gene to obtain a corresponding third type sequence; and extracting the sorting index of each codon type of the third type sequence in the first type sequence as a corresponding first codon index; and forming a corresponding codon index set of the current target gene from all the first codon indexes obtained;

[0084] Step 77, output the current experimental condition encoding vector, the current genome sequence encoding vector, the current tRNA gene copy number vector and the current target gene codon index set obtained this time as the processing result of this preprocessing.

[0085] A second aspect of an embodiment of the present invention provides a device for implementing the processing method of the tRNA adaptation index weight prediction model described in the first aspect, the device comprising: a model construction module, a model training module, a data preparation module and a model application module;

[0086] The model building module is used to build an adaptation index weight prediction model for processing tRNA adaptation index weight prediction tasks; and to design three downstream task models: a transcription efficiency prediction model, a translation efficiency prediction model, and a gene abundance prediction model;

[0087] The model training module is used to collect data on the genome sequences of the target species under different conditions and construct a first data set based on the collected data; and to perform model training on the adaptation index weight prediction model and three downstream task models based on the first data set;

[0088] The data preparation module is used to receive the current experimental conditions, current genome sequence and current target gene input by the user after the model training is completed; and pre-process the current experimental conditions, the current genome sequence and the current target gene to obtain the corresponding current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set; the current genome sequence is a genome sequence obtained by an observer through manual or machine observation / experimental means under the current experimental conditions for a type of cell of the target species; the current target gene is a DNA gene or synthetic gene in the current genome sequence;

[0089] The model application module is used to perform tRNA adaptation index weight prediction processing by the adaptation index weight prediction model according to the current genome sequence encoding vector, the current experimental condition encoding vector and the current tRNA gene copy number vector to obtain the corresponding current adaptation index weight vector; and three downstream task models respectively perform corresponding transcription efficiency, translation efficiency and gene abundance prediction processing according to the current adaptation index weight vector and the current target gene codon index set to obtain the corresponding current transcription efficiency, current translation efficiency and current gene abundance composition corresponding task reports to feedback to the user.

[0090] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0091] 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;

[0092] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0093] 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.

[0094] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for a tRNA adaptation index weight prediction model. As can be seen from the above content, the embodiment of the present invention constructs an adaptation index weight prediction model and three downstream task models (transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) for processing tRNA adaptation index weight prediction tasks; and collects data on the genome sequence of a certain type of target species under different conditions and constructs a data set based on the collected data to train the adaptation index weight prediction model and the three downstream task models; after the training, the adaptation index weight prediction model is used to predict the tRNA adaptation index weight of the current target species under the current given conditions, and the gene expression level (transcription efficiency, translation efficiency, gene abundance) of a given target gene (DNA gene, synthetic gene) is further predicted according to the tRNA adaptation index weight through the three downstream task models. The four types of models (adaptation index weight prediction model, transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) provided in the embodiments of the present invention are all end-to-end prediction models. Prediction of tRNA adaptation index weight and gene expression level based on these four types of models not only reduces processing complexity but also improves model versatility. The four types of models in the embodiments of the present invention do not need to change the model structure when switching target species, and only need to use the data set of the current species for one round of training. In this way, not only the species adaptability of the model is improved, but also the prediction stability of the model for each type of species is further improved through training. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 A schematic diagram of a processing method for a tRNA adaptation index weight prediction model provided in Example 1 of the present invention;

[0096] Figure 2 A module structure diagram of the adaptive index weight prediction model provided in the first embodiment of the present invention;

[0097] Figure 3 The module structure diagram of the transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model provided in Example 1 of the present invention;

[0098] Figure 4 A module structure diagram of a processing device for a tRNA adaptation index weight prediction model provided in Example 2 of the present invention;

[0099] Figure 5 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0100] 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.

[0101] Embodiment 1 of the present invention provides a processing method for a tRNA adaptation index weight prediction model, such as Figure 1 As shown in the schematic diagram of a processing method of a tRNA adaptation index weight prediction model provided in Example 1 of the present invention, the method mainly comprises the following steps:

[0102] Step 1: construct an adaptation index weight prediction model for processing the tRNA adaptation index weight prediction task; and design three downstream task models: transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model.

[0103] Here, the adaptation index weight prediction model of the embodiment of the present invention is used to perform tRNA adaptation index weight prediction processing based on the input genome sequence encoding vector A, experimental condition encoding vector B and tRNA gene copy number vector C and output the corresponding adaptation index weight vector W.

[0104] The above genome sequence encoding vector A is composed of codon encoding vector A C and anticodon encoding vector A AC spliced; among them, the codon encoding vector A C The codons of all codon types in the corresponding genomic sequence encode c i Sequentially sorted; anticodon encoding vector A AC The anticodons of all anticodon types in the corresponding genomic sequence encode ac j Sequentially arranged; codons encode c i The total number is recorded as N C , the anticodon encodes ac j The total number is recorded as N AC , 1≤i≤N C , 1≤j≤N AC ; Codon encoding c i and the anticodon encoding ac j Each is encoded as a triplet of bases.

[0105] The above-mentioned experimental condition coding vector B at least includes temperature coding, humidity coding, light coding, nutritional condition coding, and physiological / cultivation stage coding.

[0106] The above tRNA gene copy number vector C is composed of N ACGene copy number tGCN j Composition; Among them, the gene copy number tGCN j With anticodon encoding ac j One to one correspondence.

[0107] The above adaptation index weight vector W is composed of N C The adaptation index weight w i Composition; Among them, the adaptation index weight w i With codon encoding c i One to one correspondence.

[0108] like Figure 2 As shown in the module structure diagram of the adaptation index weight prediction model provided in Example 1 of the present invention, the first model input end, the second model input end and the third model input end of the adaptation index weight prediction model are respectively used to receive the corresponding genome sequence encoding vector A, the experimental condition encoding vector B and the tRNA gene copy number vector C, and the model output end is used to output the corresponding adaptation index weight vector W.

[0109] The model components of the adaptive index weight prediction model include: Uni-RNA model, first MLP model, second MLP model, feature fusion module, third MLP model and adaptive index weight estimation module. The connection relationship of each model component is: the input end of the Uni-RNA model is connected to the input end of the first model, and the output end is respectively connected to the input ends of the first and second MLP models; the output ends of the first and second MLP models are respectively connected to the second and third input ends of the feature fusion module; the first input end of the feature fusion module is connected to the input end of the second model, and the output end is connected to the input end of the third MLP model; the output end of the third MLP model is connected to the second input end of the adaptive index weight estimation module; the first input end of the adaptive index weight estimation module is connected to the input end of the third model, and the output end is connected to the model output end.

[0110] The functions of each model component of the adaptive exponential weighted prediction model are shown below.

[0111] 1) Uni-RNA model:

[0112] The Uni-RNA model of the embodiment of the present invention has completed pre-training; the Uni-RNA model is used to perform feature encoding on the genome sequence encoding vector A to obtain the corresponding encoding feature X and send it to the first and second MLP models.

[0113] It should be noted that the model structure, function and pre-training scheme of the Uni-RNA model of the embodiment of the present invention have been disclosed in the technical document "UNI-RNA: UNIVERSAL PRE-TRAINED MODELS REVOLUTIONIZE RNA RESEARCH". It can be seen from the technical document that the Uni-RNA model can be used to perform deep learning on the association relationship of each group of codon + codon coding pairs, codon + anti-codon coding pairs, and anti-codon + anti-codon coding pairs in a genome sequence coding vector A and output a corresponding coding feature, namely, coding feature X.

[0114] 2) First MLP model:

[0115] The first MLP model of the embodiment of the present invention is used to predict the intracellular concentration of tRNA genes corresponding to each anticodon type according to the encoding feature X, and obtain the corresponding concentration prediction vector P to send to the feature fusion module.

[0116] Here, the concentration prediction vector P is composed of N AC The predicted concentration p j Composition, predicted concentration p j With anticodon encoding ac j One to one correspondence.

[0117] 3) Second MLP model:

[0118] The second MLP model of the embodiment of the present invention is used to predict the pairing probability of each group of codon-anticodon type pairs according to the encoding feature X to obtain the corresponding pairing probability matrix M and send it to the feature fusion module.

[0119] Here, the shape of the pairing probability matrix M is N C ×N AC , by N C ×N AC The pairing probability m i,j Composition; each pairing probability m i,j With a set of codons encoding c i and the anticodon encoding ac j correspond.

[0120] 4) Feature fusion module:

[0121] The feature fusion module of the embodiment of the present invention is used to perform dimensionality reduction processing on the pairing probability matrix M to obtain the corresponding pairing probability vector V M ; and pairing probability vector V M , the concentration prediction vector P and the experimental condition encoding vector B are vector-concatenated to obtain the corresponding concatenated vector R and send it to the third MLP model.

[0122] 5) The third MLP model:

[0123] The third MLP model of the embodiment of the present invention is used to predict the coupling coefficient of each group of codon-anticodon type pairs according to the splicing vector R to obtain the corresponding coupling coefficient tensor S and send it to the adaptation index weight estimation module.

[0124] Here, the coupling coefficient tensor S has the shape of N C ×N AC , by N C ×N AC Coupling coefficient s i,j Composition; each coupling coefficient s i,j With a set of codons encoding c i and the anticodon encoding ac j The coupling coefficient s mentioned here is i,j It can be seen from the following that it is a calculation parameter used to calculate the adaptation index weight.

[0125] It should be noted that the splicing vector R incorporates three types of feature information that affect the coupling coefficient (intracellular concentration of tRNA genes, codon-anticodon pairing probability, and experimental conditions). The third MLP model can learn the mapping relationship between these three types of feature information and the coupling coefficient, and the resulting coupling coefficient will have better generalization and accuracy.

[0126] 6) Adaptation index weight estimation module:

[0127] The adaptation index weight estimation module of the embodiment of the present invention is used to estimate the adaptation index weight of each codon type according to the tRNA gene copy number vector C and the coupling coefficient tensor S to obtain the corresponding adaptation index weight vector W.

[0128] Here, the adaptation index weight vector W is composed of N C The adaptation index weight w i Composition; Adaptation index weight w i With codon encoding c i One-to-one correspondence; and the adaptation index weight w i The estimation formula is:

[0129] It can be seen from the functions of each model component of the above-mentioned adaptation index weight prediction model that the adaptation index weight prediction model is used to predict the adaptation index weights of all codons in a given genome sequence. To predict the gene expression level (such as transcription efficiency, translation efficiency, gene abundance, etc.) of a certain type of target gene (DNA gene, synthetic gene) in this given genome sequence, it is necessary to further analyze the three downstream task models (transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) of the embodiment of the present invention on the basis of the known adaptation index weight vector W and the known target gene.

[0130] The model structures of the three downstream task models of the embodiment of the present invention are as follows: Figure 3 The module structure diagram of the transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model provided in Example 1 of the present invention is shown below. A brief description of the three downstream task models is given below.

[0131] 1) Transcription efficiency prediction model:

[0132] like Figure 3 As shown, the transcription efficiency prediction model of the embodiment of the present invention is used to predict the transcription efficiency of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding transcription efficiency TPE.

[0133] The target gene codon index set G mentioned here consists of codon indices corresponding to all codon types of the specified target gene.

[0134] The model components of the transcription efficiency prediction model include: a first sub-screening module and a fourth MLP model. The connection relationship between the components is: the input end of the first sub-screening module is connected to the input end of the transcription efficiency prediction model, and the output end is connected to the input end of the fourth MLP model; the output end of the fourth MLP model is connected to the output end of the transcription efficiency prediction model. The component functions of each component are: the first sub-screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G i The extracted first weight vector corresponding to the first weight vector is sent to the fourth MLP model; the fourth MLP model is used to predict the transcription efficiency according to the first weight vector to obtain the corresponding transcription efficiency TPE and output it.

[0135] It should be noted that, as can be seen from the foregoing, the adaptation index weight vector W contains the adaptation index weights of all codons in the entire genome sequence; and when predicting the transcription efficiency of a certain type of target gene (DNA gene, synthetic gene), it is only necessary to make a prediction based on the codon adaptation index weights related to the current target gene. Therefore, the transcription efficiency prediction model of the embodiment of the present invention will first select the codon adaptation index weights related to the current target gene from the adaptation index weight vector W based on the first screening module to form a first weight vector, and then the fourth MLP model will predict the transcription efficiency based on the first weight vector.

[0136] 2) Translation efficiency prediction model:

[0137] like Figure 3 As shown, the translation efficiency prediction model of the embodiment of the present invention is used to predict the translation efficiency of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding translation efficiency TLE.

[0138] The model components of the translation efficiency prediction model include: a second screening module and a fifth MLP model. The connection relationship between the components is: the input end of the second screening module is connected to the input end of the translation efficiency prediction model, and the output end is connected to the input end of the fifth MLP model; the output end of the fifth MLP model is connected to the output end of the translation efficiency prediction model. The component functions of each component are: the second screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G i The extracted corresponding second weight vector is sent to the fifth MLP model; the fifth MLP model is used to predict the translation efficiency according to the second weight vector to obtain the corresponding gene translation efficiency TLE and output it.

[0139] Similar to the transcription efficiency prediction model, when predicting the translation efficiency of a certain type of target gene (DNA gene, synthetic gene), it is only necessary to make a prediction based on the codon adaptation index weight related to the current target gene. Therefore, the translation efficiency prediction model of the embodiment of the present invention will first select the codon adaptation index weight related to the current target gene from the adaptation index weight vector W based on the second screening module to form a second weight vector, and then the fifth MLP model will predict the translation efficiency based on the second weight vector.

[0140] 3) Gene abundance prediction model:

[0141] like Figure 3 As shown, the gene abundance prediction model of the embodiment of the present invention is used to predict the gene abundance of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding gene abundance TGA.

[0142] The model components of the gene abundance prediction model include: a third screening module and a sixth MLP model. The connection relationship between the components is: the input end of the third screening module is connected to the input end of the gene abundance prediction model, and the output end is connected to the input end of the sixth MLP model; the output end of the sixth MLP model is connected to the output end of the gene abundance prediction model. The component functions of each component are: the third screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G i The extracted corresponding third weight vector is sent to the sixth MLP model; the sixth MLP model is used to predict gene abundance according to the third weight vector to obtain the corresponding gene abundance TGA and output it.

[0143] Similar to the transcription efficiency prediction model, when predicting the gene abundance of a certain type of target gene (DNA gene, synthetic gene), it is only necessary to make a prediction based on the codon adaptation index weight related to the current target gene. Therefore, the gene abundance prediction model of the embodiment of the present invention will first select the codon adaptation index weight related to the current target gene from the adaptation index weight vector W based on the third screening module to form a third weight vector, and then the sixth MLP model will predict the gene abundance based on the third weight vector.

[0144] Step 2, collecting data on the genome sequences of the target species under different conditions and constructing a first data set based on the collected data; and training the adaptation index weight prediction model and three downstream task models based on the first data set;

[0145] Specifically, it includes: step 21, collecting data on the genome sequence of the target species under different conditions and constructing a first data set based on the collected data;

[0146] Here, the first data set constructed in the embodiment of the present invention corresponds to a type of target species; the first data set includes a plurality of first data records; each first data record corresponds to a genome sequence obtained by an observer manually or by a machine under a set condition for a type of cells of the corresponding target species; and each first data record is composed of the following data: a first genome sequence, a first set condition, a first codon data table, a first anticodon data table, and a first gene expression data table;

[0147] The data format of the first genome sequence is a universal genome sequence, genome map or gene expression spectrum; the first set condition is the observation / experimental set condition corresponding to the current genome sequence, including at least temperature, humidity, light, nutritional conditions, and physiological / culture stage; the first codon data table, the first anticodon data table and the first gene expression data table are all analysis and calculation data obtained by the observer using a preset bioinformatics tool to perform targeted analysis and calculation on the first genome sequence; the bioinformatics tools mentioned here include at least tRNAscan-SE software, Galaxy software, SAMtools software, and GATK software;

[0148] The following further describes the data contents of the first codon data table, the first anticodon data table, and the first gene expression data table:

[0149] 1) First codon data table:

[0150] The first codon data table includes a plurality of first codon data; each first codon data corresponds to a codon type in the first genome sequence; that is, the first codon data table covers all codon types in the first genome sequence;

[0151] The first codon data includes a codon identifier, a codon triplet base code, and a codon usage deviation; wherein the codon identifier is a unique identifier of the current codon type; the codon triplet base code is a triplet base code of the current codon type; and the codon usage deviation is a usage deviation scalar of the current codon type in the first genome sequence;

[0152] 2) First anticodon data table:

[0153] The first anticodon data table includes a plurality of first anticodon data; each first anticodon data corresponds to an anticodon type in the first genome sequence; that is, the first anticodon data table covers all anticodon types in the first genome sequence;

[0154] The first anticodon data includes an anticodon identifier, an anticodon triplet base code, and a tRNA gene copy number; wherein the anticodon identifier is a unique identifier of the current anticodon type; the anticodon triplet base code is a triplet base code of the current anticodon type; and the tRNA gene copy number is the total number of tRNA genes corresponding to the current anticodon type in the first genome sequence;

[0155] 3) The first gene expression data table:

[0156] The first gene expression data table includes a plurality of first gene expression data; each first gene expression data corresponds to a DNA gene or a synthetic gene in the first genome sequence; that is, the first gene expression data table covers all DNA genes and synthetic genes (if any synthetic genes exist) in the first genome sequence;

[0157] The first gene expression data includes a gene identifier, a gene type, a gene base coding sequence, a gene codon sequence, a gene transcription efficiency, a gene translation efficiency and a gene abundance; wherein the gene identifier is a unique identifier of the current gene; the gene type includes at least a DNA gene and a synthetic gene; the gene base coding sequence is a base coding sequence of the current gene; the gene codon sequence is composed of codon identifiers of all codon types of the current gene; the gene transcription efficiency, the gene translation efficiency and the gene abundance are the transcription efficiency scalar, the translation efficiency scalar and the abundance scalar of the current gene under the first set condition;

[0158] Step 22, and performing model training on the adaptation index weight prediction model and three downstream task models based on the first data set;

[0159] Specifically, it includes: step 2201, preparing training data based on the first data set to obtain a corresponding second data set; extracting the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector and the first codon usage deviation vector of each second data record in the second data set to form a corresponding third data record; and deduplicating all the obtained third data records; and forming a corresponding third data set from all the deduplicated third data records;

[0160] Specifically comprising: step 22011, preparing training data based on the first data set to obtain a corresponding second data set;

[0161] The second data set includes a plurality of second data records; the second data record includes a first training genome sequence encoding vector, a first training experimental condition encoding vector, a first training gene copy number vector, a first codon usage deviation vector, a first training codon index set, a first tag transcription efficiency, a first tag translation efficiency, and a first tag gene abundance;

[0162] Specifically comprising: step 220111, taking each first data record of the first data set as a corresponding current data record;

[0163] Step 220112, using the first setting condition, the first codon data table, the first anti-codon data table and the first gene expression data table of the current data record as the corresponding current setting condition, the current codon data table, the current anti-codon data table and the current gene expression data table;

[0164] Step 220113, extract all codon triplet base codes in the current codon data table and sort them in sequence to form a corresponding first codon code vector; extract all anticodon triplet base codes in the current anticodon data table and sort them in sequence to form a corresponding first anticodon code vector; and perform vector concatenation on the first codon code vector and the first anticodon code vector to obtain a corresponding first training genome sequence code vector;

[0165] Step 220114, based on the preset condition coding rules, the temperature, humidity, light, nutritional condition, and physiological / cultivation stage in the current set conditions are respectively encoded to obtain corresponding temperature codes, humidity codes, light codes, nutritional condition codes, and physiological / cultivation stage codes to form a corresponding first training experiment condition coding vector;

[0166] Here, the conditional coding rules are a set of pre-specified coding rules for data coding of conditional factors such as temperature, humidity, light, nutritional conditions, physiological / cultivation stages, etc. The adopted coding rules can be customized according to actual application requirements; for example, temperature, humidity, and light are encoded in the form of numerical or normalized numerical coding with positive and negative relationships, different classifications of nutritional conditions are encoded in the form of positive integer coding, and the degrees of different classifications are encoded in the form of decimal / percent coding, and the corresponding nutritional condition code is composed of the classification code + degree code, and different stages of the physiological / cultivation stage are encoded in the form of positive integer coding;

[0167] Step 220115, extract all tRNA gene copy numbers in the current anticodon data table and sort them in order to form a corresponding first training gene copy number vector;

[0168] Step 220116, extract all codon usage deviations in the current codon data table, sort them in order to form a corresponding first codon usage deviation vector;

[0169] Step 220117, perform a round of traversal on all the first gene expression data in the current gene expression data table; and in this round of traversal, use the currently traversed first gene expression data as the corresponding current gene expression data; and use the gene codon sequence, gene transcription efficiency, gene translation efficiency and gene abundance of the current gene expression data as the corresponding first training codon index set, first label transcription efficiency, first label translation efficiency and first label gene abundance; and form a corresponding second data record with the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector, the first codon usage deviation vector and the first training codon index set, first label transcription efficiency, first label translation efficiency and first label gene abundance corresponding to the current gene expression data; and at the end of this round of traversal, form a first record set corresponding to the current data record with all the second data records obtained in this round of traversal;

[0170] Step 220118, merging the first record sets corresponding to all the first data records to obtain the corresponding second data sets;

[0171] Here, through the above steps 22011 [220111-220118], it can be known that the first training codon index set, the first label transcription efficiency, the first label translation efficiency and the first label gene abundance of each second data record in the second data set are personalized, but the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector and the first codon usage deviation vector of all second data records generated by the same first data record are overlapped;

[0172] It should be noted that, as will be known below, the embodiment of the present invention will train the adaptation index weight prediction model based on the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector, and the first codon usage deviation vector of the second data record. If the adaptation index weight prediction model is trained with the second data set as the reference data set, repeated training scenarios with the same training data will inevitably occur multiple times, and the training cycle will also be relatively long. In order to shorten the training cycle and improve the training efficiency, the embodiment of the present invention hereby constructs a data set dedicated to training the adaptation index weight prediction model and without repeated records, i.e., a third data set, according to the second data set through the following steps 22012;

[0173] Step 22012, extract the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector and the first codon usage deviation vector of each second data record in the second data set to form a corresponding third data record; and perform deduplication on all the obtained third data records; and form a corresponding third data set from all the deduplicated third data records;

[0174] The third data set is composed of a plurality of third data records, each of which is composed of a corresponding first training genome sequence encoding vector, a first training experimental condition encoding vector, a first training gene copy number vector, and a first codon usage deviation vector;

[0175] Step 2202, randomly splitting the third data set based on a preset first split ratio to obtain a corresponding first training set and a first evaluation set;

[0176] The first segmentation ratio is a preset ratio value, for example, 9:1: the first training set and the first evaluation set are both composed of a plurality of third 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;

[0177] Step 2203, taking the first third data record of the first training set as the corresponding current training record;

[0178] Step 2204, inputting the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current training record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain the corresponding first predicted adaptation index weight vector;

[0179] Step 2205, bringing the first predicted adaptation index weight vector and the first codon usage deviation vector of the current training record into a preset first correlation function; and performing a round of parameter modulation on the adaptation index weight prediction model in a direction that makes the first correlation function reach a maximum positive correlation based on the preset first model optimization algorithm; and at the end of this round of parameter modulation, identifying whether the current training record is the last third data record of the first training set; if so, turning to step 2206; if not, extracting the next third data record of the first training set as a new current training record and returning to step 2204 to continue training;

[0180] Wherein, the first correlation function includes at least a Pearson correlation function, a Spearman correlation function, and a Kendall correlation function; the first model optimization algorithm includes at least a hill climbing algorithm and a genetic algorithm;

[0181] Step 2206, perform a round of traversal on all third data records of the first evaluation set; and in this round of traversal, use the third data record currently traversed as the corresponding current evaluation record; and input the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current evaluation record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain the corresponding second predicted adaptation index weight vector; and bring the second predicted adaptation index weight vector and the first codon usage deviation vector of the current training record into the first correlation function to calculate and obtain the corresponding first correlation; and at the end of this round of traversal, bring all the obtained first correlations into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value;

[0182] Wherein, the first model evaluation function includes at least a MAE function, an MSE function, and a RMSE function;

[0183] Step 2207, identifying whether the first evaluation value satisfies the preset first evaluation value range; if the first evaluation value satisfies the first evaluation value range, identifying whether all the obtained first correlations satisfy the preset first correlation range, if not, returning to step 2201 to continue training, if yes, solidifying the model parameters of the adaptive index weight prediction model and going to step 2208; if the first evaluation value does not satisfy the first evaluation value range, returning to step 2201 to continue training;

[0184] Wherein, the first evaluation value range is a preset evaluation value range; the first correlation range is a preset correlation range;

[0185] Step 2208, randomly splitting the second data set based on a preset second split ratio to obtain a corresponding second training set and a second evaluation set;

[0186] The second segmentation ratio is a preset ratio value, for example, 8:2: the second training set and the second evaluation set are both composed of a plurality of second data records; the ratio of the total number of records in the second training set to the total number of records in the second evaluation set satisfies the second segmentation ratio;

[0187] Step 2209, taking the first second data record of the second training set as the corresponding current training record;

[0188] Step 2210, inputting the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current training record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain a corresponding third predicted adaptation index weight vector;

[0189] Step 2211, inputting the third prediction adaptation index weight vector and the first training codon index set of the current training record into the transcription efficiency prediction model, the translation efficiency prediction model and the gene abundance prediction model respectively to predict and obtain the corresponding first predicted transcription efficiency, first predicted translation efficiency and first predicted gene abundance;

[0190] Step 2212, bringing the first predicted transcription efficiency and the first label transcription efficiency of the current training record into a preset first transcription loss function to calculate a corresponding first loss value; bringing the first predicted translation efficiency and the first label translation efficiency of the current training record into a preset first translation loss function to calculate a corresponding second loss value; bringing the first predicted gene abundance and the first label gene abundance of the current training record into a preset first abundance loss function to calculate a corresponding third loss value;

[0191] Wherein, the first transcription loss function, the first translation loss function and the first abundance loss function all include at least an L1 loss function and an L2 loss function;

[0192] Step 2213, identifying whether the first, second and third loss values ​​satisfy the preset first, second and third loss value ranges; if the first, second and third loss values ​​all satisfy the corresponding first, second and third loss value ranges, go to step 2214; if the first loss value does not satisfy the first loss value range, a round of parameter modulation is performed on the transcription efficiency prediction model in the direction of minimizing the function value of the first transcription loss function based on the preset first model optimizer, and the method returns to step 2211 at the end of this round of parameter modulation; if the second loss value does not satisfy the second loss value range, a round of parameter modulation is performed on the translation efficiency prediction model in the direction of minimizing the function value of the first translation loss function based on the preset second model optimizer, and the method returns to step 2211 at the end of this round of parameter modulation; if the third loss value does not satisfy the third loss value range, a round of parameter modulation is performed on the gene abundance prediction model in the direction of minimizing the function value of the first abundance loss function based on the preset third model optimizer, and the method returns to step 2211 at the end of this round of parameter modulation;

[0193] The first, second and third loss value ranges are three preset loss value ranges; the first, second and third model optimizers all include at least an SGD series optimizer and an ADAM series optimizer;

[0194] Step 2214, perform a round of traversal on all second data records of the second evaluation set; and in this round of traversal, use the second data record currently traversed as the corresponding current evaluation record; and input the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current evaluation record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain the corresponding fourth predicted adaptation index weight vector; and input the fourth predicted adaptation index weight vector and the first training codon index set of the current evaluation record into the transcription efficiency prediction model, the translation efficiency prediction model and the gene abundance prediction model respectively to predict the corresponding second predicted transcription efficiency, the second predicted translation efficiency and the second predicted gene abundance; and form a corresponding first prediction vector from the second predicted transcription efficiency, the second predicted translation efficiency and the second predicted gene abundance, and form a corresponding first label vector from the first label transcription efficiency, the first label translation efficiency and the first label gene abundance of the current evaluation record, and form a corresponding first prediction-label pair from the first prediction vector and the first label vector; and at the end of this round of traversal, bring all the obtained first prediction-label pairs into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value;

[0195] Wherein, the second model evaluation function includes at least a MAE function and a RMSE function;

[0196] Step 2215, identifying whether the second evaluation value meets the preset second evaluation value range; if the second evaluation value meets the second evaluation value range, solidifying the model parameters of the transcription efficiency prediction model, the translation efficiency prediction model and the gene abundance prediction model and going to step 2216; if the second evaluation value does not meet the second evaluation value range, returning to step 2208 to continue training;

[0197] Wherein, the second evaluation value range is a preset evaluation value range;

[0198] Step 2216, stop model training and confirm that the model training of the adaptive index weight prediction model and the three downstream task models is completed.

[0199] Step 3: After the model training is completed, the current experimental conditions, current genome sequence and current target gene input by the user are received; and the current experimental conditions, current genome sequence and current target gene are preprocessed to obtain the corresponding current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set;

[0200] Specifically, it includes: step 31, after the model training is completed, receiving the current experimental conditions, the current genome sequence and the current target gene input by the user;

[0201] Here, the current genome sequence is the genome sequence of a type of cell of the target species corresponding to the most recent model training obtained by the observer through manual or machine observation / experimental means under the current experimental conditions; the current target gene is a DNA gene or synthetic gene in the current genome sequence;

[0202] Step 32, preprocessing the current experimental condition, the current genome sequence and the current target gene to obtain the corresponding current experimental condition encoding vector, the current genome sequence encoding vector, the current tRNA gene copy number vector and the current target gene codon index set;

[0203] Specifically, step 321, based on the conditional coding rule, the temperature, humidity, light, nutritional condition, and physiological / cultivation stage in the current experimental conditions are respectively encoded to obtain corresponding temperature codes, humidity codes, light codes, nutritional condition codes, and physiological / cultivation stage codes to form a corresponding current experimental condition coding vector;

[0204] Step 322, using bioinformatics tools to analyze and count the codon types of all mRNA genes in the current genome sequence, and sequentially sorting all the codon types counted to obtain a corresponding first type sequence; and sequentially sorting the triplet base codes of all codon types in the first type sequence to obtain a corresponding current codon code vector;

[0205] Step 323, using bioinformatics tools to analyze and count the anticodon types of all tRNA genes in the current genome sequence, and sequentially sorting all the statistically obtained anticodon types to obtain a corresponding second type sequence; and sequentially sorting the triplet base codes of all anticodon types in the second type sequence to obtain a corresponding current anticodon code vector;

[0206] Step 324, performing vector concatenation on the current codon encoding vector and the current anti-codon encoding vector to obtain a corresponding current genome sequence encoding vector;

[0207] Step 325, using bioinformatics tools to perform statistical calculations on the total number of gene copies of each tRNA gene in the current genome sequence and using the calculation results as the corresponding tRNA gene copy number; and sequentially sorting all the obtained tRNA gene copy numbers to obtain the corresponding current tRNA gene copy number vector;

[0208] Step 326, using bioinformatics tools to analyze and count all codon types of the current target gene to obtain a corresponding third type sequence; and extracting the sorting index of each codon type of the third type sequence in the first type sequence as a corresponding first codon index; and forming a corresponding current target gene codon index set from all the obtained first codon indexes;

[0209] Step 327, output the current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set obtained this time as the processing result of this preprocessing.

[0210] Step 4, the adaptation index weight prediction model performs tRNA adaptation index weight prediction processing according to the current genome sequence encoding vector, the current experimental condition encoding vector and the current tRNA gene copy number vector to obtain the corresponding current adaptation index weight vector; and the three downstream task models respectively perform corresponding transcription efficiency, translation efficiency and gene abundance prediction processing according to the current adaptation index weight vector and the current target gene codon index set to obtain the corresponding current transcription efficiency, current translation efficiency and current gene abundance composition corresponding task report feedback to the user.

[0211] Finally, it should be noted that the four types of models in the embodiments of the present invention can be expanded into a variety of model application scenarios. The following is a description of three typical scenarios.

[0212] Scenario 1: Under specified experimental conditions (i.e., the experimental condition encoding vector remains unchanged), the genome sequence encoding vector, tRNA gene copy number vector and target gene codon index set corresponding to each DNA / synthetic gene of a genome sequence are identified, and multiple groups of gene identification data (experimental condition encoding vector + genome sequence encoding vector + tRNA gene copy number vector + target gene codon index set) are obtained. Then, the experimental condition encoding vector + genome sequence encoding vector + tRNA gene copy number vector of each group of gene identification data is input into the adaptation index weight prediction model to obtain the adaptation index weight vector of each DNA / synthetic gene; then, the adaptation index weight vector of each DNA / synthetic gene is brought into the fixed gene-level tRNA adaptation index calculation formula to obtain the tRNA adaptation index of each DNA / synthetic gene; then, the adaptation index weight vector of each DNA / synthetic gene + target gene codon index set is brought into the three downstream task models (transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) to obtain the transcription efficiency, translation efficiency and gene abundance of each DNA / synthetic gene. In this way, the gene expression levels of different genes in a genome sequence can be compared horizontally from multiple data dimensions (tRNA adaptation index, transcription efficiency, translation efficiency, gene abundance) based on the predicted data.

[0213] Scenario 2: Switch the experimental conditions and obtain a corresponding genome sequence under different experimental conditions; and through operations similar to scenario 1, predict the multi-dimensional gene expression data (tRNA adaptation index, transcription efficiency, translation efficiency, gene abundance) of each DNA / synthetic gene under each experimental condition; thus, based on the predicted data, the gene expression levels of each DNA / synthetic gene under different experimental conditions can be longitudinally compared.

[0214] Scenario 3: Under certain specified experimental conditions, the multi-dimensional gene expression data (tRNA adaptation index, transcription efficiency, translation efficiency, gene abundance) of each DNA / synthetic gene under each experimental condition is predicted through similar operations as Scenario 1; and under the specified experimental conditions, relevant experimental detection data (such as transcription efficiency, translation efficiency, gene abundance, etc.) are obtained through experimental means, and the accuracy of the experimental detection data is ensured through other experimental or computational means; the accuracy of the four types of models is then judged by comparing the predictions of each gene expression data item (transcription efficiency, translation efficiency, gene abundance, etc.) with the experimental data; and when the accuracy decreases or does not meet the standard, the parameters of some or all of the four types of models that cause the prediction to fail to meet the standard are tuned through the model parameter optimizer with reference to the known prediction-experimental data pairs and in accordance with the model training processing method; then, the prediction data of the four types of models after the tuning can be used to evaluate the accuracy of other experimental data whose accuracy has not been verified.

[0215] Figure 4 The module structure diagram of a processing device for a tRNA adaptation index weight prediction model provided in the second embodiment of the present invention 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 4 As shown, the device includes: a model building module 201, a model training module 202, a data preparation module 203 and a model application module 204.

[0216] The model building module 201 is used to build an adaptation index weight prediction model for processing the tRNA adaptation index weight prediction task; and to design three downstream task models: a transcription efficiency prediction model, a translation efficiency prediction model and a gene abundance prediction model.

[0217] The model training module 202 is used to collect data on the genome sequences of the target species under different conditions and construct a first data set based on the collected data; and to perform model training on the adaptation index weight prediction model and three downstream task models based on the first data set.

[0218] The data preparation module 203 is used to receive the current experimental conditions, current genome sequence and current target gene input by the user after the model training is completed; and pre-process the current experimental conditions, current genome sequence and current target gene to obtain the corresponding current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set; the current genome sequence is a genome sequence obtained by an observer through manual or machine observation / experimental means under the current experimental conditions for a type of cell of the target species; the current target gene is a DNA gene or synthetic gene in the current genome sequence.

[0219] The model application module 204 is used to perform tRNA adaptation index weight prediction processing according to the current genome sequence encoding vector, the current experimental condition encoding vector and the current tRNA gene copy number vector by the adaptation index weight prediction model to obtain the corresponding current adaptation index weight vector; and the three downstream task models respectively perform corresponding transcription efficiency, translation efficiency and gene abundance prediction processing according to the current adaptation index weight vector and the current target gene codon index set to obtain the corresponding current transcription efficiency, current translation efficiency and current gene abundance composition corresponding task report feedback to the user.

[0220] A processing device for a tRNA adaptation index weight prediction model provided in an embodiment of the present invention can execute the method steps in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0221] 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.

[0222] 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).

[0223] 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.

[0224] Figure 5 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 5As 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.

[0225] exist Figure 5 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.

[0226] 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.

[0227] 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.

[0228] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for a tRNA adaptation index weight prediction model. As can be seen from the above content, the embodiment of the present invention constructs an adaptation index weight prediction model and three downstream task models (transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) for processing tRNA adaptation index weight prediction tasks; and collects data on the genome sequence of a certain type of target species under different conditions and constructs a data set based on the collected data to train the adaptation index weight prediction model and the three downstream task models; after the training, the adaptation index weight prediction model is used to predict the tRNA adaptation index weight of the current target species under the current given conditions, and the gene expression level (transcription efficiency, translation efficiency, gene abundance) of a given target gene (DNA gene, synthetic gene) is further predicted according to the tRNA adaptation index weight through the three downstream task models. The four types of models (adaptation index weight prediction model, transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model) provided in the embodiments of the present invention are all end-to-end prediction models. Prediction of tRNA adaptation index weight and gene expression level based on these four types of models not only reduces processing complexity but also improves model versatility. The four types of models in the embodiments of the present invention do not need to change the model structure when switching target species, and only need to use the data set of the current species for one round of training. In this way, not only the species adaptability of the model is improved, but also the prediction stability of the model for each type of species is further improved through training.

[0229] 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.

[0230] 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 processing method for a tRNA adaptation index weight prediction model, characterized in that: The method comprises: Construct an adaptation index weight prediction model for tRNA adaptation index weight prediction tasks; and design three downstream task models: transcription efficiency prediction model, translation efficiency prediction model and gene abundance prediction model; Collecting data on the genome sequences of the target species under different conditions and constructing a first data set based on the collected data; and training the adaptation index weight prediction model and three downstream task models based on the first data set; After the model training is completed, the current experimental conditions, current genome sequence and current target gene input by the user are received; and the current experimental conditions, the current genome sequence and the current target gene are preprocessed to obtain the corresponding current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set; the current genome sequence is a genome sequence of a type of cell of the target species obtained by an observer through manual or machine observation / experimental means under the current experimental conditions; the current target gene is a DNA gene or synthetic gene in the current genome sequence; The adaptation index weight prediction model performs tRNA adaptation index weight prediction processing according to the current genome sequence encoding vector, the current experimental condition encoding vector and the current tRNA gene copy number vector to obtain the corresponding current adaptation index weight vector; and three downstream task models respectively perform corresponding transcription efficiency, translation efficiency and gene abundance prediction processing according to the current adaptation index weight vector and the current target gene codon index set to obtain the corresponding current transcription efficiency, current translation efficiency and current gene abundance to form a corresponding task report, which is fed back to the user.

2. The processing method of the tRNA adaptation index weight prediction model according to claim 1, characterized in that: The adaptation index weight prediction model is used to perform tRNA adaptation index weight prediction processing according to the input genome sequence encoding vector A, experimental condition encoding vector B and tRNA gene copy number vector C and output the corresponding adaptation index weight vector W; The genome sequence encoding vector A is composed of the codon encoding vector A C and anticodon encoding vector A AC spliced; the codon encoding vector A C The codons of all codon types in the corresponding genomic sequence encode c i The anti-codon encoding vector A is arranged in sequence; AC The anticodons of all anticodon types in the corresponding genomic sequence encode ac j The codons are arranged in sequence; the codons encode c i The total number is recorded as N C , the anticodon encodes ac j The total number is recorded as N AC , 1≤i≤N C , 1≤j≤N AC ; The codon encodes c i and the anticodon encodes ac j Each is a triplet base code; The experimental condition coding vector B at least includes temperature coding, humidity coding, light coding, nutritional condition coding, and physiological / culture stage coding; The tRNA gene copy number vector C is composed of N AC Gene copy number tGCN j Composition; the gene copy number tGCN j The anticodon encodes ac j One to one correspondence; The adaptation index weight vector W is composed of N C The adaptation index weight w i Composition; the adaptation index weight w i The codon encoding c i One to one correspondence; The first model input end, the second model input end and the third model input end of the adaptation index weight prediction model are respectively used to receive the corresponding genome sequence encoding vector A, the experimental condition encoding vector B and the tRNA gene copy number vector C, and the model output end is used to output the corresponding adaptation index weight vector W; The adaptation index weight prediction model includes a Uni-RNA model, a first MLP model, a second MLP model, a feature fusion module, a third MLP model and an adaptation index weight estimation module; The input end of the Uni-RNA model is connected to the input end of the first model, and the output end is connected to the input ends of the first and second MLP models respectively; the output ends of the first and second MLP models are connected to the second and third input ends of the feature fusion module respectively; the first input end of the feature fusion module is connected to the input end of the second model, and the output end is connected to the input end of the third MLP model; the output end of the third MLP model is connected to the second input end of the adaptation index weight estimation module; the first input end of the adaptation index weight estimation module is connected to the input end of the third model, and the output end is connected to the output end of the model; The Uni-RNA model has completed pre-training; the Uni-RNA model is used to perform feature encoding on the genome sequence encoding vector A to obtain corresponding encoding features X and send them to the first and second MLP models; The first MLP model is used to predict the intracellular concentration of the tRNA gene corresponding to each anticodon type according to the coding feature X to obtain the corresponding concentration prediction vector P and send it to the feature fusion module; the concentration prediction vector P is composed of N AC The predicted concentration p j composition, the predicted concentration p j The anticodon encodes ac j One to one correspondence; The second MLP model is used to predict the pairing probability of each group of codon-anticodon type pairs according to the coding feature X to obtain the corresponding pairing probability matrix M and send it to the feature fusion module; the shape of the pairing probability matrix M is N C ×N AC , by N C ×N AC The pairing probability m i,j Composition; each pairing probability m i,j With a set of codons encoding c i and the anticodon encodes ac j correspond; The feature fusion module is used to perform dimensionality reduction processing on the pairing probability matrix M to obtain the corresponding pairing probability vector V M ; and the pairing probability vector V M , performing vector concatenation of the concentration prediction vector P and the experimental condition encoding vector B to obtain a corresponding concatenated vector R and sending it to the third MLP model; The third MLP model is used to predict the coupling coefficient of each group of codon-anticodon type pairs according to the splicing vector R to obtain the corresponding coupling coefficient tensor S and send it to the adaptation index weight estimation module; the shape of the coupling coefficient tensor S is N C ×N AC , by N C ×N AC Coupling coefficient s i,j Composition; each of the coupling coefficients s i,j With a set of codons encoding c i and the anticodon encodes ac j correspond; The adaptation index weight estimation module is used to estimate the adaptation index weight of each codon type according to the tRNA gene copy number vector C and the coupling coefficient tensor S to obtain the corresponding adaptation index weight vector W; the adaptation index weight vector W is composed of N C The adaptation index weight w i Composition; the adaptation index weight w i The codon encoding c i One-to-one correspondence; the adaptation index weight w i The estimation formula is:

3. The processing method of the tRNA adaptation index weight prediction model according to claim 2, characterized in that: The transcription efficiency prediction model is used to predict the transcription efficiency of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding transcription efficiency TPE; The target gene codon index set G consists of codon indexes corresponding to all codon types of the specified target gene; The transcription efficiency prediction model includes a first sub-screening module and a fourth MLP model; the first sub-screening module input end is connected to the input end of the transcription efficiency prediction model, and the output end is connected to the input end of the fourth MLP model; the output end of the fourth MLP model is connected to the output end of the transcription efficiency prediction model; The first screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G. i The first weight vector corresponding to the extracted value is sent to the fourth MLP model; the fourth MLP model is used to predict the transcription efficiency according to the first weight vector to obtain the corresponding transcription efficiency TPE and output it; The translation efficiency prediction model is used to predict the translation efficiency of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding translation efficiency TLE; The translation efficiency prediction model includes a second sub-screening module and a fifth MLP model; the second sub-screening module input end is connected to the input end of the translation efficiency prediction model, and the output end is connected to the input end of the fifth MLP model; The output end of the fifth MLP model is connected to the output end of the translation efficiency prediction model; The second screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G. i The extracted second weight vector is sent to the fifth MLP model; The fifth MLP model is used to predict the translation efficiency according to the second weight vector to obtain the corresponding gene translation efficiency TLE and output it; The gene abundance prediction model is used to predict the gene abundance of the specified target gene according to the input adaptation index weight vector W and the target gene codon index set G and output the corresponding gene abundance TGA; The gene abundance prediction model includes a third screening module and a sixth MLP model; the input end of the third screening module is connected to the input end of the gene abundance prediction model, and the output end is connected to the input end of the sixth MLP model; the output end of the sixth MLP model is connected to the output end of the gene abundance prediction model; The third screening module is used to match the index i in the adaptation index weight vector W with the adaptation index weight w of each codon index of the target gene codon index set G. i The extracted corresponding third weight vector is sent to the sixth MLP model; the sixth MLP model is used to predict gene abundance according to the third weight vector to obtain the corresponding gene abundance TGA and output it.

4. The processing method of the tRNA adaptation index weight prediction model according to claim 1, characterized in that: The first data set includes a plurality of first data records; Each of the first data records corresponds to a genome sequence of a type of cells of the target species observed manually or by a machine under set conditions by an observer; The first data record includes a first genome sequence, a first set condition, a first codon data table, a first anticodon data table, and a first gene expression data table; The data format of the first genome sequence is a universal genome sequence, genome map or gene expression spectrum; The first set condition is the observation / experimental set condition corresponding to the current genome sequence, including at least temperature, humidity, light, nutritional conditions, and physiological / cultivation stage; The first codon data table, the first anticodon data table and the first gene expression data table are all analysis and calculation data obtained by an observer using a preset bioinformatics tool to perform targeted analysis and calculation on the first genome sequence; Wherein, the bioinformatics tools at least include tRNAscan-SE software, Galaxy software, SAMtools software, and GATK software; The first codon data table includes a plurality of first codon data; each of the first codon data corresponds to a codon type in the first genome sequence; the first codon data includes a codon identifier, a codon triplet base encoding and a codon usage deviation; the codon identifier is a unique identifier of the current codon type; the codon triplet base encoding is a triplet base encoding of the current codon type; the codon usage deviation is a usage deviation scalar of the current codon type in the first genome sequence; The first anticodon data table includes a plurality of first anticodon data; each of the first anticodon data corresponds to an anticodon type in the first genome sequence; the first anticodon data includes an anticodon identifier, an anticodon triplet base code and a tRNA gene copy number; the anticodon identifier is a unique identifier of the current anticodon type; the anticodon triplet base code is a triplet base code of the current anticodon type; the tRNA gene copy number is the total number of tRNA genes corresponding to the current anticodon type in the first genome sequence; The first gene expression data table includes multiple first gene expression data; each of the first gene expression data corresponds to a DNA gene or a synthetic gene in the first genome sequence; the first gene expression data includes a gene identifier, a gene type, a gene base coding sequence, a gene codon sequence, a gene transcription efficiency, a gene translation efficiency and a gene abundance; the gene identifier is a unique identifier of the current gene; the gene type includes at least a DNA gene and a synthetic gene; the gene base coding sequence is a base coding sequence of the current gene; the gene codon sequence is composed of the codon identifiers of all codon types of the current gene; the gene transcription efficiency, the gene translation efficiency and the gene abundance are the transcription efficiency scalar, the translation efficiency scalar and the abundance scalar of the current gene under the first set conditions.

5. The processing method of the tRNA adaptation index weight prediction model according to claim 4, characterized in that: The performing model training on the adaptation index weight prediction model and three downstream task models based on the first data set specifically includes: Step 501, preparing training data based on the first data set to obtain a corresponding second data set; extracting the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector and the first codon usage deviation vector of each second data record in the second data set to form a corresponding third data record; deduplicating all the obtained third data records; and forming a corresponding third data set from all the deduplicated third data records; Wherein, the second data set includes a plurality of second data records; the second data record includes a first training genome sequence encoding vector, a first training experimental condition encoding vector, a first training gene copy number vector, a first codon usage deviation vector, a first training codon index set, a first tag transcription efficiency, a first tag translation efficiency and a first tag gene abundance; the third data set is composed of a plurality of the third data records, each of which is composed of the corresponding first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector and the first codon usage deviation vector; Step 502, randomly splitting the third data set based on a preset first split ratio to obtain a corresponding first training set and a first evaluation set; Wherein, both the first training set and the first evaluation set are composed of a plurality of the third 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 503, taking the first third data record of the first training set as the corresponding current training record; Step 504, inputting the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current training record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain a corresponding first predicted adaptation index weight vector; Step 505, bringing the first predicted adaptation index weight vector and the first codon usage deviation vector of the current training record into a preset first correlation function; and performing a round of parameter modulation on the adaptation index weight prediction model in a direction that makes the first correlation function reach a maximum positive correlation based on a preset first model optimization algorithm; and at the end of this round of parameter modulation, identifying whether the current training record is the last third data record of the first training set; if so, go to step 506; if not, extracting the next third data record of the first training set as the new current training record and returning to step 504 to continue training; Wherein, the first correlation function includes at least Pearson correlation function, Spearman correlation function, and Kendall correlation function; the first model optimization algorithm includes at least hill climbing algorithm and genetic algorithm; Step 506, perform a round of traversal on all the third data records of the first evaluation set; and in this round of traversal, use the third data record currently traversed as the corresponding current evaluation record; and input the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current evaluation record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain the corresponding second predicted adaptation index weight vector; and bring the second predicted adaptation index weight vector and the first codon usage deviation vector of the current training record into the first correlation function to calculate and obtain the corresponding first correlation; and at the end of this round of traversal, bring all the obtained first correlations into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value; Wherein, the first model evaluation function at least includes a MAE function, an MSE function, and a RMSE function; Step 507, identifying whether the first evaluation value satisfies a preset first evaluation value range; if the first evaluation value satisfies the first evaluation value range, identifying whether all the first correlations obtained satisfy the preset first correlation range, if not, returning to step 501 to continue training, if yes, solidifying the model parameters of the adaptive index weight prediction model and going to step 508; if the first evaluation value does not satisfy the first evaluation value range, returning to step 501 to continue training; Step 508, randomly splitting the second data set based on a preset second split ratio to obtain a corresponding second training set and a second evaluation set; Wherein, both the second training set and the second evaluation set are composed of a plurality of the second data records; the ratio of the total number of records in the second training set to the total number of records in the second evaluation set satisfies the second segmentation ratio; Step 509, taking the first second data record of the second training set as the corresponding current training record; Step 510, inputting the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current training record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain a corresponding third predicted adaptation index weight vector; Step 511, inputting the third predicted adaptation index weight vector and the first training codon index set of the current training record into the transcription efficiency prediction model, the translation efficiency prediction model and the gene abundance prediction model respectively to predict and obtain the corresponding first predicted transcription efficiency, first predicted translation efficiency and first predicted gene abundance; Step 512: bring the first predicted transcription efficiency and the first label transcription efficiency of the current training record into a preset first transcription loss function to calculate a corresponding first loss value; bring the first predicted translation efficiency and the first label translation efficiency of the current training record into a preset first translation loss function to calculate a corresponding second loss value; bring the first predicted gene abundance and the first label gene abundance of the current training record into a preset first abundance loss function to calculate a corresponding third loss value; Wherein, the first transcription loss function, the first translation loss function and the first abundance loss function all include at least an L1 loss function and an L2 loss function; Step 513, identifying whether the first, second and third loss values ​​satisfy the preset first, second and third loss value ranges; if the first, second and third loss values ​​all satisfy the corresponding first, second and third loss value ranges, go to step 514; if the first loss value does not satisfy the first loss value range, a round of parameter modulation is performed on the transcription efficiency prediction model in a direction in which the function value of the first transcription loss function reaches a minimum value based on the preset first model optimizer, and the method returns to step 511 at the end of this round of parameter modulation; if the second loss value does not satisfy the second loss value range, a round of parameter modulation is performed on the translation efficiency prediction model in a direction in which the function value of the first translation loss function reaches a minimum value based on the preset second model optimizer, and the method returns to step 511 at the end of this round of parameter modulation; if the third loss value does not satisfy the third loss value range, a round of parameter modulation is performed on the gene abundance prediction model in a direction in which the function value of the first abundance loss function reaches a minimum value based on the preset third model optimizer, and the method returns to step 511 at the end of this round of parameter modulation; Wherein, the first, second and third model optimizers all include at least an SGD series optimizer and an ADAM series optimizer; Step 514, perform a round of traversal on all the second data records of the second evaluation set; and in this round of traversal, use the second data record currently traversed as the corresponding current evaluation record; and input the first training genome sequence encoding vector, the first training experimental condition encoding vector and the first training gene copy number vector of the current evaluation record into the adaptation index weight prediction model to perform tRNA adaptation index weight prediction processing to obtain the corresponding fourth predicted adaptation index weight vector; and input the fourth predicted adaptation index weight vector and the first training codon index set of the current evaluation record into the transcription efficiency prediction model, the translation efficiency prediction model and the The gene abundance prediction model is used to predict the corresponding second predicted transcription efficiency, the second predicted translation efficiency and the second predicted gene abundance; and the second predicted transcription efficiency, the second predicted translation efficiency and the second predicted gene abundance form a corresponding first prediction vector, and the first label transcription efficiency, the first label translation efficiency and the first label gene abundance of the current evaluation record form a corresponding first label vector, and the first prediction vector and the first label vector form a corresponding first prediction-label pair; and at the end of this round of traversal, all the first prediction-label pairs obtained are brought into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value; Wherein, the second model evaluation function at least includes a MAE function and a RMSE function; Step 515, identifying whether the second evaluation value meets the preset second evaluation value range; if the second evaluation value meets the second evaluation value range, solidifying the model parameters of the transcription efficiency prediction model, the translation efficiency prediction model and the gene abundance prediction model and going to step 516; if the second evaluation value does not meet the second evaluation value range, returning to step 508 to continue training; Step 516, stop model training and confirm that the model training of the adaptive index weight prediction model and the three downstream task models is completed.

6. The processing method of the tRNA adaptation index weight prediction model according to claim 5, characterized in that: The preparing training data based on the first data set to obtain a corresponding second data set specifically includes: Step 61, taking each of the first data records of the first data set as a corresponding current data record; Step 62, using the first set condition, the first codon data table, the first anti-codon data table and the first gene expression data table of the current data record as the corresponding current set condition, current codon data table, current anti-codon data table and current gene expression data table; Step 63, extract all the codon triplet base codes in the current codon data table and sort them in sequence to form a corresponding first codon coding vector; extract all the anticodon triplet base codes in the current anticodon data table and sort them in sequence to form a corresponding first anticodon coding vector; and perform vector splicing on the first codon coding vector and the first anticodon coding vector to obtain a corresponding first training genome sequence coding vector; Step 64, based on a preset condition coding rule, the temperature, humidity, light, nutritional condition, and physiological / cultivation stage in the currently set conditions are respectively encoded to obtain corresponding temperature codes, humidity codes, light codes, nutritional condition codes, and physiological / cultivation stage codes to form a corresponding first training experiment condition coding vector; Step 65, extracting all the tRNA gene copy numbers in the current anticodon data table and sequentially sorting them to form a corresponding first training gene copy number vector; Step 66, extracting all the codon usage deviations in the current codon data table and sequentially sorting them to form a corresponding first codon usage deviation vector; Step 67, perform a round of traversal on all the first gene expression data in the current gene expression data table; and in this round of traversal, use the first gene expression data currently traversed as the corresponding current gene expression data; and use the gene codon sequence, gene transcription efficiency, gene translation efficiency and gene abundance of the current gene expression data as the corresponding first training codon index set, the first label transcription efficiency, the first label translation efficiency and the first label gene abundance; and form a corresponding second data record with the first training genome sequence encoding vector, the first training experimental condition encoding vector, the first training gene copy number vector, the first codon usage deviation vector and the first training codon index set, the first label transcription efficiency, the first label translation efficiency and the first label gene abundance corresponding to the current gene expression data; and at the end of this round of traversal, form a first record set corresponding to the current data record with all the second data records obtained in this round of traversal; Step 68: Merge the first record sets corresponding to all the first data records to obtain the corresponding second data sets.

7. The processing method of the tRNA adaptation index weight prediction model according to claim 1, characterized in that: The preprocessing of the current experimental condition, the current genome sequence and the current target gene to obtain the corresponding current experimental condition encoding vector, the current genome sequence encoding vector, the current tRNA gene copy number vector and the current target gene codon index set specifically includes: Step 71, based on a preset condition coding rule, the temperature, humidity, light, nutritional condition, and physiological / cultivation stage in the current experimental condition are respectively encoded to obtain corresponding temperature codes, humidity codes, light codes, nutritional condition codes, and physiological / cultivation stage codes to form a corresponding current experimental condition coding vector; Step 72, using a preset bioinformatics tool to analyze and count the codon types of all mRNA genes in the current genome sequence, and sequentially sorting all the codon types counted to obtain a corresponding first type sequence; and sequentially sorting the triplet base codes of all codon types in the first type sequence to obtain a corresponding current codon code vector; Step 73, using the bioinformatics tool to analyze and count the anticodon types of all tRNA genes in the current genome sequence, and sequentially sorting all the statistical anticodon types to obtain a corresponding second type sequence; and sequentially sorting the triplet base codes of all anticodon types in the second type sequence to obtain a corresponding current anticodon coding vector; Step 74, performing vector concatenation on the current codon encoding vector and the current anti-codon encoding vector to obtain the corresponding current genome sequence encoding vector; Step 75, using the bioinformatics tool to perform statistical calculations on the total number of gene copies of each tRNA gene in the current genome sequence and using the calculation result as the corresponding tRNA gene copy number; and sequentially sorting all the obtained tRNA gene copy numbers to obtain the corresponding current tRNA gene copy number vector; Step 76, using the bioinformatics tool to analyze and count all codon types of the current target gene to obtain a corresponding third type sequence; and extracting the sorting index of each codon type of the third type sequence in the first type sequence as a corresponding first codon index; and forming a corresponding codon index set of the current target gene from all the first codon indexes obtained; Step 77, output the current experimental condition encoding vector, the current genome sequence encoding vector, the current tRNA gene copy number vector and the current target gene codon index set obtained this time as the processing result of this preprocessing.

8. A device for executing the processing method of the tRNA adaptation index weight prediction model according to any one of claims 1 to 7, characterized in that: The device comprises: a model building module, a model training module, a data preparation module and a model application module; The model building module is used to build an adaptation index weight prediction model for processing tRNA adaptation index weight prediction tasks; and to design three downstream task models: a transcription efficiency prediction model, a translation efficiency prediction model, and a gene abundance prediction model; The model training module is used to collect data on the genome sequences of the target species under different conditions and construct a first data set based on the collected data; and to perform model training on the adaptation index weight prediction model and three downstream task models based on the first data set; The data preparation module is used to receive the current experimental conditions, current genome sequence and current target gene input by the user after the model training is completed; and pre-process the current experimental conditions, the current genome sequence and the current target gene to obtain the corresponding current experimental condition encoding vector, current genome sequence encoding vector, current tRNA gene copy number vector and current target gene codon index set; the current genome sequence is a genome sequence obtained by an observer through manual or machine observation / experimental means under the current experimental conditions for a type of cell of the target species; the current target gene is a DNA gene or synthetic gene in the current genome sequence; The model application module is used to perform tRNA adaptation index weight prediction processing by the adaptation index weight prediction model according to the current genome sequence encoding vector, the current experimental condition encoding vector and the current tRNA gene copy number vector to obtain the corresponding current adaptation index weight vector; and three downstream task models respectively perform corresponding transcription efficiency, translation efficiency and gene abundance prediction processing according to the current adaptation index weight vector and the current target gene codon index set to obtain the corresponding current transcription efficiency, current translation efficiency and current gene abundance composition corresponding task reports to feedback to the user.

9. 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 7; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

10. 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 7.

Images