Multi-target joint optimization antibiotic molecule design system and method

Abstract

The invention discloses a multi-target joint optimized antibiotic molecule design system and method applied to the field of biological medicine, and the method comprises the steps: obtaining FabI enzyme and AcrAB-TolC efflux pump structure data, extracting double-target space and physicochemical characteristics, and constructing a joint pharmacophore model containing a FabI inhibition domain and an efflux pump blocking domain; screening candidate molecules meeting a double-domain space matching threshold, screening high-affinity double-target molecules through conformation sampling and free energy calculation, and forming a lead compound library through multi-layer filtration of metabolic stability, membrane penetrability, quantum chemical properties and synthesis feasibility; carrying out electron effect, chain length or heteroatom fine tuning on dominant molecules through in-vitro bacteriostasis and efflux inhibition experiment feedback, and carrying out iterative optimization until MIClt is met; 2 [mu] g / mL and the efflux inhibition rate is gt; according to the present invention, the antibacterial effect and the drug resistance inhibition ability are significantly improved, the molecular synthesizability and the biological safety are ensured, and the engineering design new path is provided for the Gram-negative bacterium drug resistance crisis.

Description

Technical Field

[0001] This invention relates to the field of biomedicine, and in particular to a system and method for multi-target combined optimization of antibiotic molecular design. Background Technology

[0002] Drug-loaded contact lenses (DLCLs) are a novel ocular drug delivery system that combines drugs with a corneal contact lens. Their core advantage lies in achieving long-lasting and precise drug release on the corneal surface, avoiding the drawbacks of traditional eye drops such as low bioavailability and the need for frequent administration, thus providing a revolutionary approach to the treatment of ocular infections. However, the clinical value of drug-loaded contact lenses is highly dependent on "what drug is delivered"—that is, the drug molecule used as the drug delivery target must possess potent antibacterial activity, low risk of drug resistance, and good biocompatibility. This requirement places stringent demands on the generation and screening technologies of candidate drug molecules. Among common ocular infectious pathogens, Gram-negative bacteria, due to their prominent drug resistance, have become a key focus and challenge in the development of drug delivery for drug-loaded contact lenses, urgently requiring targeted and highly effective antibiotic molecules to support their practical application.

[0003] With the accelerating spread of antibiotic resistance globally, traditional single-target antibacterial drug development strategies are facing severe challenges. Current mainstream antibiotic design largely focuses on high-affinity inhibition of a single enzyme or receptor, with its core mechanism relying on precise matching and binding energy optimization of specific target structures. However, under evolutionary pressure, bacteria readily bypass single mechanisms of action through target gene mutations, activation of efflux pump systems, or metabolic bypass remodeling, leading to rapid drug inactivation. Particularly in Gram-negative bacteria, multiple efflux pump systems such as AcrAB–TolC actively pump intracellular drugs out, significantly reducing effective intracellular concentrations; while point mutations in key metabolic enzymes like Fabi directly weaken drug binding ability. The synergistic effect of these two factors renders single-target strategies ineffective in terms of broad-spectrum coverage and delaying resistance, resulting in an increasingly narrow clinical treatment window.

[0004] Multi-target combined intervention is considered an important direction for overcoming drug resistance bottlenecks. Its basic principle lies in simultaneously acting on multiple key pathways on which bacterial survival depends, increasing the cost of mutation escape and inhibiting the efficiency of efflux pumps. Ideally, drug molecules should have a strong inhibitory ability on the Fabi enzyme active site while circumventing the AcrAB–TolC efflux recognition characteristics, thereby maintaining a stable and effective therapeutic concentration locally in the cornea, synergizing with the long-acting release characteristics of the lens. However, existing technologies still have fundamental defects in multi-target synergistic optimization: most schemes only perform linear superposition of target activities, lacking dynamic quantification of the contribution weight of each target; no physicochemical boundary model for efflux pump avoidance has been established, resulting in molecules that are highly active in vitro but ineffective in vivo; and no combined constraint mechanism for off-target toxicity and pharmacokinetic parameters has been introduced, resulting in low overall druggability of candidate molecules, failing to meet the special requirements of drug-loaded corneal contact lenses for ocular biosafety.

[0005] Current technologies fail to construct a joint optimization framework with adaptive multi-target weights, making it impossible to simultaneously balance target inhibitory efficacy, efflux avoidance capability, and safety thresholds during the molecular design stage. Especially in complex scenarios involving increased heterogeneity of clinical strains and dynamic evolution of drug resistance mechanisms, static weight allocation and single-dimensional scoring systems are prone to causing optimization deviations, making it difficult to produce candidate molecules with broad-spectrum antibacterial activity, low risk of drug resistance induction, and favorable pharmacological properties. This approach is inadequate to address the Gram-negative bacterial resistance crisis and hinders the application of drug-eluting contact lenses in the treatment of ocular infections. There is an urgent need to establish an intelligent molecular design method that integrates dynamic weighting mechanisms, efflux feature constraints, and multi-dimensional threshold control. Summary of the Invention

[0006] The core of this invention lies in constructing a dual-target synergistic model to simultaneously optimize the molecule's ability to inhibit the active site of the Fabi enzyme and its ability to escape or inhibit the efflux pump AcrAB–TolC. This achieves the triple goals of prolonging the molecule's residence time in the cell, enhancing its bactericidal efficacy, and inhibiting the rate of drug resistance evolution, thereby addressing the fundamental defects exposed by target drug design strategies in dealing with the evolution of bacterial drug resistance.

[0007] To solve the above problems, the present invention adopts the following technical solution.

[0008] A multi-target combined optimization method for antibiotic molecular design includes the following steps: S1. Construct a dual-target pharmacophore model of FabI and AcrAB–TolC, and generate an initial molecular library based on the principle of functional domain synergistic matching: S11. Obtain the crystal structure data of the FaB1 protein of the target pathogen species and the topological configuration data of the transmembrane region of the AcrAB–TolC efflux pump complex through the target structure data acquisition module. S12. Based on the FaBI protein crystal structure data, the spatial geometric constraint parameters, hydrophobicity distribution map, and set of hydrogen bond donor and acceptor site coordinates of its coenzyme binding pocket are extracted through the FaBI binding feature extraction module. S13. Based on the transmembrane topology data of the AcrAB–TolC efflux pump complex, the amino acid residue arrangement sequence, charge gradient distribution characteristics, and flexible hinge point location information of the drug transport channel inner wall were extracted using the AcrAB–TolC channel feature extraction module. S14. Construct a dual-target pharmacophore model using the dual-target pharmacophore modeling module. The pharmacophore model includes a first functional domain and a second functional domain. The first functional domain corresponds to the space-occupying group, hydrogen bond anchoring group and hydrophobic chimeric group required for FabI inhibition. The second functional domain corresponds to the cation enrichment group, rigid framework support unit and side chain spatial extension arm required for AcrAB–TolC channel blocking. S2. Perform conformational flexibility sampling and binding free energy calculation on the initial candidate molecules to screen out a set of candidate molecules that simultaneously satisfy the high affinity of FabI and the low efflux tendency of AcrAB–TolC: S21. The initial candidate molecule screening module screens the chemical molecule database for an initial set of candidate molecules that simultaneously meet the spatial matching threshold of the first functional domain and the second functional domain. The spatial matching threshold is set to be less than 0.3 nanometers between the pharmacophore feature point and the center of the molecule atom and less than 15 degrees in angle. S22. Perform conformational flexible sampling on each molecule in the initial candidate molecule set through the conformational flexible sampling and energy calculation module to generate no less than one thousand low-energy conformational isomers, and calculate the binding free energy of each low-energy conformational isomer in the FabI binding pocket and the AcrAB–TolC channel cavity. S23. Molecules with binding free energy values ​​below -40 kJ / mol at the FabI target and below -35 kJ / mol at the AcrAB–TolC target were selected as dual-target high-affinity candidate molecules using the dual-target high-affinity molecule determination module. S3. Perform metabolic penetration assessment, quantum chemical verification, and synthetic feasibility scoring on the candidate molecule set to form a priority synthesis sequence and assemble a lead compound library: S31. Metabolic stability prediction and cell membrane penetration assessment of dual-target high-affinity candidate molecules are performed through the metabolism and penetration assessment module, and molecules with a half-life of less than two hours or a transmembrane permeability coefficient of less than 10 to the power of negative six centimeters per second are eliminated. S32. Perform frontier orbital energy level calculations at the quantum chemical level on the retained molecules through the quantum chemical property verification module to ensure that the highest occupied molecular orbital energy level is higher than -8 electron volts and the lowest unoccupied molecular orbital energy level is lower than -2 electron volts. S33. The synthesis feasibility of the molecules that have passed the above screening is evaluated by the synthesis feasibility scoring module. The weighted calculation is based on the functional group complexity, the number of chiral centers and the ring strain index. Molecules with scores higher than 80 are retained to form a priority synthesis sequence and form a lead compound library. S4. Conduct in vitro activity verification experiments, and fine-tune and iterate the molecular structure based on experimental feedback until an optimized molecule that meets the preset efficacy and drug resistance control indicators is obtained: S41. The minimum inhibitory concentration (MIC) and efflux pump inhibition rate of molecules in the lead compound library were determined in vitro using the in vitro activity verification module. The changes in the diameter of the inhibition zone and the percentage decrease in efflux pump activity against standard strains and clinically resistant isolates of Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa were recorded. S42. Based on the experimental data feedback, the structure of molecules in the lead compound library whose inhibition zone diameter has increased by more than 30% and whose efflux pump activity has decreased by more than 50% is fine-tuned through the structure fine-tuning and iterative optimization module. The structure fine-tuning includes electronic effect regulation of benzene ring substituents, adding or removing one carbon atom from the length of the aliphatic chain, and replacing oxygen or nitrogen atoms with heteroatoms. S43. Re-input the structure-fine-tuned molecule into the conformation sampling and binding free energy calculation process, and iterate and optimize until the final candidate molecule output module outputs the final candidate molecule. The final candidate molecule must show a minimum inhibitory concentration of less than 2 micrograms per milliliter against at least three types of Gram-negative bacteria in three independent experiments and a stable efflux pump inhibition rate of more than 60%.

[0009] Furthermore, the specific operation of S12 includes the following process: S12a. Extract the three-dimensional coordinates of the FaB1 protein from the crystal structure file numbered EYI in the protein database; S12b, the length of the coenzyme-binding pocket is calculated to be 1.2 nm, the width of the short axis is 0.9 nm, and the depth is 0.7 nm; S12c: Identify twelve hydrophobic hotspot regions, each with a surface area of ​​not less than 0.05 square nanometers; S12d, locate the coordinates of the hydroxyl or amino nitrogen atom of the side chain of the hydrogen bond donor residue and the coordinates of the carbonyl oxygen or carboxylic acid oxygen atom of the hydrogen bond acceptor, and construct a three-dimensional coordinate set.

[0010] Furthermore, the specific operation of S13 includes the following process: S13a. Based on the three-dimensional reconstruction model with a resolution of 2.8 Å using cryo-electron microscopy, the repeating unit sequences of aspartic acid, arginine, phenylalanine, glutamine, histidine, and leucine arranged sequentially from the inside to the outside of the membrane were extracted. S13b. Calculate the charge gradient distribution characteristics from the cytoplasmic side to the periplasmic side, where the net positive charge density decreases by 40%. S13c, with glycine at position 105, proline at position 217, and serine at position 342 positioned as flexible hinge points in the dynamic gating region.

[0011] Furthermore, the specific operation of S14 includes the following process: S14a. Define the space-occupying group of the first functional domain as an aromatic fused ring structure with a volume greater than 0.15 cubic nanometers, the hydrogen bond anchoring group as a carboxylate group or sulfonamide group that can form bidentate coordination, and the hydrophobic chimeric group as a straight-chain alkyl or cycloalkyl segment with no less than six carbon atoms. S14b defines the cation enrichment group of the second functional domain as a quaternary ammonium salt structure or a guanidine structure, the rigid framework support unit as a biphenyl or naphthyl bridging structure, and the side chain spatial extension arm as a branched alkoxy or aryloxy with a halogen substituent at the end. S14c: Align the spatial vectors of the first functional domain and the second functional domain in three-dimensional space to ensure that there is no spatial conflict between the functional domains and that the cooperative inhibition geometric constraints are satisfied.

[0012] Furthermore, the specific operation of S21 includes the following procedures: S21a. Load a chemical molecule database containing more than five million pre-processed small molecule structures, with each molecule labeled with atom type, bond order, formal charge and number of rotatable bonds; S21b: Perform a gridded spatial overlap matching algorithm on each molecule to traverse the conformation space in 0.1 nanometer steps; S21c: Calculate the distance and bond angle deviation between the molecular atomic center and the pharmacophore feature point, and retain molecules with a distance of less than 0.3 nanometers and an angle deviation of less than 15 degrees. S21d outputs an initial set of candidate molecules that meet the bifunctional domain matching condition, with the number controlled between five and eight thousand.

[0013] Furthermore, the specific operation of S22 includes the following procedures: S22a. The Monte Carlo annealing algorithm is adopted, with the initial temperature set to 1,000 Kelvin, the cooling step size being 10 Kelvin, and 100 samples taken for each temperature layer. S22b. Perform no less than one thousand conformational samplings on each molecule and record the lowest-energy conformational isomer. S22c. Perform molecular mechanics Poisson-Boltzmann surface area calculations for each conformational isomer. The implicit water model is selected as the solvation model, the dielectric constant is set to 80, and the nonbonded cutoff radius is set to 1.2 nanometers. S22d outputs the binding free energy values ​​of each molecule in the FabI binding pocket and the AcrAB–TolC channel cavity.

[0014] Furthermore, the specific operations of S23 include the following procedures: S23a. Perform dual threshold determination on the FabI binding free energy and AcrAB–TolC binding free energy of each molecule; S23b. Retain molecules that simultaneously satisfy the following conditions: FabI binding free energy is less than -40 kJ / mol and AcrAB–TolC binding free energy is less than -35 kJ / mol. S23c outputs a set of dual-target high-affinity candidate molecules, reducing the number to three to five hundred.

[0015] Furthermore, the specific operation of S31 includes the following procedures: S31a, based on the cytochrome P450 enzyme system metabolic site recognition model to predict molecular half-life; S31b uses the parallel artificial membrane permeability measurement method to simulate values ​​to evaluate the transmembrane permeability coefficient; S31c, eliminating molecules with a half-life of less than two hours or a transmembrane permeability coefficient of less than 10 to the power of negative six centimeters per second; S31d retains candidate molecules that pass the metabolism and penetration assessment for the next screening stage.

[0016] Furthermore, the specific operations of S32 include the following procedures: S32a. Density functional theory B-LYP functionals are adopted, the basis set is 6-31 G stars, and the self-consistent field convergence threshold is set to 10 to the power of negative 6 Hartley. S32b, Calculate the highest occupied molecular orbital energy level and the lowest unoccupied molecular orbital energy level; S32c, retaining molecules whose highest occupied molecular orbital energy level is higher than -8 electron volts and whose lowest unoccupied molecular orbital energy level is lower than -2 electron volts; S32d, remove molecules whose frontier orbital energy levels do not meet the redox inertness requirements.

[0017] Furthermore, the specific operation of S33 includes the following process: S33a. Calculate the composite feasibility score based on the functional group complexity coefficient, the number of chiral centers, and the toroidal tension index. The formula is: score equals one hundred minus the functional group complexity coefficient multiplied by five, minus the number of chiral centers multiplied by three, and minus the toroidal tension index multiplied by two. S33b, the functional group complexity coefficient is accumulated based on the number of rare functional groups, and the ring strain index is determined by the ring size and the degree of substituent crowding. S33c, molecules with retention scores higher than 80 are included in the lead compound library; S33d outputs a lead compound library, with the number of molecules controlled between fifty and eighty.

[0018] Furthermore, the specific operation of S41 includes the following procedures: S41a. The minimum inhibitory concentration was determined by the broth microdilution method. The concentration of the inoculated bacterial solution was 5 x 10^5 colony forming units per milliliter. The incubation temperature was 37 degrees Celsius and the incubation time was 16 hours. S41b. The inhibition rate of the efflux pump was determined by the fluorescence substrate accumulation method, with an excitation wavelength of 490 nm and an emission wavelength of 520 nm. The changes in the diameter of the inhibition zone and the percentage decrease in efflux pump activity against standard strains and clinically resistant isolates of Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa were recorded. S41c molecules that exhibit an increase in inhibition zone diameter of more than 30% and a decrease in efflux pump activity of more than 50% enter the structural fine-tuning stage.

[0019] Furthermore, the specific operation of S42 includes the following process: S42a. Electronic effect regulation is applied to the substituents of the benzene ring by introducing nitro, cyano, or trifluoromethyl groups as electron-withdrawing groups, or introducing methoxy, amino, or hydroxy groups as electron-donating groups. S42b, Performing an operation to add or remove one carbon atom in the length of an aliphatic chain, limited to straight or branched structures with four to ten carbon atoms in the main chain; S42c: Perform heteroatom substitution, allowing only carbon atoms to be replaced with oxygen or nitrogen atoms; S42d, output the molecular set after structural fine-tuning, and re-enter the conformation sampling and binding free energy calculation process.

[0020] Furthermore, the specific operation of S43 includes the following process: S43a. Repeat the screening and evaluation process of claims 5 to 11 on the molecule after structural fine-tuning; S43b: Verify that the molecule has a minimum inhibitory concentration of less than 2 micrograms per milliliter against at least three Gram-negative bacteria in three independent experiments. S43c, the verification molecule showed a stable efflux pump inhibition rate of over 60% in three independent experiments; S43d outputs the final candidate molecule that meets all efficacy and drug resistance control criteria.

[0021] Furthermore, the final candidate molecules must meet the following requirements: a half-life of not less than four hours in human serum, a clearance rate of less than 20 μL / min / mg protein in liver microsomes, and no significant cumulative toxicity in renal tubular epithelial cells.

[0022] The multi-target co-optimized antibiotic molecule design system includes a target structure data acquisition module, a FabI binding feature extraction module, an AcrAB–TolC channel feature extraction module, a dual-target pharmacophore modeling module, an initial candidate molecule screening module, a conformational flexibility sampling and energy calculation module, a dual-target high-affinity molecule identification module, a metabolism and penetration ability assessment module, a quantum chemical property verification module, a synthesis feasibility scoring module, an in vitro activity verification module, a structure fine-tuning and iterative optimization module, and a final candidate molecule output module.

[0023] Compared with the prior art, the advantages of this invention are: (1) This scheme constructs a dual-pathway synergistic inhibition mechanism by simultaneously targeting the key bacterial fatty acid synthesis enzyme FabI and the efflux pump complex AcrAB–TolC, thereby achieving dual blockade of the growth and metabolic pathways and drug efflux capacity of drug-resistant strains.

[0024] (2) Establish a structure-activity relationship model at the molecular level, and combine three-dimensional conformation space sampling and energy field matching algorithm to screen and optimize the lead compound skeleton that has both high affinity for Fapi binding site occupancy and strong AcrAB–TolC channel blocking properties, thereby breaking through the technical bottleneck that single-target drugs are easily escaped by mutation or eliminated by efflux, and significantly improving the antibacterial spectrum coverage and drug resistance evolution inhibition efficacy. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall technical architecture of the multi-target joint optimization antibiotic molecular design method (FabI + AcrAB–TolC) proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the dual-target joint pharmacophore model construction and functional domain synergistic matching in this invention; Figure 3 This is a flowchart illustrating the logical process of initial candidate molecule screening, conformational flexible sampling, and binding free energy calculation in this invention. Figure 4 This is a diagram of the multidimensional screening logic framework for metabolic penetration assessment, quantum chemical verification, and synthetic feasibility scoring in this invention. Figure 5 This is a closed-loop feedback process framework diagram of in vitro activity verification and structural fine-tuning iterative optimization in this invention; Figure 6 This is a schematic diagram of the multi-level interaction relationship and data flow between the FabI target and the AcrAB–TolC external pump target in this invention. Detailed Implementation

[0026] The technical solutions will now be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention.

[0027] First implementation method: like Figure 1 and Figure 6 A multi-target combined optimization method for antibiotic molecular design includes the following steps: S1. Construct a dual-target pharmacophore model of FabI and AcrAB–TolC, and generate an initial molecular library based on the principle of functional domain synergistic matching; S11. Obtain the crystal structure data of the FaB1 protein of the target pathogen species and the topological configuration data of the transmembrane region of the AcrAB–TolC efflux pump complex through the target structure data acquisition module. S12. Based on the FaBI protein crystal structure data, the spatial geometric constraint parameters, hydrophobicity distribution map, and set of hydrogen bond donor and acceptor site coordinates of its coenzyme binding pocket are extracted through the FaBI binding feature extraction module. S12a. Extract the three-dimensional coordinates of the FaB1 protein from the crystal structure file numbered EYI in the protein database; S12b, the length of the coenzyme-binding pocket is calculated to be 1.2 nm, the width of the short axis is 0.9 nm, and the depth is 0.7 nm; S12c: Identify twelve hydrophobic hotspot regions, each with a surface area of ​​not less than 0.05 square nanometers; S12d, locate the coordinates of the hydroxyl or amino nitrogen atom of the side chain of the hydrogen bond donor residue and the coordinates of the carbonyl oxygen or carboxylic acid oxygen atom of the hydrogen bond acceptor, and construct a three-dimensional coordinate set.

[0028] S13. Based on the transmembrane topology data of the AcrAB–TolC efflux pump complex, the amino acid residue arrangement sequence, charge gradient distribution characteristics, and flexible hinge point location information of the drug transport channel inner wall were extracted using the AcrAB–TolC channel feature extraction module. S13a. Based on the three-dimensional reconstruction model with a resolution of 2.8 Å using cryo-electron microscopy, the repeating unit sequences of aspartic acid, arginine, phenylalanine, glutamine, histidine, and leucine arranged sequentially from the inside to the outside of the membrane were extracted. S13b. Calculate the charge gradient distribution characteristics from the cytoplasmic side to the periplasmic side, where the net positive charge density decreases by 40%. S13c, positioning glycine at position 105, proline at position 217, and serine at position 342 as flexible hinge points in the dynamic gating region. S14. Construct a dual-target pharmacophore model using the dual-target pharmacophore modeling module. The pharmacophore model includes a first functional domain and a second functional domain. The first functional domain corresponds to the space-occupying group, hydrogen bond anchoring group and hydrophobic chimeric group required for FabI inhibition. The second functional domain corresponds to the cation enrichment group, rigid framework support unit and side chain spatial extension arm required for AcrAB–TolC channel blocking. S14a. Define the space-occupying group of the first functional domain as an aromatic fused ring structure with a volume greater than 0.15 cubic nanometers, the hydrogen bond anchoring group as a carboxylate group or sulfonamide group that can form bidentate coordination, and the hydrophobic chimeric group as a straight-chain alkyl or cycloalkyl segment with no less than six carbon atoms. S14b defines the cation enrichment group of the second functional domain as a quaternary ammonium salt structure or a guanidine structure, the rigid framework support unit as a biphenyl or naphthyl bridging structure, and the side chain spatial extension arm as a branched alkoxy or aryloxy with a halogen substituent at the end. S14c: Align the spatial vectors of the first functional domain and the second functional domain in three-dimensional space to ensure that there is no spatial conflict between the functional domains and that the cooperative inhibition geometric constraints are satisfied.

[0029] S2. Perform conformational flexibility sampling and binding free energy calculation on the initial candidate molecules to screen out a set of candidate molecules that simultaneously satisfy the high affinity of FabI and the low exotropy tendency of AcrAB–TolC. S21. The initial candidate molecule screening module screens the chemical molecule database for an initial set of candidate molecules that simultaneously meet the spatial matching threshold of the first functional domain and the second functional domain. The spatial matching threshold is set to be less than 0.3 nanometers between the pharmacophore feature point and the center of the molecule atom and less than 15 degrees in angle. S21a. Load a chemical molecule database containing more than five million pre-processed small molecule structures, with each molecule labeled with atom type, bond order, formal charge and number of rotatable bonds; S21b: Perform a gridded spatial overlap matching algorithm on each molecule to traverse the conformation space in 0.1 nanometer steps; S21c: Calculate the distance and bond angle deviation between the molecular atomic center and the pharmacophore feature point, and retain molecules with a distance of less than 0.3 nanometers and an angle deviation of less than 15 degrees. S21d outputs an initial set of candidate molecules that meet the bifunctional domain matching condition, with the number controlled between five and eight thousand.

[0030] S22. Perform conformational flexible sampling on each molecule in the initial candidate molecule set through the conformational flexible sampling and energy calculation module to generate no less than one thousand low-energy conformational isomers, and calculate the binding free energy of each low-energy conformational isomer in the FabI binding pocket and the AcrAB–TolC channel cavity. S22a. The Monte Carlo annealing algorithm is adopted, with the initial temperature set to 1,000 Kelvin, the cooling step size being 10 Kelvin, and 100 samples taken for each temperature layer. S22b. Perform no less than one thousand conformational samplings on each molecule and record the lowest-energy conformational isomer. S22c. Perform molecular mechanics Poisson-Boltzmann surface area calculations for each conformational isomer. The implicit water model is selected as the solvation model, the dielectric constant is set to 80, and the nonbonded cutoff radius is set to 1.2 nanometers. S22d outputs the binding free energy values ​​of each molecule in the FabI binding pocket and the AcrAB–TolC channel cavity.

[0031] S23. Molecules with binding free energy values ​​below -40 kJ / mol at the FabI target and below -35 kJ / mol at the AcrAB–TolC target were selected as dual-target high-affinity candidate molecules using the dual-target high-affinity molecule determination module. S23a. Perform dual threshold determination on the FabI binding free energy and AcrAB–TolC binding free energy of each molecule; S23b. Retain molecules that simultaneously satisfy the following conditions: FabI binding free energy is less than -40 kJ / mol and AcrAB–TolC binding free energy is less than -35 kJ / mol. S23c outputs a set of dual-target high-affinity candidate molecules, reducing the number to three to five hundred.

[0032] S3. Perform metabolic penetration assessment, quantum chemical verification, and synthetic feasibility scoring on the candidate molecule set to form a preferred synthesis sequence; S31. Metabolic stability prediction and cell membrane penetration assessment of dual-target high-affinity candidate molecules are performed through the metabolism and penetration assessment module, and molecules with a half-life of less than two hours or a transmembrane permeability coefficient of less than 10 to the power of negative six centimeters per second are eliminated. S31a, based on the cytochrome P450 enzyme system metabolic site recognition model to predict molecular half-life; S31b uses the parallel artificial membrane permeability measurement method to simulate values ​​to evaluate the transmembrane permeability coefficient; S31c, eliminating molecules with a half-life of less than two hours or a transmembrane permeability coefficient of less than 10 to the power of negative six centimeters per second; S31d retains candidate molecules that pass the metabolism and penetration assessment for the next screening stage.

[0033] S32. Perform frontier orbital energy level calculations at the quantum chemical level on the retained molecules through the quantum chemical property verification module to ensure that the highest occupied molecular orbital energy level is higher than -8 electron volts and the lowest unoccupied molecular orbital energy level is lower than -2 electron volts. S31a, based on the cytochrome P450 enzyme system metabolic site recognition model to predict molecular half-life; S31b uses the parallel artificial membrane permeability measurement method to simulate values ​​to evaluate the transmembrane permeability coefficient; S31c, eliminating molecules with a half-life of less than two hours or a transmembrane permeability coefficient of less than 10 to the power of negative six centimeters per second; S31d retains candidate molecules that pass the metabolism and penetration assessment for the next screening stage.

[0034] S33. The synthesis feasibility of the molecules that have passed the above screening is evaluated by the synthesis feasibility scoring module. The weighted calculation is based on the functional group complexity, the number of chiral centers and the ring strain index. Molecules with scores higher than 80 are retained to form a priority synthesis sequence and form a lead compound library. S33a. Calculate the composite feasibility score based on the functional group complexity coefficient, the number of chiral centers, and the toroidal tension index. The formula is: score equals one hundred minus the functional group complexity coefficient multiplied by five, minus the number of chiral centers multiplied by three, and minus the toroidal tension index multiplied by two. S33b, the functional group complexity coefficient is accumulated based on the number of rare functional groups, and the ring strain index is determined by the ring size and the degree of substituent crowding. S33c, molecules with retention scores higher than 80 are included in the lead compound library; S33d outputs a lead compound library, with the number of molecules controlled between fifty and eighty.

[0035] S4. Conduct in vitro activity verification experiments, and fine-tune and iterate the molecular structure based on experimental feedback until an optimized molecule that meets the preset efficacy and drug resistance control indicators is obtained. S41. The minimum inhibitory concentration (MIC) and efflux pump inhibition rate of molecules in the lead compound library were determined in vitro using the in vitro activity verification module. The changes in the diameter of the inhibition zone and the percentage decrease in efflux pump activity against standard strains and clinically resistant isolates of Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa were recorded. S41a. The minimum inhibitory concentration was determined by the broth microdilution method. The concentration of the inoculated bacterial solution was 5 x 10^5 colony forming units per milliliter. The incubation temperature was 37 degrees Celsius and the incubation time was 16 hours. S41b. The inhibition rate of the efflux pump was determined by the fluorescence substrate accumulation method, with an excitation wavelength of 490 nm and an emission wavelength of 520 nm. The changes in the diameter of the inhibition zone and the percentage decrease in efflux pump activity against standard strains and clinically resistant isolates of Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa were recorded. S41c molecules that exhibit an increase in inhibition zone diameter of more than 30% and a decrease in efflux pump activity of more than 50% enter the structural fine-tuning stage.

[0036] S42. Based on the experimental data feedback, the structure of molecules in the lead compound library whose inhibition zone diameter has increased by more than 30% and whose efflux pump activity has decreased by more than 50% is fine-tuned through the structure fine-tuning and iterative optimization module. The structure fine-tuning includes electronic effect regulation of benzene ring substituents, adding or removing one carbon atom from the length of the aliphatic chain, and replacing oxygen or nitrogen atoms with heteroatoms. S42a. Electronic effect regulation is applied to the substituents of the benzene ring by introducing nitro, cyano, or trifluoromethyl groups as electron-withdrawing groups, or introducing methoxy, amino, or hydroxy groups as electron-donating groups. S42b, Performing an operation to add or remove one carbon atom in the length of an aliphatic chain, limited to straight or branched structures with four to ten carbon atoms in the main chain; S42c: Perform heteroatom substitution, allowing only carbon atoms to be replaced with oxygen or nitrogen atoms; S42d, output the molecular set after structural fine-tuning, and re-enter the conformation sampling and binding free energy calculation process.

[0037] S43. Re-input the structure-fine-tuned molecule into the conformation sampling and binding free energy calculation process, and iterate and optimize until the final candidate molecule output module outputs the final candidate molecule. The final candidate molecule must show a minimum inhibitory concentration of less than 2 micrograms per milliliter against at least three Gram-negative bacteria in three independent experiments and a stable efflux pump inhibition rate of more than 60%. S43a. Repeat the screening and evaluation process of claims 5 to 11 on the molecule after structural fine-tuning; S43b: Verify that the molecule has a minimum inhibitory concentration of less than 2 micrograms per milliliter against at least three Gram-negative bacteria in three independent experiments. S43c, the verification molecule showed a stable efflux pump inhibition rate of over 60% in three independent experiments; S43d: Output the final candidate molecule that meets all efficacy and drug resistance control criteria. The final candidate molecule must have a half-life of not less than four hours in human serum, a clearance rate of less than 20 μL / min / mg protein in liver microsomes, and no significant accumulation toxicity in renal tubular epithelial cells.

[0038] The design system for antibiotic molecular design based on multi-target joint optimization includes a target structure data acquisition module, a FabI binding feature extraction module, an AcrAB–TolC channel feature extraction module, a dual-target pharmacophore modeling module, an initial candidate molecule screening module, a conformational flexibility sampling and energy calculation module, a dual-target high-affinity molecule identification module, a metabolism and penetration ability assessment module, a quantum chemical property verification module, a synthesis feasibility scoring module, an in vitro activity verification module, a structure fine-tuning and iterative optimization module, and a final candidate molecule output module.

[0039] In terms of specific implementation, step S1 is implemented as follows: The pharmacophore model is an abstract expression of molecular recognition and binding ability. Its essence is the topological encoding of the spatial arrangement and chemical characteristics of key interaction sites in the target's active pocket. In single-target models, the pharmacophore only reflects the binding requirements of a single protein, neglecting the multiple biological barriers faced by molecules in the cellular environment. This application, for the first time, incorporates the substrate recognition domain characteristics of the efflux pump AcrAB–TolC into the pharmacophore construction scope, forming a dual-target joint model.

[0040] like Figure 2 and Figure 6 First, the crystal structure of the FaB1 enzyme was obtained from a protein structure database. The spatial configuration of its cofactor binding pocket and substrate channel was analyzed, and hydrogen bond donors, hydrogen bond acceptors, hydrophobic centers, aromatic ring stacking sites, and positive and negative charge distribution regions were extracted to construct a FaB1-specific pharmacophore feature vector. Simultaneously, for the AcrAB–TolC efflux pump, key residue clusters known to interact with antibiotic molecules in its transmembrane transport channel were selected, including the phenylalanine cluster located in the hydrophobic pocket of the AcrB subunit and the charged amino acid ring located in the TolC efflux channel, to construct an efflux-inhibiting pharmacophore. Its feature vector includes spatial repulsion volume, polar shielding region, flexible side chain avoidance region, and charge neutralization site.

[0041] Subsequently, the FAI pharmacophore and the AcrAB–TolC pharmacophore were functionally matched in three-dimensional space. The matching criteria were: the molecular skeleton must simultaneously meet the space-filling requirements of the FAI active pocket, and the distribution of its peripheral substituents must avoid high-affinity binding sites in the AcrAB–TolC recognition domain. A gridded spatial overlap algorithm was used for the matching process, traversing the molecular conformation space in 0.1 nm increments. The compatibility score of the dual pharmacophore feature vectors at each grid point was calculated, and only conformations with compatibility scores higher than a preset threshold were retained as valid conformations. Based on the set of valid conformations, a fragment growth algorithm was used, starting from the core aromatic ring or heterocyclic structure, to sequentially add substituents that satisfied the dual pharmacophore constraints, including alkyl chains, halogen atoms, amino groups, carboxyl groups, sulfonyl groups, and nitrogen-containing heterocycles, ultimately generating an initial molecular library containing five to eight thousand molecules. Each molecule in this initial molecular library was structurally pre-defined with potential inhibition of FAI and a low recognition probability of efflux pumps, laying a physicochemical foundation for subsequent screening.

[0042] The specific implementation process of step S2 is as follows: Although the static pharmacophore model can provide spatial matching guidance, it cannot accurately quantify the dynamic interaction energy between the molecule and the target. For example... Figure 3To improve screening accuracy, this invention introduces an evaluation mechanism combining conformational flexibility sampling and free energy perturbation calculation. First, molecular dynamics simulations are performed on each molecule in the initial candidate library, sampling its flexible conformational trajectory on a nanosecond timescale in an explicit solvent environment. The ten representative conformations with the lowest energies are extracted as inputs for subsequent calculations. Then, for the FabI target, a molecular docking procedure is used to dock each conformation to the FabI active pocket, and its binding free energy is calculated, expressed by the formula: ΔG_bind = ΔE_vdw + ΔE_elec + ΔG_solv + ΔG_conf Where ΔE_vdw represents the van der Waals interaction energy, ΔE_elec represents the electrostatic interaction energy, ΔG_solv represents the change in solvation free energy, and ΔG_conf represents the conformational entropy loss term. During docking, the spatial distance between the molecular core framework and the FabI catalytic triplet His–Tyr–Lys is forcibly constrained to within 0.3 nm to ensure the binding mode has reasonable catalytic inhibition. For the AcrAB–TolC efflux pump, a coarse-grained model of its transmembrane channel is constructed, the molecular conformation is placed at the channel entrance, a constant electric field is applied to simulate the physiological membrane potential, and the molecule's residence time and crossing energy barrier are recorded. Molecules with a residence time shorter than a preset threshold and a crossing energy barrier lower than 10 kilojoules per mole are judged to have a low efflux tendency. Finally, the screening criteria are set as follows: FabI binding free energy lower than -40 kilojoules per mole, and AcrAB–TolC crossing energy barrier higher than 15 kilojoules per mole or residence time longer than 5 nanoseconds. Molecules that meet both requirements are included in the candidate set, which is usually reduced to three to five hundred.

[0043] The specific implementation process of step S3 is as follows: Molecules with excellent in vitro binding capacity may not necessarily exert their therapeutic effects in the living environment; their ability to penetrate cell membranes, resist metabolic degradation, and the feasibility of chemical synthesis are all key limiting factors. For example... Figure 4This invention constructs a multi-dimensional evaluation system to comprehensively rank candidate molecules. Metabolic penetration assessment combines the five rules of drug-likeness with a membrane permeability prediction model to calculate the molecule's lipid-water partition coefficient, total number of hydrogen bond donors and acceptors, molecular weight, and polar surface area, predicting its permeation efficiency in the outer and inner membranes of Gram-negative bacteria. Quantum chemical verification focuses on the stability of the molecular electronic structure, using density functional theory to calculate the highest occupied molecular orbital energy level, the lowest unoccupied molecular orbital energy level, and the frontier orbital gap, ensuring that the molecule does not undergo spontaneous redox reactions or photolytic breakage under physiological conditions. Synthetic feasibility scoring is based on a retrosynthetic analysis engine, breaking down the molecular structure into commercially available starting materials, assessing the number of synthetic route steps, the stability of key intermediates, and the difficulty of constructing chiral centers, assigning each molecule a synthetic feasibility score from 0 to 100. The three evaluation results are weighted and summed, with the weights allocated as follows: metabolic penetration 40%, quantum stability 30%, and synthetic feasibility 30%. Molecules with a total score higher than 80 are included in the priority synthesis sequence, with the number controlled between 50 and 80.

[0044] The specific implementation process of step S4 is as follows: The calculated prediction needs to be verified experimentally to confirm its biological significance. For example... Figure 5 This invention establishes a closed-loop feedback mechanism, delivering molecules from the prioritized synthesis sequence to a chemical synthesis platform for milligram-level sample preparation. Subsequently, minimum inhibitory concentration (MIC) determination, time-based bactericidal curve analysis, and drug resistance induction experiments are conducted. MIC determination employs a broth microdilution method to test the inhibitory ability of molecules against representative Gram-negative strains such as *Escherichia coli*, *Klebsiella pneumoniae*, and *Pseudomonas aeruginosa*. Time-based bactericidal curves record viable bacterial counts at different time points after drug application, assessing the bactericidal rate and duration of action. Drug resistance induction experiments monitor the MIC drift rate of strains through continuous passage exposure to subinhibitory drug concentrations. Experimental data are fed back to the molecular design module. If a molecule achieves the target inhibitory activity at FAI but its efflux phenotype does not improve, its AcrAB–TolC channel retention simulation trajectory is traced back to identify structural fragments leading to high penetration, such as excessively long flexible side chains or strongly polar groups, which are then replaced with rigid ring systems or neutral substituents. If the molecule synthesis yield is less than 20%, synthetic route reconstruction is initiated, introducing a protecting group strategy or altering the cyclization sequence. The fine-tuned molecule is then cycled through steps S2 to S4 until an optimized molecule is obtained that exhibits a minimum inhibitory concentration (MIC) of less than 2 μg / mL across all tested strains, a 99% bactericidal time of less than four hours, and an MIC increase of less than fourfold after twenty consecutive passages. This molecule is the final dual-target synergistic optimized antibiotic candidate compound output by this invention.

[0045] In a preferred embodiment of the present invention, the dual-target co-pharmacophore model constructed in step S1 further incorporates a dynamic residue flexibility compensation mechanism. Specifically, during the construction of the FabI pharmacophore, not only are the static residue positions in the crystal structure considered, but also the side-chain swing range of key residues such as Tyr157 and Lys165 in the active pocket is analyzed through molecular dynamics trajectory analysis. This expands the pharmacophore feature points into a spherical tolerance domain with a certain radius, allowing the molecule to undergo slight displacement within this domain without losing the binding score. In the AcrAB–TolC pharmacophore, a mutation compatibility sub-model is constructed for residues in the AcrB subunit known to have drug-resistant mutations, such as Gly171 and Asn274, requiring candidate molecules to maintain low affinity in both wild-type and mutant channels. This mechanism significantly improves the model's robustness to natural variations in the target and avoids false positive screening caused by a single conformational assumption.

[0046] In another preferred embodiment of the present invention, the calculation of binding free energy in step S2 incorporates a solvation entropy correction term. The traditional MM / PBSA method only considers polar and nonpolar solvation energies when calculating ΔG_solv, neglecting the contribution of solvent molecule rearrangement entropy. The present invention uses a three-dimensional reference interaction site model to calculate the solvation entropy change, and the formula is modified as follows: ΔG_bind = ΔE_vdw + ΔE_elec + ΔG_polar + ΔG_nonpolar - TΔS_solvent Where TΔS_solvent is the solvation entropy term, obtained by statistically analyzing the change in the degree of order of the hydration shell on the ligand surface before and after binding of solvent molecules. This correction makes the predicted free energy value closer to the isothermal titration calorimetric experimental data, with the error controlled within ±5 kJ / mol.

[0047] In another preferred embodiment of the invention, the quantum chemical verification in step S3 includes excited-state stability analysis. In addition to the ground-state frontier orbital gap, the energies of the molecule's first singlet and triplet excited states are calculated to ensure they are above the photon energy threshold of the physiological environment, preventing photosensitization toxicity. Simultaneously, natural bond orbital analysis is used to identify strong charge transfer pathways within the molecule, avoiding irreversible electron capture in a reducing cellular environment.

[0048] In another preferred embodiment of the invention, the drug resistance induction experiment in step S4 includes monitoring the expression level of the efflux pump. After each generation of subculture, total RNA is extracted from the strain, and the expression level of the acrB gene is determined by quantitative reverse transcription polymerase chain reaction. If the expression level increases by more than twofold, the molecular test is immediately terminated, indicating that the strain has strong efflux pump induction potential, and it is not accepted even if its initial minimum inhibitory concentration meets the target. This measure blocks the drug resistance evolution pathway at its source, ensuring that the optimized molecule has long-term clinical application potential.

[0049] The above description is merely a preferred embodiment of the present invention; it encompasses all the protection scope of the present invention. Any equivalent substitutions or modifications made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solutions and improved concepts of the present invention, should be covered within the protection scope of the present invention.

Claims

1. A multi-target combined optimization method for antibiotic molecular design, characterized by: Includes the following steps: S1. Construct a dual-target pharmacophore model of FabI and AcrAB–TolC, and generate an initial molecular library based on the principle of functional domain synergistic matching: S11. Obtain the crystal structure data of the FaB1 protein of the target pathogen species and the topological configuration data of the transmembrane region of the AcrAB–TolC efflux pump complex through the target structure data acquisition module. S12. Based on the FabI protein crystal structure data, extract the spatial geometric constraint parameters, hydrophobic distribution map, and set of hydrogen bond donor and acceptor site coordinates of its coenzyme binding pocket through the FabI binding feature extraction module. S13. Based on the transmembrane topology data of the AcrAB–TolC efflux pump complex, extract the amino acid residue arrangement sequence, charge gradient distribution characteristics, and flexible hinge point location information of the drug transport channel inner wall using the AcrAB–TolC channel feature extraction module. S14. Construct a dual-target pharmacophore model using a dual-target pharmacophore modeling module. The pharmacophore model includes a first functional domain and a second functional domain. The first functional domain corresponds to the space-occupying group, hydrogen bond anchoring group, and hydrophobic chimeric group required for FabI inhibition. The second functional domain corresponds to the cation-enriching group, rigid framework support unit, and side chain spatial extension arm required for AcrAB–TolC channel blocking. S2. Perform conformational flexibility sampling and binding free energy calculation on the initial candidate molecules to screen out a set of candidate molecules that simultaneously satisfy the high affinity of FabI and the low efflux tendency of AcrAB–TolC: S21. The initial candidate molecule screening module screens the chemical molecule database for an initial set of candidate molecules that simultaneously meet the spatial matching threshold of the first functional domain and the second functional domain. The spatial matching threshold is set to be less than 0.3 nanometers between the pharmacophore feature point and the center of the molecule atom and less than 15 degrees in angle. S22. Perform conformational flexible sampling on each molecule in the initial candidate molecule set through the conformational flexible sampling and energy calculation module to generate no less than one thousand low-energy conformational isomers, and calculate the binding free energy of each low-energy conformational isomer in the FabI binding pocket and the AcrAB–TolC channel cavity. S23. Molecules with binding free energy values ​​below -40 kJ / mol at the FabI target and below -35 kJ / mol at the AcrAB–TolC target were selected as dual-target high-affinity candidate molecules using the dual-target high-affinity molecule determination module. S3. Perform metabolic penetration assessment, quantum chemical verification, and synthetic feasibility scoring on the candidate molecule set to form a priority synthesis sequence and assemble a lead compound library: S31. The metabolic stability and cell membrane penetration ability of the dual-target high-affinity candidate molecules are predicted and evaluated by the metabolism and penetration ability assessment module, and molecules with a half-life of less than two hours or a transmembrane permeability coefficient of less than 10 to the power of negative six centimeters per second are eliminated. S32. Perform frontier orbital energy level calculations at the quantum chemical level on the retained molecules through the quantum chemical property verification module to ensure that the highest occupied molecular orbital energy level is higher than -8 electron volts and the lowest unoccupied molecular orbital energy level is lower than -2 electron volts. S33. The synthesis feasibility of the molecules that have passed the above screening is evaluated by the synthesis feasibility scoring module. The weighted calculation is based on the functional group complexity, the number of chiral centers and the ring strain index. Molecules with scores higher than 80 are retained to form a priority synthesis sequence and form a lead compound library. S4. Conduct in vitro activity verification experiments, and fine-tune and iterate the molecular structure based on experimental feedback until an optimized molecule that meets the preset efficacy and drug resistance control indicators is obtained: S41. The in vitro minimum inhibitory concentration and efflux pump inhibition rate of the molecules in the lead compound library are determined by the in vitro activity verification module. The changes in the diameter of the inhibition zone and the percentage decrease in efflux pump activity against standard strains and clinical drug-resistant isolates of Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa are recorded. S42. Based on experimental data feedback, the structure of molecules in the lead compound library whose inhibition zone diameter increases by more than 30% and whose efflux pump activity decreases by more than 50% is fine-tuned through the structure fine-tuning and iterative optimization module. The structure fine-tuning includes electronic effect regulation of benzene ring substituents, adding or removing one carbon atom from the length of the aliphatic chain, and replacing oxygen or nitrogen atoms with heteroatoms. S43. Re-input the structure-fine-tuned molecule into the conformation sampling and binding free energy calculation process, and iterate and optimize until the final candidate molecule output module outputs the final candidate molecule. The final candidate molecule must show a minimum inhibitory concentration of less than 2 micrograms per milliliter against at least three types of Gram-negative bacteria and a stable efflux pump inhibition rate of more than 60% in three independent experiments.

2. The method for multi-target combined optimization of antibiotic molecular design according to claim 1, characterized in that: The specific operation of step S12 includes the following process: S12a. Extract the three-dimensional coordinates of the FaB1 protein from the crystal structure file numbered EYI in the protein database; S12b, the length of the coenzyme-binding pocket is calculated to be 1.2 nm, the width of the short axis is 0.9 nm, and the depth is 0.7 nm; S12c: Identify twelve hydrophobic hotspot regions, each with a surface area of ​​not less than 0.05 square nanometers; S12d, locate the coordinates of the hydroxyl or amino nitrogen atom of the side chain of the hydrogen bond donor residue and the coordinates of the carbonyl oxygen or carboxylic acid oxygen atom of the hydrogen bond acceptor, and construct a three-dimensional coordinate set.

3. The method for multi-target joint optimization of antibiotic molecular design according to claim 2, characterized in that: The specific operation of step S13 includes the following process: S13a. Based on the three-dimensional reconstruction model with a resolution of 2.8 Å using cryo-electron microscopy, the repeating unit sequences of aspartic acid, arginine, phenylalanine, glutamine, histidine, and leucine arranged sequentially from the inside to the outside of the membrane were extracted. S13b. Calculate the charge gradient distribution characteristics from the cytoplasmic side to the periplasmic side, where the net positive charge density decreases by 40%. S13c, with glycine at position 105, proline at position 217, and serine at position 342 positioned as flexible hinge points in the dynamic gating region.

4. The method for multi-target joint optimization of antibiotic molecular design according to claim 3, characterized in that: The specific operation of step S14 Includes the following processes: S14a. Define the space-occupying group of the first functional domain as an aromatic fused ring structure with a volume greater than 0.15 cubic nanometers, the hydrogen bond anchoring group as a carboxylate group or sulfonamide group that can form bidentate coordination, and the hydrophobic chimeric group as a straight-chain alkyl or cycloalkyl segment with no less than six carbon atoms. S14b defines the cation enrichment group of the second functional domain as a quaternary ammonium salt structure or a guanidine structure, the rigid framework support unit as a biphenyl or naphthyl bridging structure, and the side chain spatial extension arm as a branched alkoxy or aryloxy with a halogen substituent at the end. S14c: Align the spatial vectors of the first functional domain and the second functional domain in three-dimensional space to ensure that there is no spatial conflict between the functional domains and that the cooperative inhibition geometric constraints are satisfied.

5. The method for multi-target joint optimization of antibiotic molecular design according to claim 4, characterized in that: The specific operation of step S21 includes the following process: S21a. Load a chemical molecule database containing more than five million pre-processed small molecule structures, with each molecule labeled with atom type, bond order, formal charge and number of rotatable bonds; S21b: Perform a gridded spatial overlap matching algorithm on each molecule to traverse the conformation space in 0.1 nanometer steps; S21c: Calculate the distance and bond angle deviation between the molecular atomic center and the pharmacophore feature point, and retain molecules with a distance of less than 0.3 nanometers and an angle deviation of less than 15 degrees. S21d outputs an initial set of candidate molecules that meet the bifunctional domain matching condition, with the number controlled between five and eight thousand.

6. The method for multi-target combined optimization of antibiotic molecular design according to claim 5, characterized in that: The specific operation of step S22 includes the following process: S22a. The Monte Carlo annealing algorithm is adopted, with the initial temperature set to 1,000 Kelvin, the cooling step size being 10 Kelvin, and 100 samples taken for each temperature layer. S22b. Perform no less than one thousand conformational samplings on each molecule and record the lowest-energy conformational isomer. S22c. Perform molecular mechanics Poisson-Boltzmann surface area calculations for each conformational isomer. The implicit water model is selected as the solvation model, the dielectric constant is set to 80, and the nonbonded cutoff radius is set to 1.2 nanometers. S22d outputs the binding free energy values ​​of each molecule in the FabI binding pocket and the AcrAB–TolC channel cavity.

7. The method for multi-target combined optimization of antibiotic molecular design according to claim 6, characterized in that: The specific operation of step S23 includes the following process: S23a. Perform dual threshold determination on the FabI binding free energy and AcrAB–TolC binding free energy of each molecule; S23b. Retain molecules that simultaneously satisfy the following conditions: FabI binding free energy is less than -40 kJ / mol and AcrAB–TolC binding free energy is less than -35 kJ / mol. S23c outputs a set of dual-target high-affinity candidate molecules, reducing the number to three to five hundred.

8. The method for multi-target combined optimization of antibiotic molecular design according to claim 7, characterized in that: The specific operation of step S31 includes the following process: S31a, based on the cytochrome P450 enzyme system metabolic site recognition model to predict molecular half-life; S31b uses the parallel artificial membrane permeability measurement method to simulate values ​​to evaluate the transmembrane permeability coefficient; S31c, eliminating molecules with a half-life of less than two hours or a transmembrane permeability coefficient of less than 10 to the power of negative six centimeters per second; S31d retains candidate molecules that pass the metabolism and penetration assessment for the next screening stage.

9. The method for multi-target combined optimization of antibiotic molecular design according to claim 8, characterized in that: The specific operation of step S32 includes the following process: S32a. Density functional theory B-LYP functionals are adopted, the basis set is 6-31 G stars, and the self-consistent field convergence threshold is set to 10 to the power of negative 6 Hartley. S32b, Calculate the highest occupied molecular orbital energy level and the lowest unoccupied molecular orbital energy level; S32c, retaining molecules whose highest occupied molecular orbital energy level is higher than -8 electron volts and whose lowest unoccupied molecular orbital energy level is lower than -2 electron volts; S32d, remove molecules whose frontier orbital energy levels do not meet the redox inertness requirements.

10. The method for multi-target combined optimization of antibiotic molecular design according to claim 9, characterized in that: The specific operation of step S33 includes the following process: S33a. Calculate the composite feasibility score based on the functional group complexity coefficient, the number of chiral centers, and the toroidal tension index. The formula is: score equals one hundred minus the functional group complexity coefficient multiplied by five, minus the number of chiral centers multiplied by three, and minus the toroidal tension index multiplied by two. S33b, the functional group complexity coefficient is accumulated based on the number of rare functional groups, and the ring strain index is determined by the ring size and the degree of substituent crowding. S33c, molecules with retention scores higher than 80 are included in the lead compound library; S33d outputs a lead compound library, with the number of molecules controlled between fifty and eighty.

11. The method for multi-target combined optimization of antibiotic molecular design according to claim 10, characterized in that: The specific operation of step S41 includes the following process: S41a. The minimum inhibitory concentration was determined by the broth microdilution method. The concentration of the inoculated bacterial solution was 5 x 10^5 colony forming units per milliliter. The incubation temperature was 37 degrees Celsius and the incubation time was 16 hours. S41b. The inhibition rate of the efflux pump was determined by the fluorescence substrate accumulation method, with an excitation wavelength of 490 nm and an emission wavelength of 520 nm. The changes in the diameter of the inhibition zone and the percentage decrease in efflux pump activity against standard strains and clinically resistant isolates of Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa were recorded. S41c molecules that exhibit an increase in inhibition zone diameter of more than 30% and a decrease in efflux pump activity of more than 50% enter the structural fine-tuning stage.

12. The method for multi-target joint optimization of antibiotic molecular design according to claim 11, characterized in that: The specific operation of step S42 includes the following process: S42a. Electronic effect regulation is applied to the substituents of the benzene ring by introducing nitro, cyano, or trifluoromethyl groups as electron-withdrawing groups, or introducing methoxy, amino, or hydroxy groups as electron-donating groups. S42b, Performing an operation to add or remove one carbon atom in the length of an aliphatic chain, limited to straight or branched structures with four to ten carbon atoms in the main chain; S42c: Perform heteroatom substitution, allowing only carbon atoms to be replaced with oxygen or nitrogen atoms; S42d, output the molecular set after structural fine-tuning, and re-enter the conformation sampling and binding free energy calculation process.

13. The method for multi-target combined optimization of antibiotic molecular design according to claim 12, characterized in that: The specific operation of step S43 includes the following process: S43a. Repeat the screening and evaluation process of claims 5 to 11 on the molecule after structural fine-tuning; S43b: Verify that the molecule has a minimum inhibitory concentration of less than 2 micrograms per milliliter against at least three Gram-negative bacteria in three independent experiments. S43c, the verification molecule showed a stable efflux pump inhibition rate of over 60% in three independent experiments; S43d outputs the final candidate molecule that meets all efficacy and drug resistance control criteria.

14. The method for multi-target combined optimization of antibiotic molecular design according to claim 13, characterized in that: The final candidate molecules must meet the following requirements: a half-life of not less than four hours in human serum; a clearance rate of less than 20 μL / min / mg protein in liver microsomes; and no significant cumulative toxicity in renal tubular epithelial cells.

15. The design system for the multi-target joint optimization antibiotic molecular design method according to claim 14, characterized in that: It includes a target structure data acquisition module, a FaBI binding feature extraction module, an AcrAB–TolC channel feature extraction module, a dual-target pharmacophore modeling module, an initial candidate molecule screening module, a conformational flexibility sampling and energy calculation module, a dual-target high-affinity molecule identification module, a metabolism and penetration ability assessment module, a quantum chemical property verification module, a synthesis feasibility scoring module, an in vitro activity verification module, a structure fine-tuning and iterative optimization module, and a final candidate molecule output module.

Images