Preparation and delivery method of ionizable cationic lipid for lung targeting LNP system
By screening and optimizing ionizable cationic lipid molecules, a lung-targeted LNP system was constructed, which solved the problems of low lung delivery efficiency and liver enrichment in the existing mRNA delivery system, and achieved efficient lung targeting and safe mRNA delivery.
Patent Information
- Application Number
- CN202510526780.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-25
- Publication Date
- 2025-08-15
AI Technical Summary
The existing mRNA delivery system has low efficiency in lung delivery, and lipid nanoparticles are enriched in the liver in vivo, resulting in toxic reactions, making it difficult to achieve efficient lung targeting and mRNA release.
The molecular dynamics simulation and support vector machine algorithm were used to screen ionizable cationic lipid molecules suitable for lung targeting, optimize the assembly ratio and pKa regulation ability of lipid nanoparticles, and combine Monte Carlo simulation to construct a lung targeting model to optimize the assembly ratio and mRNA release efficiency of lipid nanoparticles.
It improves the targeted delivery efficiency and mRNA expression ability of lipid nanoparticles in the lungs, reduces the risk of liver enrichment, and improves the safety of treatment.
Smart Images
Figure CN120496676A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of information technology, and in particular to a method for preparing and delivering ionizable cationic lipids for a lung-targeted LNP system. Background Art
[0002] Problem background: The fields of drug delivery and nanomedicine are crucial in addressing global public health challenges, particularly the development of mRNA vaccines and therapeutic strategies for emerging respiratory infectious diseases. However, existing mRNA delivery systems have significant limitations. They primarily rely on ionizable cationic lipids, such as ALC-0315 and SM-102, which form lipid nanoparticles that accumulate primarily in the liver and have low lung delivery efficiency. This limits their application in the treatment of lung-related diseases such as acute respiratory distress syndrome, pneumonia, and viral lung injury. Furthermore, liver accumulation can trigger toxic reactions, compromising therapeutic safety. A key challenge lies in designing novel ionizable cationic lipids to optimize the lung targeting and mRNA delivery efficiency of lipid nanoparticles. Specifically, the pKa of existing lipids is insufficiently tunable, making it difficult to maintain neutrality in physiological environments and effectively charge in the acidic intracellular environment, resulting in low cellular uptake and mRNA release efficiency. Furthermore, the surface charge and component ratio of lipid nanoparticles are difficult to precisely control, limiting their accumulation and transfection capabilities in lung tissue. These technical challenges make achieving efficient lung-targeted mRNA delivery a bottleneck that needs to be addressed urgently. Therefore, how to develop new ionizable cationic lipids and improve the targeted delivery efficiency and mRNA expression ability of lipid nanoparticles in the lungs by optimizing pKa regulation, surface charge and component ratio has become a key issue in this study. Summary of the Invention
[0003] The present invention provides a method for preparing and delivering ionizable cationic lipids for lung-targeted LNP systems, which mainly includes: Obtain a pre-established database of ionizable cationic lipid molecular structures, and use molecular dynamics simulation methods to calculate the pKa values of lipid molecules and the charge distribution characteristics in an acidic environment to obtain the pKa control ability parameters and surface charge distribution data of lipid molecules; Based on the pKa control ability parameters and surface charge distribution data obtained in the first step, the support vector machine algorithm is used to classify lipid molecules. If the pKa value is between 6.0 and 7.0 and the surface charge is positive in an acidic environment, the lipid molecule is judged to be suitable for lung targeting, and a set of candidate lipid molecules is obtained; Molecular structural features are obtained from a set of candidate lipid molecules. Genetic algorithms are used to optimize the composition ratio of lipid nanoparticles. The uptake efficiency and transfection capacity of lung tissue cells are simulated to obtain the optimized composition ratio parameters. Based on the optimized component ratio parameters, a Monte Carlo simulation method was used to construct a lung-targeting model of lipid nanoparticles. The mRNA release efficiency in the lung tissue environment was calculated to obtain quantitative indicators of lung targeting ability and mRNA release efficiency. The lipid nanoparticle formula corresponding to the highest value is obtained from the quantitative indicators of lung targeting ability and mRNA release efficiency, and verified using in vitro cell experimental data. If the cell uptake efficiency and transfection ability exceed the preset threshold, the formula is judged to be the final lung-targeted mRNA delivery system, and the optimized lipid nanoparticle formula is obtained.
[0004] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: BRIEF DESCRIPTION OF THE DRAWINGS
[0005] Figure 1 The figure is a flow chart of a method for preparing and delivering ionizable cationic lipids for lung-targeted LNP systems of the present invention. DETAILED DESCRIPTION
[0006] To further understand the content of the present invention, the present invention is described in detail with reference to the accompanying drawings and examples. The present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the invention are shown in the accompanying drawings.
[0007] like Figure 1 In this embodiment, a method for preparing and delivering an ionizable cationic lipid for a lung-targeted LNP system may specifically include: Step S101, obtain a pre-established ionizable cationic lipid molecular structure database, use molecular dynamics simulation method to calculate the pKa value of lipid molecules and charge distribution characteristics in acidic environment, and obtain the pKa control ability parameters and surface charge distribution data of lipid molecules.
[0008] A molecular dynamics simulation system containing a solvation box was constructed, and simulation environment parameters with different pH values were set. A continuous protonation state sampling algorithm was used to explore the conformational space of lipid molecules. The protonation state transition data in the trajectory file was processed through the pKa calculation module to obtain the pKa value distribution curve of each lipid at a specific pH. The surface charge density distribution map was drawn, and the charge fluctuation variance of the hydrophilic head group region was calculated as a quantitative indicator of the pKa regulation ability. A symmetric matrix containing the charge optimization parameters was output.
[0009] Specifically, 100 representative molecular structures were first extracted from databases of ionizable cationic lipids, such as LIPID MAPS. These structures were converted to PDB format using the Open Babel tool and optimized for three-dimensional conformation. The force field parameters were set to GAFF2, and energy minimization was performed using AMBER software until the gradient threshold reached 0.05 kcal / mol·Å. The pKa values were then calculated using the continuous solvation method, using the COSMO-RS model to scan the pH range from 2 to 10 at intervals of 0.5. The pKa values were determined by fitting the protonation state change curves. For example, the protonation rate of a certain two-chain lipid reached 50% at pH 5.3. Molecular dynamics simulations were performed using GROMACS 2022 software, with settings of 310K temperature, 1 bar pressure, 2 fs time step, and a total simulation time of 50 ns. The PME algorithm was used to account for long-range electrostatic interactions, with a cutoff radius of 1.2 nm. By analyzing the last 10 ns of the trajectory, VMD was used to calculate the charge distribution of the lipid headgroup. For example, at pH 4, a quaternary ammonium lipid has a headgroup positive charge of +0.8e and a tail charge of -0.2e. A pKa control model was established using multiple linear regression, selecting molecular descriptors including hydrophobic carbon chain length (C12-C18), number of amino groups (1-3), and hydroxyl position (α / β). The regression equation had an R² of 0.91, with the coefficient of influence of the number of amino groups on the pKa being 0.45±0.03. Finally, principal component analysis was used to reduce the dimensionality of the charge distribution data. The cumulative contribution of the first three principal components was 85%, with the first principal component (52% of the variance) primarily reflecting the polarity of the headgroup charge.
[0010] In step S102, the lipid molecules are classified using a support vector machine algorithm based on the pKa control ability parameters and surface charge distribution data obtained in the first step. If the pKa value is in the range of 6.0 to 7.0 and the surface charge is positive in an acidic environment, the lipid molecule is judged to be suitable for lung targeting, and a set of candidate lipid molecules is obtained.
[0011] A support vector machine algorithm was used to classify the pKa control parameters and surface charge distribution data of lipid molecules. Lipid molecules with pKa values between 6.0 and 7.0 and a positive surface charge in an acidic environment were considered suitable for lung targeting, resulting in a set of candidate lipid molecules. Molecular structural features were extracted from the candidate lipid molecule set. Combined with the characteristics of phospholipid molecules in lung cell membranes, a molecular docking algorithm was used to calculate the binding affinity of the candidate lipids to lung cell membranes and identify lipid molecules with high lung cell affinity. For these high-affinity lipid molecules, a Monte Carlo simulation method was used to optimize the component ratios of the lipid nanoparticles. The particle stability parameters in the bloodstream and the permeability of the pulmonary vascular barrier were calculated to obtain the optimized lipid nanoparticle formulation. If the stability parameters of the optimized lipid nanoparticle formulation exceeded a preset threshold, a random forest algorithm was used to predict the mRNA delivery efficiency and expression level in lung cells using in vitro cell experimental data to determine the particle's lung-targeted delivery performance. Based on the judged lung-targeted delivery performance, key lipid structural features are extracted from the high-performance particle formula, and a generative adversarial network is used to generate a new lipid molecular structure. The pKa regulation ability and lung cell affinity are iteratively optimized to obtain a molecular design scheme for a new ionizable cationic lipid.
[0012] Specifically, a support vector machine algorithm was used to classify lipid molecules based on pKa regulation parameters and surface charge distribution data. The radial basis function (RBF) kernel was selected, the penalty parameter C was set to 1.0, and the kernel parameter γ was set to 0.01. Model parameters were optimized through cross-validation, achieving an accuracy of 92.3% using a five-fold cross-validation approach. The model labeled lipid molecules with pKa values between 6.0 and 7.0 and a surface charge that was positive in an acidic environment as candidates for lung targeting.
[0013] For example, a lipid molecule with a pKa value of 6.5, a headgroup charge of +0.7e at pH 5.0, and a tail chain charge of -0.1e was classified by the model as a candidate molecule for lung targeting. Further feature importance analysis revealed that the pKa value and headgroup charge contributed 0.65 and 0.35 to the classification results, respectively. Ultimately, 35 candidate lipid molecules that met the criteria were screened from the database. The hydrophobic carbon chain lengths of these molecules ranged from C14 to C18, the number of amino groups ranged from 1 to 2, and the hydroxyl group position was primarily located at the α position. Cluster analysis divided the candidate molecules into three categories, corresponding to different charge distribution patterns and pKa regulation capabilities, providing a reference for subsequent experimental verification.
[0014] Step S103, obtaining molecular structural features from the candidate lipid molecule set, optimizing the component ratio of lipid nanoparticles using a genetic algorithm, simulating the uptake efficiency and transfection ability of lung tissue cells, and obtaining optimized component ratio parameters.
[0015] A genetic algorithm was used to establish a mapping model between lipid molecular characteristics and pKa regulation ability, with molecular structural characteristics as input variables and pKa regulation ability as fitness function. If the pKa value was in the range of 6.0 to 7.5, the lipid molecule was retained in the candidate pool. Based on the characteristics of the lung cell membrane, a molecular docking algorithm was used to calculate the binding affinity of the candidate lipids to the cell membrane. Monte Carlo simulation was used to generate lipid nanoparticle configurations with different group distribution ratios. The stability parameters and permeability of the particles were calculated under each group distribution ratio. If the stability parameter was greater than the preset threshold, the uptake efficiency of the ratio was evaluated. A random forest algorithm was used to establish a prediction model for group distribution ratio and transfection ability. The group distribution ratio parameters were iteratively optimized by genetic algorithm to obtain the optimal group distribution ratio parameter set. Specifically, molecules with different hydrophilic-lipophilic balances (HLBs), such as DSPE-PEG2000 (HLB = 18), DOPE (HLB = 5), and cholesterol (HLB = 3), were first screened from a pool of candidate lipid molecules. Molecular dynamics simulations were used to calculate their critical micelle concentrations (CMCs) (0.0012 mM, 0.15 mM, and 0.003 mM, respectively) and intermolecular interaction energies (-25.6 kJ / mol, -18.3 kJ / mol, and -12.7 kJ / mol, respectively). A genetic algorithm was used to optimize the group allocation ratio, with a population size of 100, 50 iterations, a crossover probability of 0.8, a mutation probability of 0.1, and a fitness function that comprehensively considers both uptake efficiency (weight 0.6) and transfection ability (weight 0.4). During the simulation, molecular docking simulations of lung tissue-cell interactions were performed for each ratio (e.g., DSPE-PEG2000:DOPE:cholesterol = 20:55:25), calculating the binding free energy (ΔG = -32.4 kJ / mol) and the cell membrane penetration barrier (48.6 kJ / mol). A radial basis function neural network was used to establish a predictive model linking the component ratios with cellular uptake (up to 78.3%) and transfection efficiency (luciferase expression reaching 2.3×10^6 RLU / mg). The resulting Pareto optimal solution (32:43:25) was the highest-scoring ratio, with a zeta potential of +28.6 mV and a particle size distribution of 95.2±3.8 nm. Molecular dynamics analysis confirmed that this ratio exhibited a stability coefficient of 0.91 in the presence of pulmonary surfactant.
[0016] In step S104, a lung targeting model of lipid nanoparticles is constructed using the Monte Carlo simulation method based on the optimized group ratio parameters, and the mRNA release efficiency in the lung tissue environment is calculated to obtain quantitative indicators of lung targeting ability and mRNA release efficiency.
[0017] Molecular dynamics simulations were used to calculate the pKa values and surface charge distributions of lipids at different pH levels, determining their pKa controllability and charge optimization parameters. Based on these pKa controllability and charge optimization parameters, candidate lipids with pKa values between 6.0 and 7.5 were screened. A molecular docking algorithm was used to calculate the binding affinity of these candidate lipids to lung cell membranes, identifying lipid molecules with high lung cell affinity. Monte Carlo simulations were used to optimize the component ratios of the lipid nanoparticles, calculate their stability parameters in the bloodstream, and calculate their permeability through the pulmonary vascular barrier, ultimately yielding the optimized lipid nanoparticle formulation. If the stability parameters of the optimized lipid nanoparticle formulation exceeded a pre-determined threshold, a random forest algorithm was used to predict mRNA delivery efficiency and expression levels in lung cells using in vitro cell data to assess the particle's lung-targeted delivery performance. Based on this predicted lung-targeted delivery performance, key lipid structural features were extracted, and new lipid molecular structures were generated using a generative adversarial network. The pKa controllability and lung cell affinity were then iteratively optimized to yield a molecular design for a novel ionizable cationic lipid.
[0018] Specifically, when constructing a lung-targeting model for lipid nanoparticles, it is first necessary to determine the optimized component ratio parameters, such as using a cationic lipid, auxiliary lipid, and polyethylene glycol-modified lipid with a molar ratio of 50:40:10 as the basic formula. Using the Monte Carlo simulation method, 10,000 random sampling iterations were set to simulate the distribution of nanoparticles in the pulmonary capillaries, where the particle size parameters were set to 80±5 nanometers and the surface potential was +25±3 millivolts. The Brownian dynamics algorithm was used to calculate the interaction probability between nanoparticles and alveolar epithelial cells, and the diffusion coefficient was set to 3.5×10⁻. 8 cm² / s, with a capture efficiency of 68% at a binding energy threshold of -15 kT. A pH-responsive mathematical model was developed to calculate mRNA release efficiency. When the ambient pH dropped from 7.4 to 6.5, the lipid bilayer instability coefficient increased from 0.2 to 0.75. The kinetic equation for triggered mRNA release showed that 42% of the mRNA was released within the first 30 minutes, with a cumulative release of 83% within 2 hours. Finite element analysis of a 10-micron mesh of the lungs was used to quantify targeting metrics, revealing that the concentration in the right middle lobe of the lung was 4.7-fold higher than in non-targeted areas. mRNA transfection efficiency reached 61±3% in an A549 cell model. Pearson correlation analysis revealed a negative correlation between particle size and alveolar deposition efficiency (r=-0.82, p<0.01), while a positive correlation between surface potential and cellular uptake (r=0.76, p<0.05).
[0019] In step S105, the lipid nanoparticle formula corresponding to the highest value is obtained from the quantitative indicators of lung targeting ability and mRNA release efficiency, and verified using in vitro cell experimental data. If the cell uptake efficiency and transfection ability exceed the preset threshold, the formula is judged to be the final lung-targeted mRNA delivery system, and the optimized lipid nanoparticle formula is obtained.
[0020] A molecular docking algorithm was used to calculate the binding affinity of candidate lipids with pKa values ranging from 6.0 to 7.5 to lung cell membranes, and lipid molecules with high lung cell affinity were identified.
[0021] Specifically, a high-throughput screening platform was used to quantitatively evaluate the lung targeting ability and mRNA release efficiency of 100 candidate formulations from a lipid nanoparticle (LNP) library. Dynamic light scattering was used to measure particle size distribution (target range 80-120 nm) and zeta potential (threshold ±20 mV). Targeting efficiency (threshold >60%) was assessed using a microfluidic chip simulating pulmonary blood flow shear stress. For mRNA release efficiency, the encapsulation efficiency of fluorescently labeled mRNA (>90%) and the cumulative release rate (>70%) after 48 hours in phosphate-buffered saline (PBS) in vitro were used as quantitative indicators. Release curves were fitted using a nonlinear regression model, and formulations with a goodness-of-fit R² >0.95 were selected. Five Pareto-optimal solutions were identified from these two indicators using Pareto front analysis (e.g., Formulation A: 68% targeting efficiency / 82% release efficiency; Formulation B: 72% targeting efficiency / 78% release efficiency). The results were subsequently validated in the A549 human alveolar epithelial cell line, using flow cytometry to measure the cellular uptake of Cy5-labeled mRNA (threshold >50%), and fluorescence microscopy to quantify the expression efficiency of the EGFP reporter gene (threshold >40%). When Formulation B reached an uptake rate of 58% and a transfection efficiency of 45%, a decision tree algorithm was triggered to automatically determine that it met the criteria. The system then optimized the molar ratio of the ionized lipid (e.g., SM-102) to the helper lipid to 50:38.5:10:1.5 (ionized lipid: cholesterol: DSPC: PEG-lipid). The system ultimately output structural parameters (particle size 98 nm, PDI 0.12) and cryo-electron microscopy 2D classification images (vesicle integrity >95%) for this formulation as an optimized solution for the lung-targeted mRNA delivery system.
[0022] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for preparing and delivering an ionizable cationic lipid for lung-targeted LNP system, characterized in that: The method comprises: Obtain a pre-established database of ionizable cationic lipid molecular structures, and use molecular dynamics simulation methods to calculate the pKa values of lipid molecules and the charge distribution characteristics in an acidic environment to obtain the pKa control ability parameters and surface charge distribution data of lipid molecules; Based on the pKa control ability parameters and surface charge distribution data obtained in the first step, the support vector machine algorithm is used to classify lipid molecules. If the pKa value is between 6.0 and 7.0 and the surface charge is positive in an acidic environment, the lipid molecule is judged to be suitable for lung targeting, and a set of candidate lipid molecules is obtained; Molecular structural features are obtained from a set of candidate lipid molecules. Genetic algorithms are used to optimize the composition ratio of lipid nanoparticles. The uptake efficiency and transfection capacity of lung tissue cells are simulated to obtain the optimized composition ratio parameters. Based on the optimized component ratio parameters, a Monte Carlo simulation method was used to construct a lung-targeting model of lipid nanoparticles. The mRNA release efficiency in the lung tissue environment was calculated to obtain quantitative indicators of lung targeting ability and mRNA release efficiency. The lipid nanoparticle formula corresponding to the highest value is obtained from the quantitative indicators of lung targeting ability and mRNA release efficiency, and verified using in vitro cell experimental data. If the cell uptake efficiency and transfection ability exceed the preset threshold, the formula is judged to be the final lung-targeted mRNA delivery system, and the optimized lipid nanoparticle formula is obtained.
2. The method according to claim 1, characterized in that The method obtains a pre-established ionizable cationic lipid molecular structure database, uses a molecular dynamics simulation method to calculate the pKa value of the lipid molecule and the charge distribution characteristics in an acidic environment, and obtains the pKa control ability parameters and surface charge distribution data of the lipid molecule, including: A molecular dynamics simulation system containing a solvation box was constructed, and simulation environment parameters with different pH values were set. A continuous protonation state sampling algorithm was used to explore the conformational space of lipid molecules. The protonation state transition data in the trajectory file was processed through the pKa calculation module to obtain the pKa value distribution curve of each lipid at a specific pH. The surface charge density distribution map was drawn, and the charge fluctuation variance of the hydrophilic head group region was calculated as a quantitative indicator of the pKa regulation ability. A symmetric matrix containing the charge optimization parameters was output.
3. The method according to claim 1, characterized in that The lipid molecules are classified using a support vector machine algorithm based on the pKa control ability parameters and surface charge distribution data obtained in the first step. If the pKa value is within the range of 6.0 to 7.0 and the surface charge is positive in an acidic environment, the lipid molecule is judged to be suitable for lung targeting, and a set of candidate lipid molecules is obtained, including: A support vector machine algorithm was used to classify the pKa control ability parameters and surface charge distribution data of lipid molecules. If the pKa value was between 6.0 and 7.0 and the surface charge was positive in an acidic environment, the lipid molecule was judged to be suitable for lung targeting, thus obtaining a set of candidate lipid molecules. Extracting molecular structural features from the candidate lipid molecule set, combining them with the phospholipid molecular characteristics of lung cell membranes, and using a molecular docking algorithm to calculate the binding affinity between the candidate lipids and lung cell membranes, and identify lipid molecules with high lung cell affinity; Targeting high-affinity lipid molecules, the Monte Carlo simulation method was used to optimize the component ratio of lipid nanoparticles, calculate the stability parameters of the particles in the blood circulation and the permeability of the pulmonary vascular barrier, and obtain the optimized lipid nanoparticle formula; If the stability parameter of the optimized lipid nanoparticle formula is greater than a preset threshold, the random forest algorithm is used to predict the mRNA delivery efficiency and expression level of the particles in lung cells based on in vitro cell experimental data to determine the particle's lung-targeted delivery performance; Based on the judged lung-targeted delivery performance, key lipid structural features are extracted from the high-performance particle formula, and a generative adversarial network is used to generate a new lipid molecular structure. The pKa regulation ability and lung cell affinity are iteratively optimized to obtain a molecular design scheme for a new ionizable cationic lipid.
4. The method according to claim 1, wherein The molecular structure characteristics are obtained from the candidate lipid molecule set, and the component ratio of the lipid nanoparticles is optimized using a genetic algorithm. The uptake efficiency and transfection ability of lung tissue cells are simulated to obtain the optimized component ratio parameters, including: A genetic algorithm was used to establish a mapping model between lipid molecular characteristics and pKa regulation ability, with molecular structural characteristics as input variables and pKa regulation ability as fitness function. If the pKa value was in the range of 6.0 to 7.5, the lipid molecule was retained in the candidate pool. Based on the characteristics of lung cell membranes, a molecular docking algorithm was used to calculate the binding affinity of candidate lipids to cell membranes. Lipid nanoparticle configurations with different group distribution ratios were generated through Monte Carlo simulation. The stability parameters and permeability of the particles were calculated under each group distribution ratio. If the stability parameter was greater than the preset threshold, the uptake efficiency of the ratio was evaluated. A random forest algorithm was used to establish a prediction model for group distribution ratio and transfection ability. The group distribution ratio parameters were iteratively optimized through genetic algorithm to obtain the optimal group distribution ratio parameter set.
5. The method according to claim 1, wherein Based on the optimized component ratio parameters, a Monte Carlo simulation method was used to construct a lung targeting model of lipid nanoparticles, and the mRNA release efficiency in the lung tissue environment was calculated to obtain quantitative indicators of lung targeting ability and mRNA release efficiency, including: Molecular dynamics simulation technology is used to calculate the pKa value and surface charge distribution of lipids under different pH environments, and the pKa regulation ability and charge optimization parameters of lipids are obtained; Based on the pKa control ability and charge optimization parameters, candidate lipids with pKa values ranging from 6.0 to 7.5 were screened; Molecular docking algorithms were used to calculate the binding affinity of candidate lipids to lung cell membranes and identify lipid molecules with high lung cell affinity; Monte Carlo simulation was used to optimize the composition ratio of lipid nanoparticles, calculate the stability parameters of the particles in the blood circulation and the permeability of the pulmonary vascular barrier, and obtain the optimized lipid nanoparticle formula; If the stability parameter of the optimized lipid nanoparticle formula is greater than the preset threshold, the random forest algorithm is used to predict the mRNA delivery efficiency and expression level of the particles in lung cells based on in vitro cell experimental data to determine the lung-targeted delivery performance of the particles; Based on the judged lung-targeted delivery performance, key lipid structural features were extracted, and a generative adversarial network was used to generate new lipid molecular structures. The pKa regulation ability and lung cell affinity were iteratively optimized to obtain a molecular design scheme for a new type of ionizable cationic lipid.
6. The method according to claim 1, characterized in that The lipid nanoparticle formula corresponding to the highest value obtained from the quantitative indicators of lung targeting ability and mRNA release efficiency is verified using in vitro cell experimental data. If the cell uptake efficiency and transfection ability exceed the preset threshold, the formula is judged to be the final lung-targeted mRNA delivery system, and the optimized lipid nanoparticle formula is obtained, including: A molecular docking algorithm was used to calculate the binding affinity of candidate lipids with pKa values ranging from 6.0 to 7.5 to lung cell membranes, and lipid molecules with high lung cell affinity were identified.