A method and system for fibroblast transfection based on automated parameter optimization
By using automated parameter optimization methods, a hardware system was constructed and random forest model and particle swarm optimization algorithm were used to solve the shortcomings of parameter design and data processing in traditional fibroblast transfection methods, and efficient and stable transfection results were achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 天下秀(北京)再生医学技术有限公司
- Filing Date
- 2026-02-11
- Publication Date
- 2026-05-29
AI Technical Summary
Traditional fibroblast transfection methods lack systematic parameter design, leading to unstable efficiency and imperfect data processing, which affects the reproducibility and accuracy of experimental results.
An automated parameter optimization method was adopted, and a hardware system was built using an automated liquid handling workstation, a fluorescence microscope, an electroporation module, and a cell counter. Key transfection parameters were screened, data preprocessing and optimization were performed, and the optimal parameter combination was selected through a random forest model and a particle swarm optimization algorithm.
It improves the stability of transfection efficiency and the integrity and accuracy of data, solves the problems of random parameter design and imperfect data processing, and realizes efficient transfection operation.
Smart Images
Figure CN122104809A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cell biology experiments, specifically to a method and system for transfecting fibroblasts based on automated parameter optimization. Background Technology
[0002] In the field of cell biology experiments, fibroblast transfection is a crucial step in gene function research and protein expression analysis. Its operational quality and efficiency directly impact the accuracy and progress of subsequent experiments. With the increasing demands for standardization and efficiency in experiments, traditional fibroblast transfection methods have gradually revealed several problems that are difficult to adapt, specifically in the following two aspects: First, the parameter design lacks a systematic approach and relies on human experience, leading to unstable results. The transfection process involves multiple key parameters, and traditional methods often rely on the subjective experience of experimenters to adjust parameter combinations. This fails to fully consider the synergistic effects and constraints between parameters, and also makes it difficult to cover parameter adaptation scenarios under different experimental conditions. As a result, transfection efficiency fluctuates greatly, experimental results have poor reproducibility, and it is difficult to form a unified standard operating procedure.
[0003] Secondly, the imperfect data processing workflow affects the reliability of subsequent optimization. Problems such as missing data and outliers are prone to occur during experiments. Traditional processing methods often involve simple filling or direct removal of abnormal samples, which not only wastes data information but also leads to a decline in the quality of the data used for analysis. Furthermore, the lack of standardized data procedures prevents data from different sources and dimensions from forming a unified analytical basis, further reducing the accuracy of parameter importance assessment and optimization algorithm calculations, and hindering the improvement of transfection effectiveness.
[0004] Based on this, a fibroblast transfection method based on automated parameter optimization is proposed, aiming to solve the above-mentioned pain points through systematic parameter optimization. Summary of the Invention
[0005] To address the technical problems mentioned in the background section, this invention proposes a method and system for fibroblast transfection based on automated parameter optimization.
[0006] Therefore, the technical solution adopted by the present invention is as follows: According to a first aspect of the present invention, a method for transfecting fibroblasts based on automated parameter optimization is provided, the method comprising the following steps: Step S1: Utilize an automated liquid handling workstation, fluorescence microscope, and electroporation module. Incubators and cell counters constitute the hardware for automated fibroblast transfection experiments. Step S2: Screen key transfection parameters and form a feature vector; set the value range for the key transfection parameters and set constraints in the automated liquid processing workstation; Step S3: Preprocess fibroblasts according to preset feature vector parameters, acquire image data using the fluorescence microscope, analyze the image data to obtain target index vectors, and combine the feature vectors and target index vectors one-to-one as initial data; Step S4: Perform missing value imputation, outlier removal, and data standardization on the initial data to obtain standard data; Step S5: Based on the standard data, construct a random forest model and use the feature importance evaluation mechanism to select the core optimization parameters; Step S6: Select the optimal transfection parameter combination from the core optimization parameters using the particle swarm optimization algorithm; Step S7: The optimal transfection parameter combination is transmitted to the automated liquid processing workstation for format parsing. If the parsed parameters meet the constraints, they are converted into instruction signals that can be recognized by each supporting hardware, and then each hardware is driven to collaboratively execute the fibroblast transfection operation according to the optimal transfection parameters.
[0007] Furthermore, the feature vector is represented as:
[0008] in, Indicates cell density; Indicates nucleic acid concentration; Indicates the concentration of the transfection reagent; Indicates the volume ratio of reagents to nucleic acids; Indicates the voltage of electrical conversion; Indicates the duration of the electrical pulse; Indicates the incubation time of the complex; This indicates the concentration of the cell suspension.
[0009] Furthermore, the initial data is represented as follows:
[0010] in, This represents the feature vector of the j-th sample; Let m represent the target index vector for the j-th sample; m represents the total number of samples. The target index vector of the j-th sample is represented as:
[0011] in, , and Let represent the target indicators for the j-th sample, and let represent the transfection efficiency, cell viability, and mean fluorescence intensity, respectively.
[0012] Furthermore, the missing value imputation of the initial data includes using K-nearest neighbor interpolation, specifically as follows: The first step is to calculate the Euclidean distance between the j-th sample (where the target indicator is missing) and all samples in the initial data that do not have missing target indicators, using the following formula:
[0013] in, Represents the j-th sample. One eigenvalue; This represents the h-th sample where the non-target indicator is missing. One eigenvalue; This represents the sample where the h-th non-target indicator is missing; The second step is to select 5 samples with the smallest Euclidean distance from the j-th sample that are missing non-target indicators, and form a nearest neighbor sample set. The third step is to calculate the weight of the nearest neighbor sample, using the following formula:
[0014] in, Indicates the first The nearest neighbor samples, ; Indicates the first The weights of the nearest neighbor samples; The fourth step is to calculate the value of the missing target indicator in the j-th sample based on the weights of the nearest neighbor samples. The calculation formula is:
[0015] in, Indicates the first The target index vector of the nearest neighbor samples.
[0016] Furthermore, the outlier removal process for the initial data includes: using the Z-score method to remove outliers from the transfection efficiency (TE), specifically as follows: The first step is to calculate the average transfection efficiency (TE) in the initial data. and standard deviation ; The second step is to calculate the Z-score of the transfection efficiency (TE) for each sample. For the j-th sample... Z-score value , The calculation formula is:
[0017] Third step, if Then determine the j-th sample. If the value is an outlier, the corresponding sample will be removed from the initial data; The outlier removal process for cell viability (CS) and mean fluorescence intensity (FI) is consistent with the outlier removal process for transfection efficiency (TE).
[0018] Furthermore, the specific process of step S5 is as follows: Step S501: Use the standard data as input; Step S502: Based on the input, train the random forest model using 5-fold cross-validation; Step S503: Based on the feature importance evaluation mechanism of the random forest model, select the parameters that contribute the most to the transfection efficiency (TE) as the core optimization parameters.
[0019] Further, step S503 includes: Step S5031: Traverse all nodes of each decision tree in the random forest one by one, and use each parameter in the feature vector to split the sample of the current node in turn; Step S5032: Calculate the impurity of the current node before splitting, using the formula:
[0020] in, Indicates the number of samples in the current node; This represents the transfection efficiency of the e-th sample within the node; This represents the average transfection efficiency of all samples within a node. Step S5033: Calculate the impurity of the two child nodes after splitting for each parameter, and obtain... and ; Step S5034: Calculate the contribution of the parameter to the reduction of impurity at the current node, using the following formula:
[0021] Step S5035: For the same decision tree, set the parameters of each parameter... Add them together to get the total contribution of each parameter in this decision tree. ;in, Represents the r-th parameter; Indicates the first A decision tree; Step S5036: Calculate the average of the total contribution values of each parameter across all decision trees. ,pass Calculate the importance score for each parameter. The calculation formula is:
[0022] in, This represents the average of the total contribution values of the r-th parameter across all decision trees; Step S5037: Filter by importance score The parameters are used as core optimization parameters, and the core optimization parameters are expressed as follows: ,in, The value of l is determined by the selection results of the random forest model.
[0023] Furthermore, the specific process of step S6 is as follows: Step S601: Use the core optimization parameters as input; Step S602: Encode the core optimization parameters into particles, with each particle being a binary vector. The vector dimension corresponding to the core optimization parameters; Based on the value range of each key transfection parameter setting, an initial feature value is generated for each particle's characteristics using a random function; definition When indicating the selection of the first Dimensional features, At that time, it meant abandoning the first. Dimensional features; Step S603: With the goal of maximizing transfection efficiency (TE) and cell viability (CS) to be greater than or equal to 80%, define the fitness function as follows:
[0024] in, Represents particles Transfection efficiency for the corresponding parameter combination; Represents particles Cell viability for the corresponding parameter combinations; Indicates the penalty item, if The penalty term is 0, and the objective function value equals TE; if The penalty term increases as CS decreases; Step S604: Update particle velocity and position, and find the global optimal position for the population; Step S605: Repeat the particle velocity and position update, and stop the update iteration when the following condition is met: Condition 1: Stop after the number of iterations reaches the preset maximum number of iterations; Condition 2: Global optimal position of the population The change was less than 0.1%; Step S606: After the iteration terminates, output the final global optimal position of the population as the optimal combination of transfection parameters.
[0025] Furthermore, step S604 includes: Step S6041: The particle velocity update formula for generation t is:
[0026] in, Represents particles In the The velocity vector of the next iteration; and The range is Random numbers within; Represents particles The velocity vector in the t-th iteration; Represents particles The historical best position; Indicates the globally optimal position of the group; Indicates inertia weight; and Indicates the learning factor; Represents particles The position vector at the t-th iteration; Step S6042: The particle position update formula for generation t is:
[0027] Step S6043: Recalculate the objective function value for each particle and update and ; According to a second aspect of the present invention, a fibroblast transfection system based on automated parameter optimization is provided. The system includes a server, an automated liquid handling workstation, and supporting hardware; the supporting hardware includes a fluorescence microscope, an electroporation module, and... An incubator and a cell counter; the server is connected to the automated liquid handling workstation; the supporting hardware communicates with the automated liquid handling workstation via a dedicated transmission protocol. The server is used to: acquire initial data; the initial data is obtained in the following way: preprocessing fibroblasts according to preset feature vector parameters, acquiring image data using the fluorescence microscope, analyzing the image data to obtain target index vectors, combining the feature vectors and the target index vectors one-to-one as initial data; performing missing value imputation, outlier removal, and data standardization on the initial data to obtain standard data; constructing a random forest model based on the standard data, and using a feature importance evaluation mechanism to screen out core optimization parameters; and using a particle swarm optimization algorithm to screen out the optimal combination of transfection parameters from the core optimization parameters. The automated liquid processing workstation is used to set constraints related to key transfection parameters; preprocess fibroblasts according to a preset feature vector parameter composed of key transfection parameters; receive the optimal transfection parameter combination and perform format parsing; if the parsed parameters meet the constraints, they are converted into instruction signals recognizable by each supporting hardware, and then drive each hardware to collaboratively execute the fibroblast transfection operation according to the optimal transfection parameters.
[0028] Compared with the prior art, the advantages of the present invention are as follows: 1. In the parameter design stage, this invention selects eight key transfection parameters to form a feature vector. At the same time, it sets parameter constraints based on cell physiological characteristics and experimental logic to ensure the effectiveness of parameter combinations. This overcomes the limitations of existing technologies, which rely on human experience for parameter design, lack systematic constraints, and have strong randomness in parameter combinations.
[0029] 2. In the data preprocessing stage, this invention uses K-nearest neighbor interpolation to fill in missing values of the target index, combined with Z-score method to remove outliers and data standardization, which significantly improves data integrity and accuracy. It solves the problem that traditional methods rely solely on mean interpolation and simple filtering, which leads to data distortion and missing information, and cannot support subsequent accurate analysis.
[0030] 3. In the parameter optimization stage, this invention first quantifies the importance of eight parameters to transfection efficiency through a random forest model, selects core optimization parameters, and then uses a particle swarm optimization algorithm to search for the optimal combination, overcoming the problems of blind parameter optimization, low efficiency of manual trial and error, and difficulty in balancing multiple objective requirements in existing technologies. Attached Figure Description
[0031] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0032] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a flowchart illustrating step S5 of the present invention; Figure 3 This is a flowchart illustrating step S6 of the present invention. Detailed Implementation
[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, other embodiments obtained by those skilled in the art without creative effort are all within the protection scope of this invention.
[0034] To achieve the above objectives, the present invention is implemented through the following technical solutions.
[0035] Example 1 This embodiment provides a fibroblast transfection method based on automated parameter optimization.
[0036] In one embodiment, the specific process of this method is described in [reference]. Figure 1 ,include: Step S1, Hardware integration of automated devices: utilizing an automated liquid handling workstation, fluorescence microscope, and electroporation module. Incubators and cell counters constitute the hardware for automated fibroblast transfection experiments.
[0037] This automated liquid handling workstation can be, for example, the Biomek i7 automated liquid handling workstation. With the Biomek i7 automated liquid handling workstation as the core control unit, supporting hardware includes a fluorescence microscope, an electroporation module, etc. The incubator and cell counter, along with the accompanying hardware, communicate with the Biomek i7 via a proprietary transmission protocol. The specific connection method and functions are as follows: Fluorescence microscope: Communicates with Biomek i7 via RS-232 serial port. When Biomek i7 transmits the 96-well plate arrival signal, the microscope performs automatic image acquisition operation to realize automated image acquisition of the transfection effect. Electroporation module: Connects to Biomek i7 via USB interface, transmits electroporation voltage and pulse duration based on Lonza protocol to ensure accurate execution of electroporation operation; Incubator: Connects to Biomek i7 via TCP / P protocol. Biomek i7 reads the temperature and humidity data inside the incubator in real time and sends gate control commands to adjust the door opening and closing, ensuring a stable cell culture environment. Cell counter: Automated detection of fibroblast concentration and viability, providing data support for subsequent cell suspension preparation.
[0038] Step S2, Design of key transfection parameters: Screen key transfection parameters and form a feature vector; set the value range for the key transfection parameters and set constraints in the automated liquid processing workstation.
[0039] Screening cell density Nucleic acid concentration Concentration of transfection reagent The volume ratio of reagents to nucleic acids Electricity to voltage Electrical pulse duration Complex incubation time Cell suspension concentration A total of 8 key parameters are used to form a feature vector, and the feature vector... Represented as:
[0040] The value range of each parameter is as follows: , , , , , , , ; Based on cell physiological characteristics and experimental logic, constraints are set in the corresponding Biomek i7 software to ensure the effectiveness of parameter combinations. These constraints include: Cell density and nucleic acid concentration constraints: when cell density The range is in At that time, nucleic acid concentration ; in, and The range of values is determined based on historical experimental data and human experience; Electro-rotation parameter constraints: Electro-rotation voltage With pulse duration The product satisfies the following condition:
[0041] Concentration and Constraints: Culture medium and The concentrations satisfy a linear relationship:
[0042] in, express The degree of quantification of concentration.
[0043] Step S3, Fibroblast Pretreatment and Data Acquisition: Fibroblasts are pretreated according to preset feature vector parameters, image data is acquired using the fluorescence microscope, the image data is analyzed to obtain target index vectors, and the feature vectors and target index vectors are combined one-to-one as initial data.
[0044] Take the cell cryopreservation tubes, thaw them in a 37°C water bath, transfer them to centrifuge tubes containing DMEM complete medium, and centrifuge at 1000 rpm for 5 minutes. Discard the supernatant from the centrifuge tube to form a cell suspension, then inoculate into a 25 cm² culture flask and incubate at 37°C with 5% [cell suspension]. Incubate in an incubator, changing the complete culture medium every 2-3 days. When the confluence of cells in the culture flask reaches 80%-90%, perform cell passage. After passage, cells were digested to prepare a single-cell suspension, stained with trypan blue, and cell viability was assessed. The cell density corresponding to this pretreatment was recorded. and cell suspension concentration ; The single-cell suspension was diluted according to multiple sets of feature vector parameters preset in the experiment, with each dilution corresponding to one independent sample; For the diluted sample j, record the feature vector of sample j. :
[0045] The diluted sample j was automatically plated using a Biomek i7, incubated in a CO2 incubator for 24 hours, and then sorted according to the eigenvector. The robotic arm prepares the transfection complex and adds it to the sample wells, and the electrotransfer group applies the electrotransfer voltage and electrotransfer pulse duration through the electrotransfer module. Images of sample wells of sample j are acquired every 6 hours using a fluorescence microscope. The target indicators for sample j are calculated using ImageJ software, including transfection efficiency. Cell viability and mean fluorescence intensity ; The target indicators are combined into a target indicator vector, and the target indicator vector for sample j is represented as follows:
[0046] The feature vectors and target index vectors are combined in a one-to-one correspondence and used as initial data, which is then transmitted to the server in real time for backup. The initial data is represented as follows:
[0047] in, This represents the feature vector of the j-th sample; Let m represent the target index vector of the j-th sample; m represents the total number of samples.
[0048] Step S4: Data preprocessing: The initial data is imputed for missing values, outliers are removed, and the data is standardized to obtain standard data.
[0049] The initial data is used as input for data preprocessing; the data preprocessing includes missing value imputation, outlier removal, and data standardization. For the j-th sample with missing target indicators, the K-nearest neighbor interpolation method is used to fill in the missing values using the feature vector of the initial data. The specific process is as follows: The first step is to calculate the Euclidean distance between the j-th sample and all samples in the initial data that are missing non-target indicators, using the following formula:
[0050] in, Represents the j-th sample. One eigenvalue; This represents the h-th sample where the non-target indicator is missing. One eigenvalue; This represents the sample where the h-th non-target indicator is missing; The second step is to select the five non-target indicator missing samples with the smallest Euclidean distance from the j-th sample, based on the calculated Euclidean distance, to form the nearest neighbor sample set, which is represented as:
[0051] The set of target index value vectors corresponding to the nearest neighbor sample set is:
[0052] The third step is to calculate the weights of the nearest neighbor samples, using the following formula:
[0053] in, Indicates the first The nearest neighbor samples, ; Indicates the first The weights of the nearest neighbor samples; Fourth step: Calculate the value of the missing target indicator in the j-th sample based on the weights of the nearest neighbor samples. The calculation formula is:
[0054] in, Indicates the first The target index vector of the nearest neighbor samples; Outliers in the target index vector are determined using the Z-score method. The target index vector includes transfection efficiency (TE), cell viability (CS), and mean fluorescence intensity (FI). The outlier removal operation is performed on the transfection efficiency TE, and the specific process is as follows: The first step is to calculate the mean and standard deviation of the transfection efficiency (TE) of the initial data; The average value The calculation formula is:
[0055] Where m represents the total number of samples; The standard deviation The calculation formula is:
[0056] The second step is to calculate the Z-score of the transfection efficiency (TE) for each sample. For the j-th sample... Z-score value , The calculation formula is:
[0057] Third step, if Then determine the j-th sample. If the value is an outlier, the corresponding sample will be removed from the initial data; Among them, the outlier removal process for cell viability (CS) and mean fluorescence intensity (FI) is the same as the outlier removal process for transfection efficiency (TE). The initial data, after missing value imputation and outlier removal, is standardized to map the values corresponding to the feature vector and the target index vector to... Within the interval, as standard data , represented as:
[0058] Where k represents the total number of samples after data preprocessing.
[0059] Step S5: Random Forest Model Construction and Parameter Importance Analysis: Based on the standard data, a random forest model is constructed, and the core optimization parameters are selected using the feature importance evaluation mechanism.
[0060] The parameters that contribute most to transfection efficiency (TE) are selected as core optimization parameters. The specific process is as follows: Figure 2 As shown, the specific process is as follows: Step S501: Transfer the standard data As input; Step S502: Train the random forest model using 5-fold cross-validation. The process includes the following sub-steps: Step S5021: Transfer the training set The sample was divided into 5 non-overlapping subsets with the same number of samples, denoted as . , ; Step S5022: Perform 5-fold cross-validation on the training set. The process is as follows: No. Round verification, , except for the Merge all subsets outside of the specified subset. , as a training subset; As a subset for verification; After training the model on the training subset, performance metrics are calculated on the validation subset. and RMSE; Step S5023: Calculate the results of 5 rounds of verification. and ;if and The random forest model training is complete; in, express The mean; This represents the mean of RMSE; Step S503: Based on the feature importance evaluation mechanism of the random forest model, the importance of each transfection parameter is quantified by its contribution to reducing the impurity of the decision tree, and the parameters that contribute the most to the transfection efficiency (TE) are selected as the core optimization parameters. The selection process in step S503 includes the following sub-steps: Step S5031: For each decision tree in the random forest, traverse all nodes of the tree one by one, and calculate the cell density in the feature vector. Nucleic acid concentration Concentration of transfection reagent The volume ratio of reagents to nucleic acids Electricity to voltage Electrical pulse duration Complex incubation time and cell suspension concentration Eight parameters are used to split the sample of the current node in turn, with each parameter applied sequentially. Step S5032: Split each parameter. First, calculate the impurity of the current node before splitting, using the formula:
[0061] in, Indicates the number of samples in the current node; This represents the transfection efficiency of the e-th sample within the node; This represents the average transfection efficiency of all samples within a node. Step S5033: Calculate the impurity of the two child nodes after splitting for each parameter, and obtain... and ; Step S5034: Calculate the contribution of the parameter to the reduction of impurity at the current node, using the following formula:
[0062] Step S5035: For the same decision tree, set the parameters of each parameter... Add them together to get the total contribution of each parameter in this decision tree. ;in, Represents the r-th parameter; Indicates the first A decision tree; Step S5036: If the random forest contains a total of T decision trees, for each parameter, calculate the average of its total contribution value across all decision trees, using the following formula:
[0063] pass The importance score for each parameter is calculated using the following formula:
[0064] in, This represents the average of the total contribution values of the r-th parameter across all decision trees; Step S5037: Filter by importance score The parameters are used as core optimization parameters, and the core optimization parameters are expressed as follows: ,in, The value of l is determined by the selection results of the random forest model.
[0065] Step S6: Particle swarm optimization algorithm to search for the optimal combination of transfection parameters: The optimal combination of transfection parameters is selected from the core optimization parameters by using the particle swarm optimization algorithm.
[0066] The optimal transfection parameter combination is selected from the core optimization parameters through particle swarm optimization. The specific process is as follows: Figure 3 As shown, the specific steps are as follows: Step S601: Use the core optimization parameters as input; Step S602: Perform particle encoding, setting the number of particles to 100, with each particle being a binary vector. The vector dimension corresponding to the core optimization parameters; Based on the value range of each key transfection parameter setting, an initial feature value is generated for each particle's characteristics using a random function; definition When indicating the selection of the first Dimensional features, At that time, it meant abandoning the first. Dimensional features; The particle represents a transfection parameter selection scheme; Step S603: With the goal of maximizing transfection efficiency (TE) and cell viability (CS) to be greater than or equal to 80%, define the fitness function as follows:
[0067] in, Represents particles Transfection efficiency for the corresponding parameter combination; Represents particles Cell viability for the corresponding parameter combinations; Indicates the penalty item, if The penalty term is 0, and the objective function value equals TE; if The penalty term increases as CS decreases; Step S604: Update particle velocity and position, and find the global optimal position for the population. The process includes the following sub-steps: Step S6041: The particle velocity update formula for generation t is:
[0068] in, Represents particles In the The velocity vector of the next iteration; and The range is Random numbers within; Represents particles The velocity vector in the t-th iteration; Represents particles The historical best position; Indicates the globally optimal position of the group; Indicates inertia weight; and Indicates the learning factor; Represents particles The position vector at the t-th iteration; Step S6042: The particle position update formula for generation t is:
[0069] Step S6043: After updating the particle position in each iteration, recalculate the objective function value for each particle and update... and ; Step S605: Repeat the particle velocity and position update in step S604. Stop the update iteration when the following condition is met: Condition 1: Stop after the number of iterations reaches the preset maximum number of iterations; Condition 2: Global optimal position of the population The change was less than 0.1%; Step S606: After the iteration terminates, output the final global optimal position of the population as the optimal combination of transfection parameters.
[0070] Step S7, Local Execution: The optimal transfection parameter combination is transmitted to the automated liquid processing workstation for format parsing. If the parsed parameters meet the constraints, they are converted into instruction signals that can be recognized by each supporting hardware, driving each hardware to collaboratively execute the fibroblast transfection operation according to the optimal transfection parameters.
[0071] The optimal transfection parameter combination is parsed according to the format of the workstation in the Biomeki7 automated liquid handling workstation. If the parsed parameters do not meet the constraints set in the Biomek i7 software, the optimal transfection parameter combination will be generated again. If the parsed parameters meet the constraints set in the Biomek i7 software, they are converted into instruction signals that can be recognized by each supporting hardware and then sent to the corresponding hardware such as the fluorescence microscope, electroporation module, and CO2 incubator, driving each hardware to collaboratively execute the fibroblast transfection operation according to the optimal transfection parameters.
[0072] Furthermore, steps S4 to S6 of the above method can be executed in a server connected to the automated liquid handling workstation for data interaction, such as acquiring initial data and transmitting optimal transfection parameter combinations. The server may include a memory and a processor for storing and processing the data.
[0073] Example 2 This embodiment provides a fibroblast transfection system based on automated parameter optimization. The system includes a server, an automated liquid handling workstation, and supporting hardware; the supporting hardware includes a fluorescence microscope, an electroporation module, and... An incubator and a cell counter; the server is connected to the automated liquid handling workstation; the supporting hardware communicates with the automated liquid handling workstation via a dedicated transmission protocol. The server is used to: acquire initial data; the initial data is obtained in the following way: preprocessing fibroblasts according to preset feature vector parameters, acquiring image data using the fluorescence microscope, analyzing the image data to obtain target index vectors, combining the feature vectors and the target index vectors one-to-one as initial data; performing missing value imputation, outlier removal, and data standardization on the initial data to obtain standard data; constructing a random forest model based on the standard data, and using a feature importance evaluation mechanism to screen out core optimization parameters; and using a particle swarm optimization algorithm to screen out the optimal combination of transfection parameters from the core optimization parameters. The automated liquid processing workstation is used to set constraints related to key transfection parameters; preprocess fibroblasts according to a preset feature vector parameter composed of key transfection parameters; receive the optimal transfection parameter combination and perform format parsing; if the parsed parameters meet the constraints, they are converted into instruction signals recognizable by each supporting hardware, and then drive each hardware to collaboratively execute the fibroblast transfection operation according to the optimal transfection parameters.
[0074] The functions of the server, automated liquid handling workstation, and supporting hardware have been described in detail in Example 1, and will not be repeated here.
[0075] Therefore, compared with the prior art, the advantages of the present invention are as follows: 1. In the parameter design stage, this invention selects eight key transfection parameters to form a feature vector. At the same time, it sets parameter constraints based on cell physiological characteristics and experimental logic to ensure the effectiveness of parameter combinations. This overcomes the limitations of existing technologies, which rely on human experience for parameter design, lack systematic constraints, and have strong randomness in parameter combinations.
[0076] 2. In the data preprocessing stage, this invention uses K-nearest neighbor interpolation to fill in missing values of the target index, combined with Z-score method to remove outliers and data standardization, which significantly improves data integrity and accuracy. It solves the problem that traditional methods rely solely on mean interpolation and simple filtering, which leads to data distortion and missing information, and cannot support subsequent accurate analysis.
[0077] 3. In the parameter optimization stage, this invention first quantifies the importance of eight parameters to transfection efficiency through a random forest model, selects core optimization parameters, and then uses a particle swarm optimization algorithm to search for the optimal combination, overcoming the problems of blind parameter optimization, low efficiency of manual trial and error, and difficulty in balancing multiple objective requirements in existing technologies.
[0078] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for transfecting fibroblasts based on automated parameter optimization, characterized in that, The method includes the following steps: Utilizing automated liquid handling workstations, fluorescence microscopes, and electroporation modules, Incubators and cell counters constitute the hardware for automated fibroblast transfection experiments. Key transfection parameters are selected and feature vectors are formed; value ranges are set for the key transfection parameters, and constraints are set in the automated liquid processing workstation. Fibroblasts are preprocessed according to preset feature vector parameters, image data is acquired using the fluorescence microscope, the image data is analyzed to obtain target index vectors, and the feature vectors and target index vectors are combined one-to-one as initial data. The initial data is imputed for missing values, removed for outliers, and standardized to obtain standard data. Based on the standard data, a random forest model is constructed, and the core optimization parameters are selected using a feature importance evaluation mechanism. The optimal combination of transfection parameters is selected from the core optimization parameters using a particle swarm optimization algorithm. The optimal transfection parameter combination is transmitted to the automated liquid processing workstation for format parsing. If the parsed parameters meet the constraints, they are converted into instruction signals that can be recognized by each supporting hardware, driving each hardware to collaboratively execute the fibroblast transfection operation according to the optimal transfection parameters.
2. The method according to claim 1, characterized in that, The feature vector is represented as follows: in, Indicates cell density; Indicates nucleic acid concentration; Indicates the concentration of the transfection reagent; Indicates the volume ratio of reagents to nucleic acids; Indicates the voltage of the electrical conversion; Indicates the duration of the electrical pulse; Indicates the incubation time of the complex; This indicates the concentration of the cell suspension.
3. The method according to claim 1, characterized in that, The initial data is represented as follows: in, This represents the feature vector of the j-th sample; Let m represent the target index vector for the j-th sample; m represents the total number of samples. The target index vector of the j-th sample is represented as: in, , and Let represent the target indicators for the j-th sample, and let represent the transfection efficiency, cell viability, and mean fluorescence intensity, respectively.
4. The method according to claim 3, characterized in that, The process of imputing missing values in the initial data includes: Missing values are filled using the K-nearest neighbor interpolation method. The specific process is as follows: For the j-th sample lacking the target indicator, calculate the Euclidean distance between the j-th sample and all samples in the initial data that do not lack the target indicator, using the following formula: in, Represents the j-th sample. One eigenvalue; This represents the h-th sample where the non-target indicator is missing. One eigenvalue; This represents the sample where the h-th non-target indicator is missing; Select the five samples with the smallest Euclidean distance from the j-th sample that are missing non-target indicators and form the nearest neighbor sample set; The weights of the nearest neighbor samples are calculated using the following formula: in, Indicates the first The nearest neighbor samples, ; Indicates the first The weights of the nearest neighbor samples; Calculate the value of the missing target indicator in the j-th sample based on the weights of the nearest neighbor samples. The calculation formula is: in, Indicates the first The target index vector of the nearest neighbor samples.
5. The method according to claim 3, characterized in that, The outlier removal process for the initial data includes: Outliers in transfection efficiency (TE) were removed using the Z-score method. The specific process is as follows: Calculate the average value of the transfection efficiency TE in the initial data. and standard deviation ; Calculate the Z-score of the transfection efficiency (TE) for each sample, for the j-th sample. Z-score value , The calculation formula is: like Then determine the j-th sample. If the value is an outlier, the corresponding sample will be removed from the initial data; The outlier removal process for cell viability (CS) and mean fluorescence intensity (FI) is consistent with the outlier removal process for transfection efficiency (TE).
6. The method according to claim 2, characterized in that, Based on the standard data, a random forest model is constructed, and the core optimization parameters are selected using a feature importance evaluation mechanism. The specific process is as follows: Use the standard data as input; Based on the input, the random forest model is trained using 5-fold cross-validation; Based on the feature importance evaluation mechanism of the random forest model, the parameters that contribute the most to the transfection efficiency (TE) are selected as optimization parameters.
7. The method according to claim 6, characterized in that, The feature importance evaluation mechanism based on the random forest model selects the parameters that contribute the most to the transfection efficiency (TE) as optimization parameters, including: Iterate through all nodes of each decision tree in the random forest, and use each parameter in the feature vector to split the sample of the current node in turn. The impurity of the current node before splitting is calculated using the following formula: in, Indicates the number of samples in the current node; This represents the transfection efficiency of the e-th sample within the node; This represents the mean transfection efficiency of all samples within a node; Calculate the impurity of the two child nodes after splitting for each parameter, and obtain... and ; The contribution of the parameters to the reduction of impurity at the current node is calculated using the following formula: For the same decision tree, each parameter Add them together to get the total contribution of each parameter in this decision tree. ;in, Represents the r-th parameter; Indicates the first A decision tree; Calculate the average of the total contribution values for each parameter across all decision trees. ,pass Calculate the importance score for each parameter. The calculation formula is: in, This represents the average of the total contribution values of the r-th parameter across all decision trees; Filtering importance score The parameters are used as core optimization parameters, and the core optimization parameters are expressed as follows: ,in, The value of l is determined by the selection results of the random forest model.
8. The method according to claim 7, characterized in that, The process of selecting the optimal transfection parameter combination from the core optimization parameters using the particle swarm optimization algorithm is as follows: Use the aforementioned core optimization parameters as input; The core optimization parameters are encoded into particles, with each particle being a binary vector. , corresponding to the vector dimension of the core optimization parameters; Based on the value range of each key transfection parameter setting, an initial feature value is generated for each particle's characteristics using a random function; definition When indicating the selection of the first dimensional features, At that time, it meant abandoning the first. Dimensional features; With the goal of maximizing transfection efficiency (TE) and cell viability (CS) to be greater than or equal to 80%, a fitness function is defined as follows: in, Represents particles Transfection efficiency for the corresponding parameter combination; Represents particles Cell viability for the corresponding parameter combinations; Indicates the penalty item, if The penalty term is 0, and the fitness function objective function value equals TE; if The penalty term increases as CS decreases; Update particle velocity and position to find the global optimal position for the population; Repeat the particle velocity and position updates, and stop the update iteration when the following condition is met: Condition 1: Stop after the number of iterations reaches the preset maximum number of iterations; Condition 2: Global optimal position of the population The change was less than 0.1%; After the iteration terminates, the final global optimal position of the population is output as the optimal combination of transfection parameters.
9. The method according to claim 8, characterized in that, The process of updating particle velocity and position and finding the optimal combination of transfection parameters includes: The particle velocity update formula for generation t is: in, Represents particles In the The velocity vector of the next iteration; and The range is Random numbers within; Represents particles The velocity vector in the t-th iteration; Represents particles The historical best position; Indicates the globally optimal position of the group; Indicates inertia weight; and Indicates the learning factor; Represents particles The position vector at the t-th iteration; The particle position update formula for generation t is: Recalculate the fitness function for each particle and update it. and .
10. A fibroblast transfection system based on automated parameter optimization, characterized in that, It includes a server, an automated liquid handling workstation, and supporting hardware; the supporting hardware includes a fluorescence microscope, an electroporation module, and... An incubator and a cell counter; the server is connected to the automated liquid handling workstation; the supporting hardware communicates with the automated liquid handling workstation via a dedicated transmission protocol. The server is used to: acquire initial data; the initial data is obtained in the following way: preprocessing fibroblasts according to preset feature vector parameters, acquiring image data using the fluorescence microscope, analyzing the image data to obtain target index vectors, combining the feature vectors and the target index vectors one-to-one as initial data; performing missing value imputation, outlier removal, and data standardization on the initial data to obtain standard data; constructing a random forest model based on the standard data, and using a feature importance evaluation mechanism to screen out core optimization parameters; and using a particle swarm optimization algorithm to screen out the optimal combination of transfection parameters from the core optimization parameters. The automated liquid processing workstation is used to set constraints related to key transfection parameters; preprocess fibroblasts according to a preset feature vector parameter composed of key transfection parameters; receive the optimal transfection parameter combination and perform format parsing; if the parsed parameters meet the constraints, they are converted into instruction signals recognizable by each supporting hardware, and then drive each hardware to collaboratively execute the fibroblast transfection operation according to the optimal transfection parameters.