Aromatic deep eutectic solvent multi-scale molecular simulation screening method and system based on a multi-objective collaborative scoring model

Abstract

The present application belongs to the field of computational chemistry and molecular simulation technology, and relates to a kind of aromatic eutectic solvent multiscale molecular simulation screening method and system based on multi-objective collaborative scoring model.Establish target molecule model library and eutectic solvent candidate component library;Quantum chemistry calculation is carried out on target molecules and eutectic solvent candidate system, and quantum chemistry descriptors are extracted;Molecular dynamics simulation system is constructed, and molecular dynamics descriptors such as aromatic space enrichment coefficient Kπ, glycosidic bond interaction index Ggly and lignin bond type affinity index Lβ are extracted;Multi-scale feature vectors are obtained by normalization processing, input into multi-objective collaborative scoring model, calculate and extract score SE, grading score SF, and judge and output active ingredient extraction priority, lignocellulose grading priority or segmented collaborative processing mode.The present application is suitable for eutectic solvent calculation screening and application mode discrimination when there is a target conflict between plant aromatic active ingredient extraction and lignocellulose grading treatment.

Description

Technical Field

[0001] This invention belongs to the field of computational chemistry and molecular simulation technology, specifically relating to a multi-scale molecular simulation screening method and system for aromatic eutectic solvents based on a multi-objective co-scoring model. In particular, it relates to a design method for aromatic eutectic solvents based on quantum chemical calculations, molecular dynamics simulations, multi-scale descriptor extraction, and a multi-objective co-scoring model. This method is applicable to the computational screening and application mode discrimination of eutectic solvents when there is an objective conflict between the extraction of plant aromatic active ingredients and the lignocellulose grading process. Background Technology

[0002] Plant biomass systems typically contain polyhydroxy aromatic active ingredients and complex lignocellulose skeletons. Extraction of aromatic active ingredients requires solvents with appropriate affinity for the target aromatic structures and avoids strong acid, strong oxidizing, or excessively solvating environments that could lead to degradation of the active ingredients. Lignocellulose fractionation, on the other hand, requires solvents to disrupt the hydrogen bond network between hemicellulose, lignin, and cellulose, as well as the lignin-carbohydrate complex structure, while minimizing excessive damage to cellulose and the risk of lignin condensation. Therefore, active ingredient extraction and lignocellulose fractionation are not simple optimization problems in the same direction, but rather involve a multi-objective conflict between "high affinity extraction," "strong fractionation treatment," and "low damage risk."

[0003] Eutectic solvents are a class of homogeneous, designable, and green solvent systems formed by hydrogen bond acceptors and donors in a specific molar ratio through hydrogen bonding, electrostatic interactions, van der Waals interactions, and other non-covalent interactions. They possess characteristics such as low volatility, tunable composition, and designable solubility, and have been used for the extraction of aromatic active ingredients such as plant polyphenols, flavonoids, and anthocyanins, as well as for the pretreatment of lignocellulose biomass, lignin dissolution, hemicellulose removal, and cellulose enrichment. However, screening eutectic solvent systems that possess both high extraction efficiency of aromatic active ingredients and good lignocellulose fractionation effects remains a challenge.

[0004] Existing molecular simulation or machine learning-assisted screening methods for eutectic solvents typically optimize the extraction rate of a single target component, mainly predicting extraction efficiency through target component-eutectic solvent binding energy, target component-matrix binding energy, selectivity index, or molecular polarity parameters. While these methods can reduce some experimental screening work, they struggle to simultaneously evaluate the interrelationships between aromatic active ingredient extraction, hemicellulose removal, lignin dissolution, cellulose damage risk, and active ingredient degradation risk. This is especially true for ternary eutectic solvents containing aromatic hydrogen bond donors, where π-π stacking and CH-π interactions introduced by aromatic rings, glycosidic interactions induced by acidic hydrogen bond donors, and selective affinity for lignin bonds coexist. Therefore, relying solely on a single binding energy or extraction rate cannot accurately determine the application mode of candidate solvents.

[0005] Therefore, there is an urgent need to establish a multi-objective screening method and system that can start from computational chemical descriptors and distinguish between the compatibility of active ingredient extraction and the compatibility of lignocellulose grading. Summary of the Invention

[0006] To address the problems of existing molecular simulation screening methods, such as over-reliance on single interaction energies, difficulty in simultaneously predicting the extraction efficiency of aromatic active ingredients and the lignocellulose grading effect, and neglect of the stability and process adaptability of active ingredients, this invention proposes a multi-scale molecular simulation screening method and system for aromatic eutectic solvents based on a multi-objective synergistic scoring model. This method quantitatively identifies the interactions between candidate solvents and plant aromatic active ingredients, polysaccharide fragments, and lignin bond types through quantum chemical and molecular dynamics simulations. Furthermore, the multi-objective synergistic scoring model distinguishes between extraction-priority, grading-priority, and segmented synergistic processing modes, thereby providing a preparable, verifiable, and process-adaptable eutectic solvent design scheme before experiments.

[0007] The technical solution of this invention is: This invention provides a multi-scale molecular simulation screening method for aromatic eutectic solvents based on a multi-objective collaborative scoring model, comprising the following steps: (1) Establish a target molecule model library and a eutectic solvent candidate component library; (2) Perform quantum chemical calculations on the target molecule and the candidate system of eutectic solvent, and extract quantum chemical descriptors, including electrostatic potential, molecular polarity, IGMH weak interaction, AIM bond critical point, and BSSE corrected interaction energy; (3) Construct a molecular dynamics simulation system containing the target molecule and the eutectic solvent component, perform equilibrium simulation and production simulation on the molecular dynamics simulation system, and obtain dynamic descriptors, including the average number of hydrogen bonds, radial distribution function, aromatic spatial enrichment coefficient Kπ, glycosidic bond interaction index Ggly and lignin bond type affinity index Lβ; (4) Normalize the quantum chemical descriptor from step (2) and the kinetic descriptor from step (3) to form a multi-scale feature vector; (5) The multi-objective collaborative scoring model reads multi-scale feature vectors and calculates the active ingredient extraction adaptation index A, lignocellulose grading adaptation index F, process adaptation index T, and risk penalty term R respectively, and further calculates the active ingredient extraction score SE and lignocellulose grading score SF; wherein, the calculation formula for SE is: SE = 0.60A + 0.25T - 0.15Ract; The formula for calculating SF is: SF = 0.60F + 0.20T - 0.20Rfrac; Among them, Ract represents the degradation risk of active ingredients, and Rfrac represents the lignocellulose grading risk; (6) Based on SE and SF and their difference, the candidate systems of eutectic solvents are sorted and classified, and the output mode is determined, including the active ingredient extraction priority mode, the lignocellulose grading priority mode, the combined treatment mode or the segmented synergistic treatment mode.

[0008] Furthermore, in step (1), the target molecule model library includes plant aromatic active ingredients, polysaccharide fragments, and lignin bond types, wherein the polysaccharide fragments include glucan fragments and xylan fragments, and the lignin bond types include one or more of β-O-4, 4-O-5, β-1, and β-5 linkage structures; The eutectic solvent candidate component library includes hydrogen bond acceptors, aliphatic hydrogen bond donors, aromatic hydrogen bond donors, and candidate molar ratios; wherein, the hydrogen bond acceptor is choline chloride, the aliphatic hydrogen bond donor is ethylene glycol or glycerol, and the aromatic hydrogen bond donor is guaiacol or p-toluenesulfonic acid; when the eutectic solvent candidate system is a binary system, the candidate molar ratio is 1:2, and when the eutectic solvent candidate system is a ternary system, the candidate molar ratio is 1:2:1.

[0009] Furthermore, in step (2), the quantum chemical calculation includes the following steps: The target molecule is input into the quantum chemistry calculation module for geometric optimization and single-point energy calculation, resulting in the optimized structure file, energy file and wave function file of the target molecule. Based on the candidate systems of eutectic solvents, initial molecular clusters of eutectic solvents are constructed; the optimized target molecules are combined with the initial molecular clusters of eutectic solvents to construct eutectic solvent-target molecule composite clusters. The initial molecular clusters of the eutectic solvent and the eutectic solvent-target molecular complex clusters are input into the conformation search program. The lowest energy conformation is selected as the initial structure for semi-empirical quantum chemical screening and then optimized. Density functional theory calculations were performed on the molecular clusters and composite clusters after initial screening. Geometric optimization and single-point energy calculations were performed on the initial molecular clusters in eutectic solvent. Basis set overlap error correction was performed on the eutectic solvent-target molecular composite clusters. Quantum chemical descriptors were extracted based on the generated quantum chemical wavefunction data.

[0010] Furthermore, in step (2), the electrostatic potential descriptor includes the maximum positive electrostatic potential ESPmax, the minimum negative electrostatic potential ESPmin, the electrostatic potential range ΔESP, and the local electrostatic potential. Molecular polarity descriptors include molecular polarity index (MPI), polar surface area (PSA), and polar surface area ratio (FPSA). The AIM bond critical point includes the bond critical point electron density ρ(rBCP), the bond critical point electron density Laplace value ∇²ρ(rBCP), the energy density E(rBCP), and the hydrogen bond binding energy; The BSSE-corrected interaction energy is the interaction energy corrected for basis set overlap error and is decomposed into electrostatic interaction energy Eele, van der Waals interaction energy EvdW, repulsion energy Erep, and total interaction energy Etotal.

[0011] Furthermore, in step (3), the aromatic plane of the target aromatic active ingredient or lignin aromatic unit is used as the reference plane. The probability density ρπ of the aromatic ring centroid of the aromatic hydrogen bond donor in the region of 0.30-0.50 nm above and below the plane is calculated, and the average probability density ρ0 of the aromatic ring centroid in the entire simulation box is calculated. The aromatic spatial enrichment coefficient Kπ is calculated according to the formula Kπ=ρπ / ρ0.

[0012] Furthermore, in step (3), the formula for calculating the glycosidic bond interaction index Ggly is: Ggly=0.40N⁺[ggly(rp)]+0.30N⁺(NHBgly)+0.30N⁺(-Eele,gly) Where ggly(rp) is the intensity of the first peak of the radial distribution function of the key atom in the eutectic solvent relative to the oxygen in the glycosidic bond, NHBgly is the average number of hydrogen bonds near the glycosidic bond, Eele,gly is the electrostatic interaction energy between the eutectic solvent and the fragment near the glycosidic bond, and N + This is a favorable normalization function.

[0013] Furthermore, in step (3), the formula for calculating the lignin bond affinity index Lβ is: Lβ=0.40N⁺(-Etotal,β)+0.30N⁺(Kπ,lignin)+0.30N⁺(NHBβ) Wherein, Etotal,β is the total interaction energy between the eutectic solvent and the corresponding lignin bond type segment; the corresponding lignin bond type segment is β-O-4, β-1, β-5 or 4-O-5; Kπ,lignin is the aromatic space enrichment coefficient of the aromatic hydrogen bond donor near the lignin aromatic unit; NHBβ is the average number of hydrogen bonds formed between the eutectic solvent and the vicinity of the lignin bond type; and N⁺ is the favorable normalization function.

[0014] Furthermore, in step (6), when SE≥0.70 and SE-SF≥0.10, the active ingredient extraction priority mode is output; when SF≥0.70 and SF-SE≥0.10, the lignocellulose grading priority mode is output; when SE≥0.60, SF≥0.60 and |SE-SF|<0.10, the balanced processing mode is output; each eutectic solvent candidate system is sorted from high to low according to SE to obtain the first active ingredient extraction candidate system, and sorted from high to low according to SF to obtain the first lignocellulose grading candidate system; when the first active ingredient extraction candidate system and the first lignocellulose grading candidate system are not the same eutectic solvent candidate system, and the first active ingredient extraction candidate system satisfies SE≥0.70 and the first lignocellulose grading candidate system satisfies SF≥0.70, the segmented collaborative processing mode is output.

[0015] This invention also provides a multi-scale molecular simulation screening system for aromatic eutectic solvents based on a multi-objective collaborative scoring model, comprising: The model input module is used to build a target molecule model library and a eutectic solvent candidate component library; The quantum chemical calculation module is used to perform quantum chemical calculations on target molecules and eutectic solvent candidate systems and output quantum chemical descriptors. The molecular dynamics simulation module is used to construct molecular dynamics simulation systems that include target molecules and eutectic solvent components, and output molecular dynamics descriptors; The multi-scale feature extraction and fusion module is used to receive quantum chemical descriptors and dynamic descriptors, normalize them, group and fuse them, and output multi-scale feature vectors. The multi-objective collaborative scoring module is used to read multi-scale feature vectors and calculate the active ingredient extraction score (SE) and the lignocellulose grading score (SF). The mode discrimination and output module is used to determine the preferred mode based on SE, SF and the difference between the two, and output the eutectic solvent formulation, molar ratio and application mode.

[0016] The beneficial effects of this invention are: (1) The present invention expands the computational object from a single target molecule to a composite target system of "plant aromatic active ingredients - polysaccharide fragments - lignin bond fragments", which can simultaneously evaluate the selective effect of eutectic solvents on different biomass components.

[0017] (2) The screening method provided by the present invention is based on the target conflict between the extraction of aromatic active ingredients and the graded treatment of lignocellulose. By introducing the spatial enrichment factor Kπ of aromatic hydrogen bond donors, the π-π stacking and CH-π interaction are transformed from qualitative structural interpretation into calculable and sortable descriptors, avoiding the judgment of the extraction ability of aromatic active ingredients based solely on the total binding energy. By introducing the polysaccharide glycosidic bond interaction index Ggly and the lignin bond type affinity index Lβ, the risk of hemicellulose removal, lignin dissolution and cellulose damage are evaluated separately, thereby avoiding the mistaken equation of "strong graded ability" with "comprehensive solvent optimization".

[0018] (3) The present invention uses the difference between the active ingredient extraction score (SE) and the lignocellulose grading score (SF) to determine the extraction priority, grading priority or segmented synergistic processing modes in the same candidate library, which is suitable for the high-value utilization of complex plant biomass.

[0019] (4) The method of the present invention can distinguish the selective effects of aromatic eutectic solvents on different components of plant aromatic active ingredients and lignocellulose before the experiment, reduce empirical screening and erroneous formulation verification, and improve the interpretability and feasibility of eutectic solvent design. Through comparative verification, the system of the present invention can avoid the misselection of candidate solvents caused by interaction energy models or graded intensity models alone, and has a clearer computational chemistry effect and experimental verifiability. Attached Figure Description

[0020] Figure 1 The flowchart of the multi-scale molecular simulation screening method for aromatic eutectic solvents based on a multi-objective collaborative scoring model provided by this invention is shown.

[0021] Figure 2 This is a schematic diagram of the module structure of the computational chemistry design simulation screening system provided by the present invention.

[0022] Figure 3 This is a schematic diagram illustrating the calculation relationships of A, F, T, R, SE, and SF in the multi-objective collaborative scoring model provided by this invention.

[0023] Figure 4 This is a schematic diagram illustrating the comparison and verification logic between the method of the present invention and Comparative Examples 1-3. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Here, some terms or nouns, as well as computing devices and software environments involved in this invention, are explained. The following related explanations are optional and can be combined with the technical solutions of the embodiments of this invention in any way, all of which fall within the protection scope of the embodiments of this invention.

[0026] Hydrogen bond acceptor: A component capable of accepting hydrogen bond protons. In this embodiment, choline chloride is preferably the hydrogen bond acceptor.

[0027] Hydrogen bond donor: A component capable of providing hydrogen bond protons. In this embodiment, hydrogen bond donors include ethylene glycol, glycerol, guaiacol, and p-toluenesulfonic acid.

[0028] Aromatic hydrogen bond donor: A eutectic solvent component containing both an aromatic ring and a hydrogen bond donor group. In this embodiment, guaiacol and p-toluenesulfonic acid are selected as aromatic hydrogen bond donors.

[0029] Density functional theory calculations: quantum chemical calculation methods used to obtain information on molecular configuration, wave function, electrostatic potential, and non-covalent interactions.

[0030] Molecular dynamics simulation: An atomic-scale simulation method for calculating the evolution of target molecules and solvent components over time in a periodic simulation box.

[0031] Radial distribution function: A function that statistically measures the probability of a specified atom or group appearing at different distances around a target site.

[0032] Spatial distribution function: A function that statistically analyzes the three-dimensional spatial enrichment regions of a specified molecular fragment around a target molecule.

[0033] Aromatic spatial enrichment coefficient Kπ: the ratio of the probability density of the aromatic ring centroid of the aromatic hydrogen bond donor in the local region above and below the target aromatic plane to the average probability density of the system.

[0034] Glycosidic interaction index (Ggly): an indicator of the strength of short-range interactions between eutectic solvent components and sites near polysaccharide glycosidic bonds.

[0035] Lignin bond type affinity index Lβ: an indicator of the degree of interaction affinity between eutectic solvents and lignin linkage fragments of different bond types.

[0036] This embodiment is implemented using the following computing devices and software environment. The software names are only used to describe the repeatable implementation method; software with the same computing function can be used as an equivalent substitute.

[0037] 1) The computing device is a workstation with no less than 16-core central processing unit, 64 GB of memory and 2 TB of storage space, or a high-performance computing node with equivalent performance.

[0038] 2) The operating system is Linux.

[0039] 3) The molecular modeling program uses Gaussian 09 or software with equivalent molecular construction capabilities.

[0040] 4) The conformation search program uses Molclus or software with equivalent low-energy conformation search capabilities.

[0041] 5) The semi-empirical quantum chemistry screening program uses MOPAC or software with PM6-DH+ calculation capabilities.

[0042] 6) Wavefunction analysis programs should use Multiwfn or other software capable of calculating electrostatic potential, molecular polarity index, polar surface area, weak interactions, and molecular topological parameters in atoms.

[0043] 7) The molecular dynamics program used is GROMACS.

[0044] 8) Packmol was used as the initial box construction program for molecular dynamics.

[0045] 9) The force field parameters of small molecules were obtained from the CGenFF platform and input into the molecular dynamics program using the CHARMM force field format.

[0046] Example 1 like Figure 1 As shown, this embodiment provides a multi-scale molecular simulation screening method for aromatic eutectic solvents based on a multi-objective collaborative scoring model, including the following steps: Step 1: Establish a target molecule model library and a eutectic solvent candidate component library; The target molecule model library includes: (1) Model of plant aromatic active ingredients Rutin was selected as a representative model of plant aromatic active ingredients. The rutin molecule contains a flavonoid core, multiple phenolic hydroxyl groups, glycoside substituents, and a conjugated aromatic structure, which can represent the common flavonoids and polyphenols aromatic active ingredients found in rose petals, leaves, and other plant materials.

[0047] In modeling, a three-dimensional structure of the rutin molecule was established, retaining the phenolic hydroxyl oxygen atom, phenolic hydroxyl hydrogen atom, flavonoid parent aromatic ring, glycosidic bond oxygen atom, glycosyl side chain hydroxyl group, and carbonyl oxygen atom. The rutin aromatic ring plane was used for subsequent calculation of the aromatic spatial enrichment coefficient Kπ.

[0048] (2) Polysaccharide fragment model Dextran and xylan fragments were constructed to represent cellulose and hemicellulose structural units, respectively. Both fragments retained the glycosidic oxygen atom, epoxy atom, side-chain hydroxyl group, and terminal hydroxyl group. The glycosidic oxygen atom was defined as the major analytical site for the glycosidic interaction index Ggly.

[0049] (3) Lignin bond-type fragment model A lignin model consisting of six basic structural units was constructed, including β-O-4, β-1, β-5, and 4-O-5 bonds. For each lignin model, aromatic rings, ether oxygen, benzylic carbon, and phenolic hydroxyl groups near the connecting bonds were extracted as analytical sites to calculate the lignin bond affinity indices Lβ-O-4, Lβ-1, Lβ-5, and L4-O-5.

[0050] The library of candidate components for eutectic solvents includes: This embodiment establishes six candidate eutectic solvent systems, as shown in Table 1 below. The hydrogen bond acceptor is choline chloride, the aliphatic hydrogen bond donor is ethylene glycol or glycerol, and the aromatic hydrogen bond donor is guaiacol or p-toluenesulfonic acid.

[0051] Table 1 Candidate Eutectic Solvent Systems

[0052] For the binary systems CE and CG, one choline chloride molecule and two aliphatic hydrogen bond donor molecules are used to construct the minimum molecular cluster. For the ternary systems CEG, CEP, CGG, and CGP, one choline chloride molecule, two aliphatic hydrogen bond donor molecules, and one aromatic hydrogen bond donor molecule are used to construct the minimum molecular cluster.

[0053] Step 2, Quantum Chemical Calculations Quantum chemical calculations include the following steps: (1) Target molecule optimization The plant aromatic active ingredient model (rutin), polysaccharide fragment model (glucan fragment, xylan fragment), and lignin bond-type fragment model (lignin hexa-unit model) were input into the quantum chemical calculation module for geometric optimization and single-point energy calculation. The geometric optimization method was B3LYP-D3(BJ) / def2-SVP, and the single-point energy calculation method was B3LYP-D3(BJ) / def2-TZVP. After the calculation was completed, the optimized structure file, energy file, and wavefunction file of the target molecule were retained, and it was checked whether each target molecule had imaginary frequencies; if imaginary frequencies were found, the initial configuration was readjusted and optimized again.

[0054] (2) Construction of eutectic solvent molecular clusters: Initial eutectic solvent clusters were constructed based on the composition and molar ratio of candidate eutectic solvents (e.g., six eutectic solvents: CE, CEG, CEP, CG, CGG, and CGP). For binary eutectic solvent clusters, one hydrogen bond acceptor molecule and two aliphatic hydrogen bond donor molecules were used; for ternary aromatic eutectic solvent clusters, one hydrogen bond acceptor molecule, two aliphatic hydrogen bond donor molecules, and one aromatic hydrogen bond donor molecule were used.

[0055] (3) Construction of eutectic solvent-target molecule complex clusters: After obtaining the optimized structure of the target molecule and the initial molecular cluster of the eutectic solvent, the optimized plant aromatic active ingredient model, polysaccharide fragment model or lignin bond type fragment model are combined with the eutectic solvent molecular cluster to construct the eutectic solvent-target molecule complex cluster.

[0056] For aromatic active ingredient complexes, the aromatic ring of the aromatic hydrogen bond donor is positioned close to the aromatic plane of the target aromatic active ingredient. For example, for a rutin-eutectic solvent complex, the eutectic solvent molecule cluster is placed near the aromatic plane of the rutin flavonoid core, with the initial distance between the centroid of the aromatic ring of the aromatic hydrogen bond donor and the centroid of the rutin aromatic ring being 0.35-0.50 nm, while simultaneously placing the chloride ion or hydroxyl hydrogen close to the hydroxyl oxygen, glycosidic bond oxygen, or carbonyl oxygen site of rutin phenol.

[0057] For polysaccharide complexes, chloride ions, hydroxyl hydrogens, or sulfonic acid hydrogens are brought close to the oxygen atom of the polysaccharide glycosidic bond. For example, for dextran or xylan fragment-eutectic solvent complexes, chloride ions, hydroxyl hydrogens, or sulfonic acid hydrogens in the eutectic solvent are placed near the oxygen atom, epoxy atom, or side-chain hydroxyl group of the polysaccharide glycosidic bond at an initial distance of 0.25-0.35 nm to simulate the short-range effect of the eutectic solvent on sites near the polysaccharide glycosidic bond.

[0058] For lignin complex clusters, the aromatic ring of the aromatic hydrogen bond donor is positioned close to the lignin aromatic unit, and the hydrogen bond donor group is positioned close to the ether bond oxygen or phenolic hydroxyl site. For example, for lignin bond-fragment-eutectic solvent complex clusters, the aromatic ring of the aromatic hydrogen bond donor is positioned close to the lignin aromatic unit, with a centroid distance of 0.35-0.50 nm between the aromatic rings, and the hydrogen bond donor group is positioned close to the ether bond oxygen, benzylic carbon, or phenolic hydroxyl site near the β-O-4, β-1, β-5, or 4-O-5 linkage structure.

[0059] The aforementioned composite clusters are used for subsequent conformational searches, semi-empirical screening, density functional theory optimization, BSSE corrected interaction energy calculation, IGMH weak interaction analysis, and AIM topology analysis.

[0060] (4) Conformation search and semi-empirical screening The eutectic solvent molecular clusters and eutectic solvent-target molecular complex clusters are input into the conformation search program, with an initial number of no less than 100 conformations. The conformation screening energy window is within 10 kcal / mol of the lowest relative energy. The candidate conformations are sorted from low to high energy, and the lowest energy conformation is selected as the initial structure for semi-empirical quantum chemical screening. The PM6-DH+ method is used for initial screening optimization.

[0061] (5) Density functional theory calculations and basis set overlap error (BSSE) correction Density functional theory calculations were performed on the molecular clusters and complex clusters after semi-empirical initial screening. For eutectic solvent molecules, geometry optimization was performed using M06-2X / 6-31G(d), and single-point energies were calculated using M06-2X / 6-311+G(d,p). For the eutectic solvent-target molecule complex cluster, basis set overlap error correction was performed using the keyword Counterpoise=2 to obtain the corrected interaction energies, and a file for wavefunction analysis was generated.

[0062] (6) Quantum chemical descriptor extraction electrostatic potential The electrostatic potential of the molecular surface was calculated using a wavefunction analysis program. The maximum positive electrostatic potential ESPmax, the minimum negative electrostatic potential ESPmin, the electrostatic potential range ΔESP=ESPmax-ESPmin, and the local electrostatic potentials near chloride ions, hydroxyl hydrogens, sulfonic acid oxygens, phenolic hydroxyl oxygens, and glycosidic oxygens were extracted.

[0063] Molecular polarity Calculate the molecular polarity index (MPI), polar surface area (PSA), and polar surface area ratio (FPSA). A higher MPI indicates a more uneven distribution of electrostatic potential on the molecular surface, and the stronger the electrostatic interaction between the eutectic solvent and polyhydroxy aromatic molecules and polysaccharide structures may be.

[0064] IGMH weak interactions Weak interaction analysis was performed using an independent gradient model based on Hirshfeld partitioning, with the isosurface parameter set to 0.008 au. Hydrogen bonding regions, van der Waals interactions, π-π stacking interactions, CH-π interactions, and repulsion regions were observed. When a continuous weak interaction region appeared between the aromatic ring of the aromatic hydrogen bond donor and the aromatic ring of rutin or lignin, and this region primarily exhibited van der Waals attraction, it was recorded as the presence of aromatic synergistic interaction.

[0065] Atom-in-Molecular Topological Analysis (AIM Topological Analysis) Atomic molecular topological analysis was used to locate bond critical points, and the bond critical point electron density ρ(rBCP), the Laplace value of the bond critical point electron density ∇²ρ(rBCP), the energy density E(rBCP), and the hydrogen bond binding energy were extracted. The hydrogen bond binding energy was calculated using the formula EH-bond = -223.08 × ρ(rBCP) + 0.7423, where ρ(rBCP) is the bond critical point electron density, and the unit of hydrogen bond binding energy is kcal / mol.

[0066] In this embodiment, representative hydrogen bond energies selected for the Rutin-CEG complex include -8.41513, -4.71647, -7.14135, -6.54126, -4.33723, and -5.34778 kcal / mol; representative hydrogen bond energies for the Xyl-CEP complex include -4.30374, -4.04204, -7.04491, -6.02470, -6.61358, and -11.56302 kcal / mol. These results indicate that CEG exhibits moderate hydrogen bond strength with rutin, while CEP demonstrates stronger directional hydrogen bond strength with xylan.

[0067] (5) Decomposition of interaction energy The interaction energies of the eutectic solvent-target molecule complexes, corrected for basis set overlap errors, were calculated and decomposed into electrostatic interaction energy Eele, van der Waals interaction energy EvdW, repulsion energy Erep, and total interaction energy Etotal. If the absolute value of EvdW increases upon the addition of an aromatic hydrogen bond donor, and weak interaction analysis shows a continuous attraction region near the aromatic ring, then the aromatic eutectic solvent is determined to have an enhanced van der Waals synergistic effect on the aromatic active ingredient or lignin aromatic structure.

[0068] Step 3: Molecular Dynamics Simulation (1) Generation of force field parameters Rutin, dextran fragments, xylan fragments, lignin models, choline chloride, ethylene glycol, glycerol, guaiacol, and p-toluenesulfonic acid were input into the CGenFF platform to generate CHARMM force field parameter files. The small molecule force field parameters output by the CGenFF platform include parameter penalty scores. These scores characterize the reliability of automatically matched bond lengths, bond angles, dihedral angles, or non-bonded interaction parameters; higher penalty scores indicate a lower degree of match between the parameter and the existing force field parameter library. For bond lengths, bond angles, or dihedral angle parameters with high penalty scores, quantum chemical optimization configurations were used to verify the corresponding parameters, thereby improving the reliability of the molecular dynamics simulation topology files.

[0069] (2) Construction of periodic simulation boxes A cubic periodic simulation box with a side length of 6.0 nm was constructed using Packmol. Each simulation box contained one rutin molecule, one dextran fragment, one xylan fragment, one lignin hexaunit model, and 300 choline chloride molecules. For the CE or CG system, 600 aliphatic hydrogen bond donor molecules were added; for the CEG, CEP, CGG, or CGP system, 600 aliphatic hydrogen bond donor molecules and 300 aromatic hydrogen bond donor molecules were added. These quantities ensured that the molar ratio of hydrogen bond acceptor to aliphatic hydrogen bond donor was 1:2 in binary systems and 1:2:1 in ternary systems.

[0070] (3) Energy minimization and balance simulation The periodic simulation chamber was subjected to energy minimization, canonical ensemble equilibration (NVT equilibration), isothermal-isobaric ensemble equilibration (NPT equilibration), and production simulation in sequence. The steps and parameters are shown in Table 2 below: Energy minimization employs the steepest descent method with a maximum step size of 50,000 steps. The process stops when the maximum force in the system is less than 1,000 kJ·mol⁻¹·nm⁻¹, in order to eliminate unreasonable atomic contacts in the initial structure.

[0071] The NVT equilibrium was carried out under a canonical ensemble, with the temperature set at 298 K, the equilibrium time at 1 ns, and the time step at 2 fs, to stabilize the system temperature.

[0072] NPT equilibrium was carried out under isothermal and isobaric ensemble conditions, with the temperature set at 298 K, the pressure at 1 bar, the equilibrium time at 5 ns, and the time step at 2 fs, to stabilize the system density and volume.

[0073] Table 2 Steps and Parameters

[0074] Production simulations were conducted under the NPT ensemble with a simulation time of 100 ns and a time step of 2 fs. Electrostatic interactions were performed using the particle mesh Ewald method, with a van der Waals interaction cutoff radius of 1.2 nm. Hydrogen bond length constraints were handled using the LINCS algorithm, and trajectories were saved every 10 ps.

[0075] After the production simulation is completed, the trajectory of 20-100 ns is used for statistical analysis; if the system density, temperature and total energy do not reach stability after 20 ns, the NPT equilibrium time is extended to 10 ns and the production simulation is repeated.

[0076] (4) Molecular dynamics descriptor extraction Average number of hydrogen bonds (NHB) The average number of hydrogen bonds between the eutectic solvent component and rutin, dextran, xylan, and lignin was statistically analyzed. The criteria for hydrogen bond determination were a donor-acceptor distance less than or equal to 0.35 nm and a hydrogen bond angle deviating from the linear direction by no more than 30°. The average number of hydrogen bonds between rutin and the eutectic solvent (NHBrutin), dextran and lignin (NHBglucan), xylan and lignin (NHBlignin) was calculated separately.

[0077] radial distribution function This includes the position rp of the first peak of the radial distribution function and the intensity g(rp) of the first peak of the radial distribution function.

[0078] Calculate the radial distribution functions of chloride ions, hydroxyl oxygen, sulfonic acid oxygen, and aromatic ring centroids relative to the target site. The bin width of the radial distribution function is set to 0.002 nm, and the maximum statistical distance is set to 1.2 nm. The radial distributions of chloride ions relative to the hydroxyl oxygen of rutin, chloride ions relative to the hydroxyl oxygen of rutin heterocycles, sulfonic acid oxygen relative to the oxygen of xylan glycosidic bonds, sulfonic acid hydrogen relative to the oxygen of xylan glycosidic bonds, the radial distribution of the aromatic ring centroid of the aromatic hydrogen bond donor relative to the aromatic ring centroid of rutin, and the radial distribution of the aromatic ring centroid of the aromatic hydrogen bond donor relative to the aromatic ring centroid of lignin.

[0079] If the first peak of the radial distribution function is located between 0.25 and 0.35 nm, and the intensity of the first peak is greater than or equal to 1.5, then the candidate eutectic solvent is considered to have an effective short-range effect on the target site. For the CEG-rutin system, the first peak of the radial distribution function of chloride ions near the rutin O4 site is located at 0.314 nm, with a peak intensity of 9.573, indicating that chloride ions in CEG have a significant short-range hydrogen bonding effect on the aromatic hydroxyl sites of rutin.

[0080] Aromatic spatial enrichment coefficient Kπ Spatial distribution functions (SDFs) were used to analyze the three-dimensional enrichment regions of each eutectic solvent component around the target molecule. SDFs characterize the spatial distribution of components such as aromatic hydrogen bond donors, chloride ions, hydroxyl hydrogens, or sulfonic acid oxygens around the target molecule.

[0081] Based on the spatial distribution function analysis, the aromatic spatial enrichment coefficient Kπ is further calculated. For the rutin model, the plane of the flavonoid parent aromatic ring is used as the reference plane; for the lignin model, the plane of the lignin aromatic unit is used as the reference plane. The spatial distribution probability density ρπ of the aromatic hydrogen bond donor aromatic ring centroid is statistically analyzed in the region 0.30-0.50 nm above and below the reference plane, and the average probability density ρ0 of the aromatic ring centroid in the entire simulation box is also statistically analyzed. The aromatic spatial enrichment coefficient Kπ is calculated according to the formula Kπ=ρπ / ρ0.

[0082] When Kπ≥1.20, it is determined that the aromatic hydrogen bond donor forms a spatially enriched configuration near the target aromatic structure; when Kπ≥1.50, it is determined that it has significant π-π stacking or CH-π synergistic effect.

[0083] Glycosidic bond interaction index (Ggly) The glycosidic interaction index is used to evaluate the ability of candidate eutectic solvents to interact with sites near glycosidic bonds in polysaccharides. It is calculated using the formula Ggly = 0.40N⁺[ggly(rp)] + 0.30N⁺(NHBgly) + 0.30N⁺(-Eele,gly). Where ggly(rp) is the intensity of the first peak of the radial distribution function of the key atom in the eutectic solvent relative to the oxygen in the glycosidic bond, NHBgly is the average number of hydrogen bonds near the glycosidic bond, Eele,gly is the electrostatic interaction energy between the eutectic solvent and the fragment near the glycosidic bond, and N... + This is a favorable normalization function.

[0084] When the position of the first peak rp is not in the range of 0.25-0.35 nm, the ggly(rp) term of the candidate system is recorded as 0. When rp is in the range of 0.25-0.35 nm and g(rp) ≥ 1.5, the candidate eutectic solvent is determined to have an effective short-range effect on sites near the glycosidic bond.

[0085] Lignin bond affinity index Lβ Affinity indices were calculated for four lignin bond types: β-O-4, β-1, β-5, and 4-O-5, using the formula Lβ = 0.40N⁺(-Etotal,β) + 0.30N⁺(Kπ,lignin) + 0.30N⁺(NHBβ). Here, Etotal,β is the total interaction energy between the eutectic solvent and the corresponding lignin bond type fragment, which is β-O-4, β-1, β-5, or 4-O-5; Kπ,lignin is the aromatic space enrichment coefficient of the aromatic hydrogen bond donor near the lignin aromatic unit; NHBβ is the average number of hydrogen bonds formed between the eutectic solvent and the vicinity of this lignin bond type; and N... + This is a favorable normalization function.

[0086] Lβ-O-4, Lβ-1, Lβ-5 and L4-O-5 were obtained to distinguish the selective effects of candidate eutectic solvents on different lignin bond types.

[0087] Step 4: Multi-scale feature extraction and fusion After the extraction of quantum chemical and molecular dynamics descriptors, the multi-scale feature extraction and fusion module first normalizes, groups, and fuses the descriptors to form multi-scale feature vectors.

[0088] (1) Normalization In this invention, N⁺ and N⁻ are both normalization functions. N⁺ represents a favorable normalization function, used to handle descriptors where larger values ​​are more favorable to the target score; N⁻ represents a penalized normalization function, used to handle descriptors where larger values ​​are less favorable to the target score.

[0089] Based on the direction of each descriptor's contribution to the target score, the descriptors are divided into beneficial descriptors and penalized descriptors.

[0090] Favorable descriptors are those whose values, when increased, are beneficial to improving the compatibility of active ingredient extraction, lignocellulose grading, or process compatibility. These include, but are not limited to, the aromatic spatial enrichment coefficient Kπ, the average number of hydrogen bonds NHB, the intensity of the first peak of the radial distribution function g(rp), the glycosidic bond interaction index Ggly, the lignin bond type affinity index Lβ, the liquid stability Sliq, and the polarity compatibility Pfit.

[0091] Penalty descriptors are those whose values, when increased, may reduce the stability of active ingredients, increase the risk of cellulose damage, increase the risk of lignin condensation, or reduce process operability. These include, but are not limited to, viscosity η, acid risk Racid, redox potential ORP, cellulose damage risk Cdam, and lignin condensation risk Rcond.

[0092] For the favorable descriptor x, normalization is performed using the formula N⁺(x) = (x - xmin) / (xmax - xmin). For the penalized descriptor y, normalization is performed using the formula N⁻(y) = (ymax - y) / (ymax - ymin). Here, xmin, xmax, ymin, and ymax are the minimum and maximum values ​​of the corresponding descriptor in the candidate eutectic solvent system set, respectively. In this embodiment, the candidate eutectic solvent system set includes six candidate systems: CE, CEG, CEP, CG, CGG, and CGP. When the denominator is 0, the corresponding normalization value is set to 0.5.

[0093] (2) Group fusion The normalized descriptors are divided into three feature groups according to their functions: The first category is the active ingredient extraction characteristic group, including -EvdW,rutin-arom, Kπ,rutin, NHBrutin and gO4 / O5(rp); The second category is the lignocellulose grading characteristic group, including -Exylan, -Elignin, Ggly, and Lβ-O-4; The third category is the process compatibility and risk characteristics group, which includes viscosity η, liquid stability Sliq, polarity compatibility Pfit, redox potential ORP, acid degradation risk Racid, and cellulose damage risk Cdam.

[0094] The multi-scale feature vector includes an active ingredient extraction feature group, a lignocellulose grading feature group, a process adaptation feature group, and a risk feature group. Specifically, the active ingredient extraction feature group includes -EvdW,rutin-arom, Kπ,rutin, NHBrutin, and gO4 / O5(rp); the lignocellulose grading feature group includes -Exylan, -Elignin, Ggly, and Lβ-O-4; the process adaptation feature group includes viscosity η, liquid stability Sliq, polarity fit Pfit, and redox potential ORP; and the risk feature group includes acid risk Racid, cellulose damage risk Cdam, and lignin condensation risk Rcond.

[0095] Step 5: Multi-objective collaborative scoring model The multi-objective collaborative scoring model reads multi-scale feature vectors and calculates the active ingredient extraction suitability index A, lignocellulose grading suitability index F, process suitability index T, and risk penalty term R, respectively. Then, it calculates the active ingredient extraction score SE and the lignocellulose grading score SF (see...). Figure 3 ).

[0096] (1) Active ingredient extraction suitability index A A=0.30N⁺(-EvdW,rutin-arom)+0.25N⁺(Kπ,rutin)+0.20N⁺(NHBrutin)+0.15N⁺[gO4 / O5(rp)]+0.10N⁻(Racid) Wherein, EvdW,rutin-arom is the van der Waals contribution energy between the aromatic hydrogen bond donor and rutin, Kπ,rutin is the aromatic spatial enrichment coefficient near the aromatic plane of rutin, NHBrutin is the average number of hydrogen bonds between rutin and the eutectic solvent, gO4 / O5(rp) is the radial distribution function peak intensity of chloride ions or hydrogen bond donor atoms near the O4 / O5 sites of rutin, and Racid is the acid degradation risk penalty term.

[0097] (2) Lignocellulose grading fit index F F=0.30N⁺(-Exylan)+0.20N⁺(-Elignin)+0.20N⁺(Ggly)+0.20N⁺(Lβ-O-4)+0.10N⁻(Cdam) Among them, Exylan is the interaction energy between the eutectic solvent and xylan fragments, Elignin is the interaction energy between the eutectic solvent and lignin model, Ggly is the glycosidic bond interaction index, Lβ-O-4 is the β-O-4 bond type affinity index, and Cdam is the cellulose damage risk penalty term.

[0098] (3) Process adaptability index T T=0.40N⁻(η)+0.25N⁺(Sliq)+0.20N⁺(Pfit)+0.15N⁻(ORP) Where η is the viscosity at 25℃, Sliq is the liquid stability (1 for homogeneous and transparent liquids, 0 for turbidity or crystallization), Pfit is the polarity compatibility, and ORP is the redox potential. The process compatibility index T is used to evaluate the feasibility of candidate eutectic solvents in preparation, mass transfer, and processing; a higher T value indicates that the candidate eutectic solvent has better liquid stability, polarity compatibility, and process operability.

[0099] (4) Risk penalty item R The risk penalty term R is used to characterize the potential adverse effects of candidate eutectic solvents during active ingredient extraction or lignocellulose fractionation. The risk penalty term R includes the active ingredient degradation risk (Ract) and the lignocellulose fractionation risk (Rfrac).

[0100] The active ingredient degradation risk (Ract) is used to characterize the adverse effects of candidate eutectic solvents on the stability of aromatic active ingredients, and is calculated using the following formula: Ract=0.50N⁺(Racid)+0.30N⁺(ORP)+0.20N⁺(η) Where Racid is the acidity risk parameter, ORP is the redox potential, and η is the viscosity at 25°C. The higher the Ract value, the higher the risk of degradation of the aromatic active ingredient by the candidate eutectic solvent.

[0101] The risk Rfrac for lignocellulose fractionation is used to characterize the adverse effects of candidate eutectic solvents on excessive cellulose damage, lignin condensation, or limited mass transfer during lignocellulose fractionation, and is calculated using the following formula: Rfrac=0.50N⁺(Cdam)+0.30N⁺(Rcond)+0.20N⁺(η) Wherein, Cdam is the cellulose damage risk parameter, Rcond is the lignin condensation risk parameter, and η is the viscosity at 25°C. The larger the Rfrac value, the higher the risk of the candidate eutectic solvent producing side reactions or adverse effects during the lignocellulose fractionation process.

[0102] Ract and Rfrac are risk values. The higher the values ​​of risk factors such as Racid, ORP, η, Cdam, and Rcond, the greater the risk. They are normalized using N⁺. However, Ract and Rfrac are deducted when they are finally included in SE / SF.

[0103] (5) Extracting scores and grading scores The active ingredient extraction score (SE) was calculated as follows: SE = 0.60A + 0.25T - 0.15Ract. The lignocellulose grading score (SF) was calculated as follows: SF = 0.60F + 0.20T - 0.20Rfrac. Here, Ract represents the degradation risk of the active ingredient, and Rfrac represents the lignocellulose grading risk.

[0104] Step 6: Pattern Determination and Output The judgment rules are as follows: ① When SE≥0.70 and SE-SF≥0.10, output the active ingredient extraction priority mode.

[0105] ② When SF≥0.70 and SF-SE≥0.10, output the lignocellulose grading priority mode.

[0106] ③ When SE≥0.60, SF≥0.60 and |SE-SF|<0.10, output is processed in a balanced manner.

[0107] ④ Sort each eutectic solvent candidate system from high to low SE to obtain the first active ingredient extraction candidate system, and sort each eutectic solvent candidate system from high to low SF to obtain the first lignocellulose grading candidate system. When the first active ingredient extraction candidate system and the first lignocellulose grading candidate system are not the same eutectic solvent candidate system, and the first active ingredient extraction candidate system satisfies SE≥0.70 and the first lignocellulose grading candidate system satisfies SF≥0.70, the segmented collaborative processing mode is output.

[0108] In this embodiment, the six candidate eutectic solvents were processed according to the above calculation process, and the judgment results are shown in Table 3 below.

[0109] Table 3. Pattern Determination Results

[0110] Therefore, the first candidate eutectic solvent is CEG (choline chloride: ethylene glycol: guaiacol = 1:2:1), used in the priority mode for extracting plant aromatic active ingredients; the second candidate eutectic solvent is CEP (choline chloride: ethylene glycol: p-toluenesulfonic acid = 1:2:1), used in the priority mode for lignocellulose grading. When both high flavonoid / polyphenol extraction yields and high lignocellulose grading effects are required for the same plant material, a segmented synergistic processing mode is output, i.e., CEG is first used for active ingredient extraction, and then CEP is used for lignocellulose grading of the extracted solid residue.

[0111] Experimental Example 1 1. Raw material pretreatment Rose petals and leaves from Pingyin rose were mixed at a mass ratio of 1:1. The sample was freeze-dried at -50℃ for 24 h, pulverized, and passed through a 60-mesh sieve to obtain rose powder. The powder was placed in a sealed bag and stored in a cool, dry place away from light.

[0112] 2. Eutectic solvent treatment Six eutectic solvents, CE, CEG, CEP, CG, CGG, and CGP, as shown in Table 1, were used. Specifically, CE was choline chloride:ethylene glycol = 1:2, CEG was choline chloride:ethylene glycol:guaiacol = 1:2:1, CEP was choline chloride:ethylene glycol:p-toluenesulfonic acid = 1:2:1, CG was choline chloride:glycerol = 1:2, CGG was choline chloride:glycerol:guaiacol = 1:2:1, and CGP was choline chloride:glycerol:p-toluenesulfonic acid = 1:2:1.

[0113] The rose powder mixture and the aforementioned eutectic solvent were mixed at a solid-liquid ratio of 1:10 g / mL and then transferred to a microwave reactor for treatment. The treatment temperature was 70℃, the stirring speed was 500 rpm, and the treatment time was 20 min.

[0114] After processing, the slurry was transferred to centrifuge tubes for centrifugation; the supernatant was collected, filtered through a 0.45 μm nylon 66 organic filter membrane, and used for the determination of total flavonoids and total polyphenols. The solid residue was repeatedly washed with deionized water until neutral, and then freeze-dried at -50℃ for storage.

[0115] 3. Determination of total flavonoid content Dilute the supernatant by a factor of 2, and add 5.0 mL of the diluted solution to a test tube. Add 0.30 mL of 5% sodium nitrite solution sequentially; after a 6-minute interval, add 0.30 mL of 10% aluminum chloride solution; after another 6-minute interval, add 2.0 mL of 1 mol / L sodium hydroxide solution; then add 2.4 mL of deionized water. Mix well and incubate in the dark for 15 minutes, then measure the absorbance at 510 nm. Establish a standard curve using rutin as a standard. The total flavonoid content is expressed as mg RE / g DW.

[0116] 4. Determination of total polyphenol content Take 0.2 mL of the diluted extract, add 15.8 mL of deionized water and 1.0 mL of Folin-Ciocalteu reagent, mix well and let stand for 5 min. Then add 3.0 mL of 20% sodium carbonate solution, react in the dark for 2 h, and measure the absorbance at 765 nm. The total polyphenol content is expressed as mg GAE / g DW.

[0117] 5. Solid component analysis The cellulose, hemicellulose, and lignin contents in the solid residue were determined according to the standard component analysis method for herbal biomass, and calculated using the following formulas: Cellulose retention rate = (glucose content in solid residue / glucose content in raw material) × 100%; Hemicellulose removal rate = [1 - (arabinose content in solid residue / arabinose content in raw material)] × 100%; Lignin removal rate = [1 - (lignin content in solid residue / lignin content in raw material)] × 100%.

[0118] 6. Verification Results The results of the above determination of total polyphenol content, total flavonoid content, and cellulose, hemicellulose and lignin content in solid residue are shown in Table 4 below.

[0119] Table 4 Validation Results

[0120] The results show that CEG had the highest total polyphenol and total flavonoid content, at 215.91 mg GAE / gDW and 105.48 mg RE / gDW, respectively; CEP had the highest hemicellulose and lignin removal rates, at 97.60% and 77.68%, respectively. These experimental results are consistent with the determination that "CEG is the preferred system for active ingredient extraction, and CEP is the preferred system for lignocellulose grading" based on the screening method of this invention.

[0121] Comparative Example 1: Screening based solely on total interaction energy or average number of hydrogen bonds In contrast, a screening model considering only the total interaction energy or the average number of hydrogen bonds was used to rank candidate eutectic solvents. This model tends to prioritize PTSA-like systems with stronger acidity and hydrogen bonding, due to their strong short-range interactions with polysaccharide and lignin fragments. However, this model does not incorporate Kπ, active ingredient degradation risk, and acid penalty terms, and cannot distinguish between "strong fractionation ability" and "adaptability for active ingredient extraction," potentially misselecting ChCl:EG:PTSA as the preferred solvent for active ingredient extraction.

[0122] Compared to the comparative example, Example 1 of the present invention introduces Kπ and EvdW,act-arom in A, and introduces acid degradation risk and redox risk in Ract. Therefore, it can identify ChCl:EG:Gua as the preferred system for active ingredient extraction without misjudging due to the strong interaction of the PTSA system.

[0123] Comparative Example 2: Screening was conducted based solely on hemicellulose removal rate or lignin removal rate. In contrast, a screening model considering only hemicellulose or lignin removal rates was used. This model prioritizes PTSA-based acidic systems and tends to use them as solvents for one-pot integrated treatment. However, because this model does not constrain active ingredient extraction efficiency, aromatic spatial enrichment, active ingredient stability, and cellulose damage risk, it may lead to insufficient active ingredient extraction or increased cellulose damage.

[0124] Compared to the comparative example, Embodiment 1 of the present invention models the grading capability F and the extraction capability A separately, and outputs a segmented collaborative processing mode through the difference between SE and SF. When the extraction priority candidate and the grading priority candidate are inconsistent, the system does not force the output of a single solvent, but outputs the first eutectic solvent and the second eutectic solvent, which is more in line with the technical requirements for the high-value utilization of complex plant biomass.

[0125] Comparative Example 3: Scoring model without introducing the aromatic spatial enrichment factor Kπ In contrast, a scoring model that does not incorporate Kπ was used, considering only the number of hydrogen bonds, electrostatic interactions, and total van der Waals interactions. This model can determine whether non-covalent interactions exist between the solvent and the target molecule, but it is difficult to identify the local enrichment configurations formed above and below the target aromatic plane by aromatic hydrogen bond donors. Therefore, it cannot fully explain the improvement in the extraction performance of aromatic active ingredients by the guaiacol system.

[0126] Compared with the comparative example, Embodiment 1 of the present invention extracts Kπ through SDF and uses it as an important input variable of A, making π-π stacking and CH-π interaction calculable and sortable technical features, thus enhancing the correspondence between molecular simulation and experimental extraction results.

[0127] To further illustrate the effectiveness of the multi-objective collaborative scoring model of this invention, the screening results of the model of this invention are compared and analyzed with those of the three types of scoring models in Comparative Examples 1-3. The results are as follows: Figure 4 As shown in Table 5.

[0128] Table 5 Comparison of Screening Results

[0129] Table 5 shows that Comparative Examples 1 and 2 readily select PTSA-based acidic systems, but this result cannot simultaneously meet the requirements of stable extraction of aromatic active ingredients and lignocellulose fractionation. While Comparative Example 3 can consider some non-covalent interactions, it cannot convert the spatial enrichment configuration of aromatic hydrogen bond donors near the target aromatic plane into a sortable descriptor due to the lack of Kπ. This invention, by introducing Kπ, Ggly, and Lβ, and further constructing a dual-scoring model of SE and SF, can identify CEG as the preferred system for active ingredient extraction and CEP as the preferred system for lignocellulose fractionation, and output a segmented synergistic processing mode when the two are inconsistent.

[0130] Example 2 This embodiment provides a multi-scale molecular simulation screening system for aromatic eutectic solvents based on a multi-objective collaborative scoring model. This system can be deployed on a local server, workstation, high-performance computing cluster, or cloud computing platform.

[0131] like Figure 2 As shown, the system includes a model input module, a quantum chemical calculation module, a molecular dynamics simulation module, a multi-scale feature extraction and fusion module, a multi-objective collaborative scoring module, and a pattern discrimination and output module. The multi-scale feature extraction and fusion module is a pre-module of the multi-objective collaborative scoring module. After the quantum chemical calculation module and the molecular dynamics simulation module output the original descriptors, they first enter the multi-scale feature extraction and fusion module for normalization, grouping, and feature vector construction. Subsequently, the multi-objective collaborative scoring module reads the multi-scale feature vector and calculates A, F, T, R, SE, and SF according to weights and thresholds. Specifically: The model input module is used to import or generate target molecule structures, eutectic solvent component structures and candidate molar ratios, and to establish a target molecule model library and a eutectic solvent candidate component library. The quantum chemistry calculation module is used to call quantum chemistry programs to complete conformation search, geometry optimization, single-point energy calculation and wavefunction analysis, and extract quantum chemistry descriptors. The molecular dynamics simulation module is used to call molecular dynamics programs to complete system construction, energy minimization, equilibrium simulation, production simulation and trajectory analysis, and obtain molecular dynamics descriptors. The multi-scale feature extraction and fusion module is used to receive the quantum chemical descriptor output by the quantum chemical calculation module and the molecular dynamics descriptor output by the molecular dynamics simulation module, and normalize and fuse the quantum chemical descriptor and the dynamics descriptor into a multi-scale feature vector. The multi-objective collaborative scoring module is used to calculate A, F, T, and R based on weights and thresholds, and finally calculate the active ingredient extraction score SE and the lignocellulose grading score SF. The mode discrimination and output module is used to determine the preferred mode and output the eutectic solvent formulation, molar ratio, application mode, and experimental verification plan.

[0132] Finally, the system outputs the results.

[0133] This invention transforms the design of aromatic eutectic solvents from empirical screening into a computational chemistry workflow that is computationally achievable, orderable, and verifiable, through multi-scale molecular simulations and a multi-objective synergistic scoring model. Its core lies not in a single solvent formulation, but in constructing a computational design method using Kπ, Ggly, Lβ, A, F, T, R, SE, and SF to identify different target conflicts and output application patterns. This approach is beneficial for covering subsequent application scenarios involving replacing aromatic hydrogen bond donors, plant materials, or target active ingredients.

[0134] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, alterations, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-scale molecular simulation screening method for aromatic eutectic solvents based on a multi-objective collaborative scoring model, characterized in that, Includes the following steps: (1) Establish a target molecule model library and a eutectic solvent candidate component library; (2) Perform quantum chemical calculations on the target molecule and the candidate system of eutectic solvent, and extract quantum chemical descriptors, including electrostatic potential, molecular polarity, IGMH weak interaction, AIM bond critical point, and BSSE corrected interaction energy; (3) Construct a molecular dynamics simulation system containing the target molecule and the eutectic solvent component, perform equilibrium simulation and production simulation on the molecular dynamics simulation system, and obtain dynamic descriptors, including the average number of hydrogen bonds, radial distribution function, aromatic spatial enrichment coefficient Kπ, glycosidic bond interaction index Ggly and lignin bond type affinity index Lβ; (4) Normalize the descriptors extracted in steps (2) and (3) to form multi-scale feature vectors; (5) The multi-objective collaborative scoring model reads multi-scale feature vectors and calculates the active ingredient extraction adaptation index A, lignocellulose grading adaptation index F, process adaptation index T, and risk penalty term R respectively, and further calculates the active ingredient extraction score SE and lignocellulose grading score SF; wherein, the calculation formula for SE is: SE = 0.60A + 0.25T - 0.15Ract; The formula for calculating SF is: SF = 0.60F + 0.20T - 0.20Rfrac; Among them, Ract represents the degradation risk of active ingredients, and Rfrac represents the lignocellulose grading risk; (6) Based on SE and SF and their difference, the candidate systems of eutectic solvents are sorted and classified, and the output mode is determined, including the active ingredient extraction priority mode, the lignocellulose grading priority mode, the combined treatment mode or the segmented synergistic treatment mode.

2. The method for multi-scale molecular simulation screening of aromatic eutectic solvents based on a multi-objective collaborative scoring model according to claim 1, characterized in that, In step (1), the target molecule model library includes plant aromatic active ingredients, polysaccharide fragments, and lignin bond types, wherein the polysaccharide fragments include glucan fragments and xylan fragments, and the lignin bond types include one or more of β-O-4, 4-O-5, β-1, and β-5 linkage structures; The eutectic solvent candidate component library includes hydrogen bond acceptors, aliphatic hydrogen bond donors, aromatic hydrogen bond donors, and candidate molar ratios; wherein, the hydrogen bond acceptor is choline chloride, the aliphatic hydrogen bond donor is ethylene glycol or glycerol, and the aromatic hydrogen bond donor is guaiacol or p-toluenesulfonic acid; when the eutectic solvent candidate system is a binary system, the candidate molar ratio is 1:2, and when the eutectic solvent candidate system is a ternary system, the candidate molar ratio is 1:2:

1.

3. The method for multi-scale molecular simulation screening of aromatic eutectic solvents based on a multi-objective collaborative scoring model according to claim 1, characterized in that, In step (2), the quantum chemical calculation includes the following steps: The target molecule is input into the quantum chemistry calculation module for geometric optimization and single-point energy calculation, resulting in the optimized structure file, energy file and wave function file of the target molecule. Based on the candidate systems of eutectic solvents, initial molecular clusters of eutectic solvents are constructed; the optimized target molecules are combined with the initial molecular clusters of eutectic solvents to construct eutectic solvent-target molecule composite clusters. The initial molecular clusters of the eutectic solvent and the eutectic solvent-target molecular complex clusters are input into the conformation search program. The lowest energy conformation is selected as the initial structure for semi-empirical quantum chemical screening and then optimized. Density functional theory calculations were performed on the molecular clusters and composite clusters after initial screening. Geometric optimization and single-point energy calculations were performed on the initial molecular clusters in eutectic solvent. Basis set overlap error correction was performed on the eutectic solvent-target molecular composite clusters. Quantum chemical descriptors were extracted based on the generated quantum chemical wavefunction data.

4. The method for multi-scale molecular simulation screening of aromatic eutectic solvents based on a multi-objective collaborative scoring model according to claim 1, characterized in that, In step (2), the electrostatic potential descriptor includes the maximum positive electrostatic potential ESPmax, the minimum negative electrostatic potential ESPmin, the electrostatic potential range ΔESP, and the local electrostatic potential. Molecular polarity descriptors include molecular polarity index (MPI), polar surface area (PSA), and polar surface area ratio (FPSA). The AIM bond critical point includes the bond critical point electron density ρ(rBCP), the bond critical point electron density Laplace value ∇²ρ(rBCP), the energy density E(rBCP), and the hydrogen bond binding energy; The BSSE-corrected interaction energy is the interaction energy corrected for basis set overlap error and is decomposed into electrostatic interaction energy Eele, van der Waals interaction energy EvdW, repulsion energy Erep, and total interaction energy Etotal.

5. The method for multi-scale molecular simulation screening of aromatic eutectic solvents based on a multi-objective collaborative scoring model according to claim 1, characterized in that, In step (3), the aromatic plane of the target aromatic active ingredient or lignin aromatic unit is used as the reference plane. The probability density ρπ of the aromatic ring centroid of the aromatic hydrogen bond donor in the region of 0.30-0.50 nm above and below the plane is calculated, and the average probability density ρ0 of the aromatic ring centroid in the entire simulation box is calculated. The aromatic spatial enrichment coefficient Kπ is calculated according to the formula Kπ=ρπ / ρ0.

6. The method for multi-scale molecular simulation screening of aromatic eutectic solvents based on a multi-objective collaborative scoring model according to claim 1, characterized in that, In step (3), the formula for calculating the glycosidic bond interaction index Ggly is: Ggly=0.40N⁺[ggly(rp)]+0.30N⁺(NHBgly)+0.30N⁺(-Eele,gly) Where ggly(rp) is the intensity of the first peak of the radial distribution function of the key atom in the eutectic solvent relative to the oxygen in the glycosidic bond, NHBgly is the average number of hydrogen bonds near the glycosidic bond, Eele,gly is the electrostatic interaction energy between the eutectic solvent and the fragment near the glycosidic bond, and N + This is a favorable normalization function.

7. The method for multi-scale molecular simulation screening of aromatic eutectic solvents based on a multi-objective collaborative scoring model according to claim 1, characterized in that, In step (3), the formula for calculating the lignin bond affinity index Lβ is: Lβ=0.40N⁺(-Etotal,β)+0.30N⁺(Kπ,lignin)+0.30N⁺(NHBβ) Wherein, Etotal,β is the total interaction energy between the eutectic solvent and the corresponding lignin bond type segment; the corresponding lignin bond type segment is β-O-4, β-1, β-5 or 4-O-5; Kπ,lignin is the aromatic space enrichment coefficient of the aromatic hydrogen bond donor near the lignin aromatic unit; NHBβ is the average number of hydrogen bonds formed between the eutectic solvent and the vicinity of the lignin bond type; and N⁺ is the favorable normalization function.

8. The method for multi-scale molecular simulation screening of aromatic eutectic solvents based on a multi-objective collaborative scoring model according to claim 1, characterized in that, In step (6), when SE≥0.70 and SE-SF≥0.10, the active ingredient extraction priority mode is output; when SF≥0.70 and SF-SE≥0.10, the lignocellulose grading priority mode is output; when SE≥0.60, SF≥0.60 and |SE-SF|<0.10, the balanced treatment mode is output. Each eutectic solvent candidate system is sorted from highest to lowest SE to obtain the first active ingredient extraction candidate system, and sorted from highest to lowest SF to obtain the first lignocellulose grading candidate system. When the first active ingredient extraction candidate system and the first lignocellulose grading candidate system are not the same eutectic solvent candidate system, and the first active ingredient extraction candidate system satisfies SE≥0.70 and the first lignocellulose grading candidate system satisfies SF≥0.70, the segmented collaborative processing mode is output.

9. A multi-scale molecular simulation screening system for aromatic eutectic solvents based on a multi-objective collaborative scoring model, characterized in that, include: The model input module is used to build a target molecule model library and a eutectic solvent candidate component library; The quantum chemical calculation module is used to perform quantum chemical calculations on target molecules and eutectic solvent candidate systems and output quantum chemical descriptors. The molecular dynamics simulation module is used to construct molecular dynamics simulation systems that include target molecules and eutectic solvent components, and output molecular dynamics descriptors; The multi-scale feature extraction and fusion module is used to receive quantum chemical descriptors and dynamic descriptors, normalize them, group and fuse them, and output multi-scale feature vectors. The multi-objective collaborative scoring module is used to read multi-scale feature vectors and calculate the active ingredient extraction score (SE) and the lignocellulose grading score (SF). The mode discrimination and output module is used to determine the preferred mode based on SE, SF and the difference between the two, and output the eutectic solvent formulation, molar ratio and application mode.

Images