Plant exosome extraction process based on ai large model

Abstract

The application discloses a plant exosome extraction process based on an AI large model, which comprises the following steps: pretreating plant raw materials to obtain plant homogenate; sequentially performing differential centrifugation and polymer precipitation on the plant homogenate to obtain exosome crude extract; sequentially performing tangential flow filtration and size exclusion chromatography purification on the exosome crude extract; performing component analysis on the purified plant exosome by using liquid chromatography-mass spectrometry; inputting the compound identification data into the AI large model for function screening, and outputting a compound list with target function activity, and the application relates to the technical field of plant extracts, and through the synergistic effect of the combination of a separation and purification system and an AI auxiliary function screening system, the problems of low extraction efficiency and low purity and low function screening efficiency in the prior art are effectively solved, significant technical progress is achieved, and the application has good industrial practical value and market application prospect.

Description

Technical Field

[0001] This invention relates to the field of plant extract technology, specifically to a plant exosome extraction process based on an AI large model. Background Technology

[0002] Plant exosome-like nanovesicles are a type of membrane-bound vesicle with a diameter of 30-200 nm secreted by plant cells. They are rich in bioactive components such as proteins, lipids, nucleic acids, and small molecule metabolites, and have broad application prospects in drug delivery, tissue repair, anti-inflammatory and antioxidant fields.

[0003] Currently, methods for extracting plant exosomes mainly include ultracentrifugation, polymer precipitation, size exclusion chromatography, and immunoaffinity chromatography. However, existing technologies have the following drawbacks: First, the lack of systematic optimization of extraction process parameters makes it difficult to achieve both exosome yield and purity. For example, although ultracentrifugation yields high purity, the equipment is expensive, the processing capacity is small, and the time consumption is long; although polyethylene glycol (PEG) precipitation is simple to operate, the purity is low, and key parameters such as PEG concentration, incubation time, and centrifugal force are mostly set based on experience, lacking scientific parameter optimization.

[0004] Second, existing extraction processes only focus on physical separation and lack functional screening of active ingredients in the extracted products. Plant exosomes contain hundreds or even thousands of compounds, and traditional methods require bioactivity testing of each one, which is inefficient and makes it difficult to quickly identify target components with specific functions.

[0005] Third, existing technologies have not systematically integrated artificial intelligence into the entire process of plant exosome extraction and screening, making it impossible to achieve a paradigm shift from "blind extraction" to "function-oriented extraction".

[0006] Therefore, it is of great significance to develop a plant exosome extraction process that can balance extraction efficiency and purity and efficiently screen functional active ingredients. Summary of the Invention

[0007] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a plant exosome extraction process based on an AI large model, which solves the technical problems of lack of optimization of extraction process parameters and low efficiency of product function screening in existing technologies.

[0008] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a plant exosome extraction process based on an AI large model, comprising the following steps: S1: Pre-treat the plant raw materials to obtain a plant homogenate; S2: The plant homogenate is subjected to differential centrifugation and polymer precipitation in sequence to obtain crude exosome extract; S3: The crude exosome extract was sequentially purified by tangential flow filtration and size exclusion chromatography to obtain purified plant exosomes; S4: The purified plant exosomes were analyzed by liquid chromatography-mass spectrometry to obtain compound identification data; S5: Input the compound identification data into the AI ​​large model for functional screening, and output a list of compounds with target functional activities.

[0009] In the above-mentioned technical solution of the present invention, steps S1-S3 constitute a physical separation process for the efficient extraction of exosomes from plant raw materials. Large particulate impurities are removed by differential centrifugation, and exosomes are initially enriched by polymer precipitation. Then, deep purification is carried out by tangential flow filtration and size exclusion chromatography to obtain high-purity and high-activity plant exosomes. Steps S4-S5 use liquid chromatography-mass spectrometry to perform non-targeted metabolomics analysis on the extracted exosomes, and then use an AI big data model to intelligently screen the identified compounds to quickly identify active ingredients with specific functions (such as antioxidant, anti-inflammatory, tissue repair, etc.). This technical solution realizes the integration of the entire process of plant exosome extraction, purification, analysis and screening, which significantly improves the process efficiency and the accuracy of product function discovery.

[0010] In some embodiments, the pretreatment in step S1 includes: mixing plant material with pre-cooled phosphate buffer at a weight-to-volume ratio of 1:2 to 1:5, homogenizing at high speed at 0 to 10°C, and filtering through a 0.22 to 0.45 μm filter membrane.

[0011] In the above technical solution, the material-liquid ratio of 1:2 to 1:5 can ensure that the plant cells are fully broken down without affecting the subsequent separation; the low temperature of 0 to 10℃ can prevent the exosomes from degrading during the extraction process; and filtration through a 0.22 to 0.45 μm filter membrane can effectively remove unbroken cells and large particles of residue, laying a good foundation for subsequent centrifugation purification.

[0012] In some embodiments, the polymer precipitation in step S2 is carried out using polyethylene glycol, wherein the mass-volume concentration of the polyethylene glycol is 8% to 15%, and the precipitation incubation time is 16 to 24 hours.

[0013] In the above technical solution, polyethylene glycol (PEG) competitively binds with water molecules, causing exosomes to precipitate and accumulate under hydrophobic conditions. This invention uses response surface methodology to optimize and determine the parameter range of PEG concentration (8%~15%) and incubation time (16~24 hours). Within this range, a high exosome yield and stable precipitation effect can be obtained. If the concentration is too low or the time is too short, the precipitation will be incomplete; if the concentration is too high or the time is too long, too many impurities may be introduced, affecting subsequent purification.

[0014] In some embodiments, the tangential flow filtration in step S3 uses a polyethersulfone hollow fiber membrane module with a pore size of 500~1000kDa, and uses a continuous percolation mode for buffer replacement, with a percolation volume of 3~12 times the sample volume.

[0015] In the above technical solution, hollow fiber membranes with pore sizes of 500~1000kDa are used, which can effectively retain exosomes (particle size 30-200nm) while removing small molecular weight PEG residues and impurity proteins. The continuous percolation mode can realize the simultaneous replacement of buffer and removal of impurities. A percolation volume of 3~12 times can ensure an impurity removal rate of >99.7%. Compared with the traditional ultracentrifugation method, this process reduces the processing time by more than 50% and is easy to scale up to industrial scale.

[0016] In some embodiments, the size exclusion chromatography in step S3 uses an agarose gel column and isocratic elution with phosphate buffer as the mobile phase.

[0017] In the above technical solution, agarose gel (such as Sepharose CL-2B) has a suitable pore size distribution, which allows exosomes to be eluted in the pore volume, while small molecule impurities (such as residual PEG, salt ions, etc.) are retained in the gel pores, thereby achieving further purification. The isocratic elution operation is simple, has good repeatability, and the purity of the obtained exosomes can meet the requirements of subsequent LC-MS analysis.

[0018] In some embodiments, the liquid chromatography-mass spectrometry (LC-MS) technique in step S4 employs an ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS) system. The chromatographic conditions are as follows: a C18 column is used, the column temperature is 35-45°C, mobile phase A is an aqueous solution containing 0.05%-0.2% formic acid, and mobile phase B is an acetonitrile solution containing 0.05%-0.2% formic acid, with linear gradient elution of phase B at a volume fraction of 2%-98%. The mass spectrometry conditions are as follows: an electrospray ionization source is used, and data-dependent acquisition is performed in positive and negative ion modes.

[0019] In the above technical solutions, the C18 chromatographic column has broad applicability to plant-derived secondary metabolites; formic acid, as a mobile phase additive, can improve ionization efficiency; linear gradient elution can effectively separate compounds with large polarity ranges within 12 minutes; positive and negative ion mode switching and data-dependent acquisition can maximize the acquisition of mass spectrometry information of compounds, providing a high-quality data foundation for subsequent AI screening.

[0020] In some embodiments, before step S4, step S3-1 is further included to characterize the physicochemical properties of the purified plant exosomes; The characterization includes observing morphology using transmission electron microscopy, determining particle size distribution using nanoparticle tracking analysis, determining zeta potential using dynamic light scattering, and determining protein concentration using protein quantification.

[0021] In the above technical solution, systematic physicochemical characterization of exosomes before LC-MS analysis ensures that the sample quality for component analysis meets the requirements. Transmission electron microscopy can confirm the morphological integrity of exosomes; nanoparticle tracking analysis can determine the particle size distribution (usually 80-200 nm); Zeta potential can characterize the surface charge state of exosomes (usually negative); and protein quantification can assess sample concentration. These characterization data serve as a verification basis for the quality of exosome extraction and can also be used as auxiliary information for subsequent learning of AI models.

[0022] In some embodiments, the large AI model in step S5 is a DeepSeek model, and the function filtering includes: The compound identification data was input into the DeepSeek-DF model for matching confidence filtering. The compounds filtered by matching confidence are input into the DeepSeek-Bio model for biological relevance screening; The DeepSeek-R1 model was used to perform semantic recognition and functional annotation on the screened compounds.

[0023] In the above technical solution, the DeepSeek-DF model is used to filter compounds with low matching confidence to ensure the reliability of the screening results; the DeepSeek-Bio model is based on a biological knowledge graph to screen compounds related to target functions (such as anti-oxidation, anti-inflammation, anti-apoptosis, cell proliferation, tissue regeneration, etc.); the DeepSeek-R1 model uses semantic recognition technology to annotate and classify the screened compounds. This three-layer screening architecture can quickly identify active ingredients with target functions from hundreds or thousands of compounds, and the screening efficiency is tens of times higher than that of traditional bioactivity testing.

[0024] In some embodiments, steps S2 to S5 are all performed at 0 to 10°C, and in step S3, 0.1% to 0.5% bovine serum albumin is added to the buffer solution for tangential flow filtration to prevent exosome adhesion.

[0025] In the above technical solution, the low temperature condition of 0~10℃ can effectively inhibit the activity of endogenous enzymes and prevent the degradation of exosomes; the addition of 0.1%~0.5% bovine serum albumin (BSA) can form a protective layer on the membrane surface and tube wall, reduce the non-specific adsorption of exosomes, and improve the recovery rate.

[0026] In some embodiments, the differential centrifugation includes sequential low-speed centrifugation, medium-speed centrifugation, and high-speed centrifugation; the low-speed centrifugation has a centrifugal force of 200~500×g and a time of 5~15 minutes; the medium-speed centrifugation has a centrifugal force of 1500~3000×g and a time of 15~30 minutes; and the high-speed centrifugation has a centrifugal force of 8000~15000×g and a time of 20~40 minutes.

[0027] In the above technical solution, gradient centrifugation can remove impurities of different sizes, such as cell debris, organelles, and microvesicles, step by step, providing a clear supernatant for subsequent PEG precipitation. The centrifugation force range and time window have been optimized to ensure effective impurity removal while avoiding the loss of exosomes during centrifugation.

[0028] (III) Beneficial Effects The beneficial effects of this invention are: This invention employs a combined separation and purification system of differential centrifugation, PEG precipitation, tangential flow filtration, and size exclusion chromatography in the extraction process. The optimal parameter ranges for PEG mass-volume concentration, precipitation incubation, and precipitation centrifugation force were determined using response surface methodology. Within these parameter ranges, the exosome yield can reach over 4.24 g / kg of raw material. Simultaneously, synergistic purification through tangential flow filtration and size exclusion chromatography is achieved. Tangential flow filtration utilizes a hollow fiber membrane module with a pore size of 500–1000 kDa for continuous percolation, with a percolation volume of 3–12 times the sample volume. Size exclusion chromatography uses an agarose gel column for isocratic elution. The combined use of these two methods effectively removes PEG residues and impurities, achieving an impurity removal rate of over 99.7%. This significantly solves the problems of low purity in existing technologies such as PEG precipitation and small throughput in ultracentrifugation, achieving a simultaneous improvement in both high yield and high purity.

[0029] This invention introduces liquid chromatography-mass spectrometry (LC-MS) combined with an AI-powered large-scale model for component analysis and functional screening after purification. The LC-MS utilizes a C18 column, a column temperature of 35–45°C, a mobile phase system of 0.05%–0.2% formic acid, and a data-dependent acquisition method that switches between positive and negative ion modes, enabling comprehensive acquisition of compound information from exosomes. The AI-powered large-scale model employs a three-layer progressive screening architecture, rapidly identifying active ingredients with target functions such as antioxidant, anti-inflammatory, anti-apoptotic, cell proliferation-promoting, and tissue regeneration from hundreds or thousands of compounds. Compared to traditional manual literature retrieval, this AI screening scheme significantly improves efficiency and effectively solves the problem of low efficiency in functional screening in existing technologies.

[0030] In steps S2 to S5 of this invention, a low-temperature operation of 0~10℃ is used throughout the process, and 0.1%~0.5% bovine serum albumin is added to the tangential flow filtration buffer. The low temperature condition can effectively inhibit the activity of endogenous enzymes and prevent exosome degradation. Bovine serum albumin can form a protective layer on the membrane surface and tube wall and reduce the non-specific adsorption of exosomes. The combined effect of the two can increase the exosome recovery rate by more than 30% and maintain a uniform and stable particle size distribution, effectively solving the problems of easy degradation and adsorption loss of exosomes during the extraction process.

[0031] This invention systematically optimizes the process parameters for each step, determining the optimal parameter ranges and operating conditions for differential centrifugation, PEG precipitation, tangential flow filtration, size exclusion chromatography, liquid chromatography-mass spectrometry, and AI screening. The process is stable and controllable, and easy to implement for industrial production. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the process flow of the present invention. Detailed Implementation

[0033] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0034] Example 1: Please refer to Figure 1 This embodiment provides a plant exosome extraction process based on an AI large model, and the specific steps are as follows: S1. Raw material pretreatment: Harvest 500g of fresh black goji berries that are uniformly ripe and free from pests and diseases. Remove diseased, damaged, and impurity berries to ensure the purity of the raw material. Mix the selected berries with pre-cooled phosphate buffered saline (PBS, pH 7.4) at a ratio of 1:3 (w / v). Homogenize at 10,000 rpm for 5 minutes at 4°C using a high-speed homogenizer. Maintain the low temperature during homogenization using an ice bath. Filter the homogenate sequentially through 3μm and 0.45μm regenerated cellulose membranes at a pressure of 1.5 bar and 4°C to remove unbroken cells and large particles, yielding approximately 1500mL of clear homogenate.

[0035] S2, Differential centrifugation and PEG precipitation: The filtered homogenate was subjected to three-stage differential centrifugation at 4°C: low speed centrifugation at 300×g for 10 minutes, supernatant was collected and precipitate was discarded; medium speed centrifugation at 2000×g for 20 minutes, supernatant was collected and precipitate was discarded; high speed centrifugation at 10000×g for 30 minutes, supernatant was collected and precipitate was discarded, yielding approximately 1450 mL of supernatant.

[0036] Slowly add PEG6000 stock solution (50% w / v) to the supernatant while stirring until the final PEG concentration reaches 12% (w / v). Place the mixture in a 4°C refrigerator and incubate for 22 hours. After incubation, centrifuge at 9720×g for 60 minutes (4°C), collect the precipitate, and resuspend the precipitate in 10 mL of pre-cooled PBS to obtain crude exosome extract.

[0037] S3. Tangential flow filtration and size exclusion chromatography purification: A 750 kDa polyethersulfone hollow fiber membrane module (membrane area 0.01 m²) was connected using the ÄKTA Pure FPLC system. 2 The fiber had an inner diameter of 300 μm and a length of 60 mm. The injection flow rate was set at 8 mL / min, the transmembrane pressure was controlled below 1 bar, and the temperature was 4 °C. Continuous percolation was used for buffer replacement, with the percolation buffer being PBS (pH 7.4) containing 0.2% BSA. The percolation volume was 8 times the sample volume (i.e., 80 mL). After percolation, the sample was concentrated to 10 mL, and the retentate was collected.

[0038] Using a Sepharose CL-2B gel column (20 mL column volume, 1.6 cm inner diameter, 10 cm column height), isocratic elution was performed with PBS (pH 7.4) as the mobile phase at a flow rate of 0.8 mL / min. 1 mL of the tangential flow filtration concentrate was loaded onto the column, and the elution peak (approximately 6-10 mL) was collected to obtain the purified plant exosome suspension.

[0039] S3-1, Physicochemical property characterization: The purified sample was subjected to multi-dimensional characterization: morphology was observed using transmission electron microscopy. 10 μL of sample was added to a copper grid, allowed to stand for 5 minutes, and excess liquid was absorbed with filter paper. Then, 10 μL of 2% phosphotungstic acid (pH 6.8) was added for negative staining for 2 minutes, excess stain was absorbed with filter paper, and the sample was dried at room temperature before observation under a transmission electron microscope. Particle size distribution was determined using nanoparticle tracking analysis. The sample was diluted to a suitable concentration (approximately 1 × 10⁻⁶). 8 The concentration of particles (particles / mL) was measured using a NanoSight NS300. Zeta potential was determined using dynamic light scattering; samples were diluted to an appropriate concentration and measured using a Zetasizer NanoZS. Protein concentration was determined using a protein quantification method, employing the BCA method. 10 μL of sample was added to 200 μL of BCA working solution, incubated at 37°C for 30 minutes, and absorbance was measured at 562 nm. Protein concentration was calculated based on the standard curve.

[0040] S4, LC-MS component analysis: Untargeted metabolomics analysis was performed using an ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS / MS) system. Chromatographic conditions: HSST3 column (100 mm × 2.1 mm, 1.8 μm), column temperature 40 °C; mobile phase A: 0.1% formic acid aqueous solution; mobile phase B: 0.1% formic acid acetonitrile solution; gradient elution program: 0 min 2% B, 12 min 98% B, linear gradient; flow rate 0.3 mL / min; injection volume 5 μL. Mass spectrometry conditions: electrospray ionization (ESI), switching between positive and negative ion modes, spray voltage 3.5 kV (positive ion mode) and 3.0 kV (negative ion mode), sheath gas flow rate 45 arb, auxiliary gas flow rate 15 arb, capillary temperature 320 °C; data-dependent acquisition mode, mass range m / z 50-1500, primary mass spectrometry resolution 70000, secondary mass spectrometry resolution 17500, collision energy stepwise (20 / 40 / 60 eV).

[0041] The raw data were converted to mzML format using ProteoWizard software; peak extraction, peak alignment, and compound identification were performed using CompoundDiscoverer 3.3 software with the following parameter settings: maximum retention time offset 0.5 min, quality tolerance 10 ppm, and signal-to-noise ratio threshold 3; database matching: mzCloud, KEGG, HMDB, and MassBank, with a matching score ≥80 retained.

[0042] S5, AI Large Model Function Filtering: The list of exosome-specific compounds identified by LC-MS was input into the DeepSeek model for a three-layer progressive screening: First, the compound identification data was input into the DeepSeek-DF model for matching confidence filtering, retaining compounds with an MS / MS matching score ≥80; Second, the compounds after matching confidence filtering were input into the DeepSeek-Bio model for biological relevance screening, selecting compounds related to five functional categories: antioxidant, anti-inflammatory, anti-apoptotic, cell proliferation-promoting, and tissue regeneration-promoting, retaining compounds with a relevance score ≥0.7; Third, the DeepSeek-R1 model was used to perform semantic recognition and functional annotation on the screened compounds, outputting a list of compounds with the target functional activity.

[0043] This invention proposes a plant exosome extraction process based on an AI large-scale model. The extraction process employs a combined separation and purification system of differential centrifugation, PEG precipitation, tangential flow filtration, and size exclusion chromatography. Differential centrifugation removes large particulate impurities, providing a clear environment for PEG precipitation. Response surface methodology is used to optimize PEG concentration and incubation time to achieve high yield. Tangential flow filtration removes PEG residues and impurities, and size exclusion chromatography further refines the purity, achieving high purity. After purification, LC-MS combined with the AI ​​large-scale model is introduced: LC-MS comprehensively identifies compounds, and the DeepSeek model rapidly identifies target active ingredients through a three-layer progressive screening process (DF, Bio, R1), solving the problem of low efficiency in functional screening. Simultaneously, low-temperature operation throughout the process and the addition of BSA inhibit degradation and reduce adsorption, synergistically protecting exosome activity and recovery rate. This addresses the technical problems of existing technologies that struggle to balance extraction efficiency and purity, and suffer from low efficiency in functional screening. Example 2

[0044] The difference between this embodiment and Embodiment 1 is that some process parameters have been adjusted, as follows: S2. Differential centrifugation and PEG precipitation: The final PEG concentration was adjusted to 8% (w / v), and the incubation time was adjusted to 24 hours. The remaining steps and parameters were the same as in Example 1. Example 3

[0045] The difference between this embodiment and Embodiment 1 is that some process parameters have been adjusted, as follows: S2. Differential centrifugation and PEG precipitation: The final PEG concentration was adjusted to 15% (w / v), and the incubation time was adjusted to 16 hours. The remaining steps and parameters were the same as in Example 1. Example 4

[0046] The difference between this embodiment and Embodiment 1 is that some process parameters have been adjusted, as follows: S3. Tangential flow filtration and size exclusion chromatography purification: The pore size of the tangential flow filtration membrane was adjusted to 500 kDa. The remaining steps and parameters were the same as in Example 1. Example 5

[0047] The difference between this embodiment and Embodiment 1 is that some process parameters have been adjusted, as follows: S3. Tangential flow filtration and size exclusion chromatography purification: The pore size of the tangential flow filtration membrane was adjusted to 1000 kDa. The remaining steps and parameters were the same as in Example 1. Example 6

[0048] The difference between this embodiment and Embodiment 1 is that the plant material has been replaced, as detailed below: S1. Raw material pretreatment: Harvest 500 g of fresh mulberry root bark (Morus alba L.), remove the outer bark and impurities, mix with pre-cooled phosphate buffer (PBS, pH 7.4) at a ratio of 1:4 (w / v), and homogenize at 10,000 rpm for 5 minutes at 4°C using a high-speed homogenizer. The remaining pretreatment steps are the same as in Example 1. Steps S2 to S5 are the same as in Example 1. Example 7

[0049] The difference between this embodiment and Embodiment 1 is that the plant material has been replaced, as detailed below: S1. Raw material pretreatment: Harvest 500 g of fresh lemon (Citrus limon) fruit, remove the peel and only take the pulp, mix it with pre-cooled phosphate buffer (PBS, pH 7.4) at a ratio of 1:2 (w / v), and homogenize it at 10,000 rpm for 5 minutes at 4°C using a high-speed homogenizer. The remaining pretreatment steps are the same as in Example 1. Steps S2 to S5 are the same as in Example 1. Example 8

[0050] The difference between this embodiment and Embodiment 1 is that the operating temperature has been adjusted, as detailed below: Steps S2 to S5 are all carried out at 0°C (using an ice water bath and a 4°C cold storage), and the remaining steps and parameters are the same as in Example 1.

[0051] Comparative Example 1 The difference between this comparative example and Example 1 is that step S6 (AI large model functional screening) is omitted, and the compounds identified by LC-MS are screened for function by manual literature search. The remaining steps are the same as in Example 1.

[0052] The specific procedure is as follows: Three researchers searched the PubMed, Web of Science, and CNKI databases using compound names and functional keywords, with each person working 8 hours a day.

[0053] Comparative Example 2 The difference between this comparative example and Example 1 is that the S3 tangential flow filtration and size exclusion chromatography purification steps are replaced by traditional ultracentrifugation, and tangential flow filtration and size exclusion chromatography are omitted. The remaining steps are the same as in Example 1.

[0054] The specific procedure is as follows: Centrifuge the crude exosome extract obtained by S2 at 100,000×g for 70 minutes (4℃), discard the supernatant, resuspend the precipitate in PBS, and repeat once.

[0055] Comparative Example 3 The difference between this comparative example and Example 1 is that the PEG precipitation parameters in S2 exceed the preferred range of this invention, the final PEG concentration is adjusted to 5% (less than 8%), the incubation time is adjusted to 12 hours (less than 16 hours), and the remaining steps are the same as in Example 1.

[0056] Comparative Example 4 The difference between this comparative example and Example 1 is that the PEG precipitation parameters in S2 exceed the preferred range of this invention, the final PEG concentration is adjusted to 20% (higher than 15%), and the incubation time is adjusted to 30 hours (higher than 24 hours). The remaining steps are the same as in Example 1.

[0057] Comparative Example 5 The difference between this comparative example and Example 1 is that steps S2 to S5 were all performed at room temperature (25°C), and BSA was not added to the tangential flow filtration buffer in S3. The remaining steps were the same as in Example 1.

[0058] Comparative Example 6 The difference between this comparative example and Example 1 is that the size exclusion chromatography step in S3 is omitted, and only tangential flow filtration is used for purification. The remaining steps are the same as in Example 1.

[0059] Comparative Example 7 The difference between this comparative example and Example 1 is that the differential centrifugation step in S2 is omitted, and the homogenate filtered in S1 is directly subjected to PEG precipitation. The remaining steps are the same as in Example 1.

[0060] Performance testing The following performance tests were performed on the products obtained in Examples 1-8 and Comparative Examples 1-7: Exosome yield determination: The protein concentration of the purified exosome suspension was determined by the BCA method, and the yield was calculated according to the formula: Yield (g / kg) = (protein concentration × suspension volume) / raw material mass. The result is expressed as g / kg raw material.

[0061] Purity determination (removal rate of contaminating proteins): SDS-PAGE gel electrophoresis combined with grayscale analysis was used to compare the relative intensity of contaminating protein bands in the samples before and after purification. The contaminating protein removal rate (%) was calculated as follows: (grayscale value of contaminating protein before purification - grayscale value of contaminating protein after purification) / grayscale value of contaminating protein before purification × 100%.

[0062] Particle size distribution determination: The particle size distribution of exosomes was determined using a nanoparticle tracking analyzer, and the average particle size and particle size range were recorded.

[0063] Zeta potential measurement: The surface potential of exosomes was measured using a Zeta potential analyzer.

[0064] AI-screened active compound count: Record the number of compounds with target functional activity output by the DeepSeek model in step S6; Comparative Example 1 records the number of active compounds identified by manual literature retrieval.

[0065] Processing time: Record the total time from the start of raw material pretreatment of S1 to obtaining the final screening result (excluding the time for physicochemical property characterization). Comparative Example 1 records the time spent on manual screening separately.

[0066] The results of each performance test are shown in the table below. The comparison between Example 1 and Comparative Example 2 shows that the yield of Example 1 (4.24 g / kg) was significantly higher than that of Comparative Example 2 (2.87 g / kg), an increase of 32.3%. This indicates that the present invention, by using tangential flow filtration coupled with size exclusion chromatography to replace traditional ultracentrifugation, can effectively improve the exosome recovery rate. Meanwhile, the purity of Example 1 (>99.8%) was better than that of Comparative Example 2 (>99.2%), indicating that the purification effect of TFF+SEC coupled with ultracentrifugation alone is superior.

[0067] The comparison between Example 1 and Comparative Example 6 shows that: after omitting the size exclusion chromatography step, the yield (4.31 g / kg) of Comparative Example 6 was slightly higher than that of Example 1 (due to sample loss without the SEC step), but the purity (>98.2%) decreased significantly, the impurity removal rate decreased by 1.6 percentage points, and the number of active compounds screened by AI (35 kinds) decreased significantly. This indicates that the SEC step plays an irreplaceable role in removing residual PEG and trace amounts of impurities and improving sample purity.

[0068] The comparison between Example 1 and Comparative Example 7 shows that after omitting the differential centrifugation step, the yield of Comparative Example 7 was artificially high (5.12 g / kg, due to the presence of a large amount of non-exosome proteins), but the purity dropped sharply to 85.3%, and the particle size distribution became significantly wider (100-450 nm). This indicates the necessity of differential centrifugation for removing cell debris and organelles and ensuring the purity of exosomes.

[0069] The comparison of Examples 1, 2, and 3 shows that good yields and purity can be obtained within the parameter range of 8% to 15% PEG concentration and 16 to 24 hours of incubation time. Among them, Example 1 (12% PEG, 22h) showed the best results, while Comparative Example 3 (5% PEG, 12h) saw a sharp drop in yield of 71.0%, and Comparative Example 4 (20% PEG, 30h) showed a significant decrease in purity, which verifies the critical significance of the preferred parameter range of the present invention.

[0070] The comparison between Example 1 and Comparative Example 1 shows that Example 1 uses an AI large-scale model for screening, which can screen 46 active compounds from 1881 compounds in just a few minutes; while Comparative Example 1 uses manual literature search, which takes 240 hours and only identifies 12 active compounds, with a false negative rate of 74%. This indicates that the AI ​​screening scheme of the present invention has a higher efficiency and significant advantages in the comprehensiveness and accuracy of the screening results.

[0071] The comparison between Example 1 and Comparative Example 5 shows that: Comparative Example 5, operated at room temperature without the addition of BSA, resulted in a 39.6% decrease in yield (from 4.24 g / kg to 2.56 g / kg), a wider particle size distribution (80-250 nm), and a decrease in the absolute value of the Zeta potential (-3.21 mV), indicating that exosomes aggregated and degraded. This verifies the protective effect of the low temperature conditions of 0-10℃ and the addition of 0.1%-0.5% BSA on the activity and recovery rate of exosomes in this invention.

[0072] The results of Examples 6 (mulberry bark) and 7 (lemon) show that the process of the present invention has good adaptability to different plant sources, and the yield, purity, particle size distribution and other indicators all meet the requirements. The number of active compounds screened by AI is 51 and 38 respectively, which verifies the universality of the technical solution of the present invention.

[0073] In summary, this invention, through a combined separation and purification system of differential centrifugation, PEG precipitation, tangential flow filtration, and size exclusion chromatography, a screening system using LC-MS combined with DeepSeek's AI-assisted function, and the combination of low-temperature protection throughout the process and BSA anti-adhesion design, achieved excellent results in Examples 1-8, with yields ≥2.98 g / kg, purity ≥99.5%, and ≥38 AI-screened active compounds. In contrast, Comparative Examples 1-7 showed significant deficiencies in extraction efficiency, purity, screening efficiency, or exosome activity. This demonstrates the integrity and synergy of the technical solution of this invention and proves that the technical solution of this invention has significant progress and inventiveness compared to the prior art.

[0074] This invention effectively solves the problems of difficulty in balancing extraction efficiency and purity and low functional screening efficiency in existing technologies by combining a separation and purification system with an AI-assisted functional screening system. It has achieved significant technological progress and has good industrial practical value and market application prospects.

[0075] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0076] In the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0077] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make modifications, alterations, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A plant exosome extraction process based on an AI large-scale model, characterized in that, Includes the following steps: S1: Pre-treat the plant raw materials to obtain a plant homogenate; S2: The plant homogenate is subjected to differential centrifugation and polymer precipitation in sequence to obtain crude exosome extract; S3: The crude exosome extract was sequentially purified by tangential flow filtration and size exclusion chromatography to obtain purified plant exosomes; S4: The purified plant exosomes were analyzed by liquid chromatography-mass spectrometry to obtain compound identification data; S5: Input the compound identification data into the AI ​​large model for functional screening, and output a list of compounds with target functional activities.

2. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, The pretreatment in step S1 includes: mixing plant raw materials with pre-cooled phosphate buffer at a weight-to-volume ratio of 1:2 to 1:5, homogenizing at high speed at 0 to 10°C, and filtering through a 0.22 to 0.45 μm filter membrane.

3. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, In step S2, the polymer precipitation is carried out using polyethylene glycol, wherein the mass-volume concentration of polyethylene glycol is 8% to 15%, and the precipitation incubation time is 16 to 24 hours.

4. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, The tangential flow filtration described in step S3 uses a polyethersulfone hollow fiber membrane module with a pore size of 500~1000kDa, and adopts a continuous percolation mode for buffer replacement, with a percolation volume of 3~12 times the sample volume.

5. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, The size exclusion chromatography described in step S3 uses an agarose gel column and isocratic elution with phosphate buffer as the mobile phase.

6. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, The liquid chromatography-mass spectrometry (LC-MS) technique described in step S4 employs an ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS) system. The chromatographic conditions are as follows: a C18 column is used, with a column temperature of 35–45 °C; mobile phase A is an aqueous solution containing 0.05%–0.2% formic acid, and mobile phase B is an acetonitrile solution containing 0.05%–0.2% formic acid, using a linear gradient elution of phase B with a volume fraction of 2%–98%. The mass spectrometry conditions are as follows: an electrospray ionization source is used, and data-dependent acquisition is performed in positive and negative ion modes.

7. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, Before step S4, step S3-1 is also included, which characterizes the physicochemical properties of the purified plant exosomes; The characterization includes observing morphology using transmission electron microscopy, determining particle size distribution using nanoparticle tracking analysis, determining zeta potential using dynamic light scattering, and determining protein concentration using protein quantification.

8. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, The AI ​​large model mentioned in step S5 is the DeepSeek model, and the function filtering includes: The compound identification data was input into the DeepSeek-DF model for matching confidence filtering. The compounds filtered by matching confidence are input into the DeepSeek-Bio model for biological relevance screening; The DeepSeek-R1 model was used to perform semantic recognition and functional annotation on the screened compounds.

9. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, Steps S2 to S5 are all performed at 0~10℃, and in step S3, 0.1%~0.5% bovine serum albumin is added to the buffer solution of tangential flow filtration to prevent exosome adhesion.

10. The plant exosome extraction process based on an AI large model according to claim 1, characterized in that, The differential centrifugation includes low-speed centrifugation, medium-speed centrifugation, and high-speed centrifugation performed sequentially; the centrifugal force of the low-speed centrifugation is 200~500×g, and the time is 5~15 minutes; the centrifugal force of the medium-speed centrifugation is 1500~3000×g, and the time is 15~30 minutes; the centrifugal force of the high-speed centrifugation is 8000~15000×g, and the time is 20~40 minutes.

Images