Method for screening hericium erinaceus functional polypeptide based on artificial intelligence assistance and polypeptide or polypeptide composition

By combining the large language model and database of the DeepSeek platform, the significant functions of Hericium erinaceus peptides were successfully identified and verified, solving the problem of low efficiency in identifying Hericium erinaceus peptide functions, realizing efficient screening and functional verification, and promoting its application in functional foods and nutritional products.

CN121737244APending Publication Date: 2026-03-27ZHEJIANG UNIV OF TECH

Patent Information

Application Number
CN202511980777.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-25
Publication Date
2026-03-27

AI Technical Summary

Technical Problem

Existing technologies suffer from low efficiency in functional identification of Hericium erinaceus peptides, difficulty in short-chain sequence annotation, and a lack of systematic verification processes, making it difficult to achieve efficient functional peptide development.

Method used

Using a large language model based on the DeepSeek platform combined with the BioChemia and InterPro databases, high-throughput sequence-function matching was performed. Combined with mass spectrometry identification and in vitro activity assay, the significant ACE inhibitory and antioxidant functions of Hericium erinaceus peptides were identified and verified.

Benefits of technology

This study achieved efficient screening of Hericium erinaceus peptides, identifying short-chain peptides with significant functions, shortening the development cycle, reducing costs, improving identification accuracy and efficiency, and expanding their application in functional foods and nutritional products.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure FT_1
    Figure FT_1
  • Figure FT_2
    Figure FT_2
  • Figure FT_3
    Figure FT_3
Patent Text Reader

Abstract

The invention discloses a screening method of hericium erinaceus functional polypeptide based on artificial intelligence assistance and polypeptide or a polypeptide composition. The method comprises the following steps: (1) constructing an AI auxiliary annotation model; (2) extracting hericium erinaceus crude protein; (3) preparing hericium erinaceus protein enzymatic hydrolysate through composite enzymatic hydrolysis; (4) performing polypeptide identification on the hericium erinaceus protein enzymatic hydrolysate to obtain a polypeptide sequence; and (5) carrying out AI auxiliary annotation and polypeptide screening by using a large language model based on a DeepSeek platform so as to obtain a polypeptide sequence with assumed biological activity. The polypeptide sequence and the polypeptide composition which have remarkable ACE inhibition and oxidation resistance dual functions are successfully obtained by utilizing the method disclosed by the invention. Compared with a traditional experience screening path, the method realizes a high-throughput active peptide development process which is targeted in structure, accurate in prediction and capable of verifying a closed loop.
Need to check novelty before this filing date? Find Prior Art

Description

(I) Technical Field

[0001] This invention relates to the fields of food science, bioinformatics and functional peptide screening technology, and in particular to an artificial intelligence-assisted method for screening functional peptides of Hericium erinaceus and the obtained Hericium erinaceus peptides or peptide compositions. (II) Technical Background

[0002] Lion's mane mushroom ( Hericium erinaceus Hericium erinaceus (Lion's Mane Mushroom) is a medicinal and edible fungus widely distributed in Northeast, East, and Central China. Current research indicates that Hericium erinaceus is rich in various active ingredients, including proteins and terpenoids, with a particularly high protein content (15%–25%, dry weight), exhibiting potential bioactivity in antioxidation and blood pressure reduction. Despite its high nutritional value and functional development potential, current research and product development primarily focus on the overall biological functions of its crude extracts, lacking specific efficacy assessments of individual active ingredients. This limits its development and application as a high-value-added food or nutritional product.

[0003] Hericium erinaceus polypeptides are small molecular fragments obtained from its proteins through enzymatic hydrolysis or other processing methods. They possess advantages such as small molecular weight, high absorption and utilization rate, and well-defined biological activities. Studies have shown that compared to the parent protein, polypeptides are more easily absorbed by the gastrointestinal tract and can bind to biological targets more efficiently, thus exhibiting more significant biological activities in areas such as antioxidation and blood pressure reduction. However, current research on Hericium erinaceus polypeptides is still relatively limited. Their specific structures and mechanisms of action have not yet been fully elucidated, restricting their efficient development and application in functional foods and nutritional health products.

[0004] In recent years, the application of artificial intelligence (AI) in the field of functional factor identification has gradually emerged, especially the semantic mining and sequence annotation technologies based on language models, which provide a new path for the efficient screening of active ingredients in natural products. Large Language Models (LLMs), as open semantic understanding and generation frameworks, can acquire cross-domain knowledge transfer capabilities through large-scale corpus pre-training and, with the support of strategies such as instruction fine-tuning and retrieval enhancement, achieve joint analysis of unstructured documentary evidence and structured database information. Based on this, DeepSeek, as an emerging AI-based open semantic mining platform, can integrate cross-document and multi-database data using LLM large language models as its core. Through keyword recognition, context modeling, and semantic alignment, it performs functional association inference and preliminary annotation of protein or peptide sequences, thus demonstrating significant advantages in improving screening efficiency and shortening development cycles.

[0005] However, existing AI models, limited by peptide chain length, still face challenges in identifying and analyzing short-chain peptides due to limitations in accuracy, insufficient annotation granularity, and weak structural feature recognition capabilities when dealing with natural protein sources. For example, patent CN117510586A discloses an AI-based method for identifying umami peptides, capable of recognizing umami peptide sequences up to 12 oz in length. However, its model relies on pure computational simulation, lacks verification of the lower limit of short peptide recognition, and its functional evaluation depends on sensory evaluation using mixed peptide solutions, making it difficult to clarify the specific role of individual peptide segments. Patent CN 119775357A discloses a screening method and application for anti-fatigue peptides in yak blood. By separating yak blood proteins and conducting a series of in vitro activity tests, a specific peptide with potential activity, KD-8 (KTYFPHFD), was finally screened and its anti-fatigue efficacy was verified. However, this method still adopts the traditional "separation-activity verification-structure analysis" process, which suffers from long screening cycles, high experimental costs, and reliance on human experience, making it difficult to meet the diverse and high-throughput functional peptide development needs. (III) Summary of the Invention

[0006] This invention aims to overcome the shortcomings of existing technologies in terms of low efficiency in identifying functional peptides from Hericium erinaceus, difficulties in short-chain sequence annotation, and lack of systematic verification procedures. It proposes an efficient screening method for functional peptides from Hericium erinaceus based on artificial intelligence-assisted identification. This method utilizes a large language model to achieve high-throughput sequence-function matching, combined with mass spectrometry identification and in vitro activity assays, successfully obtaining key peptide sequences with significant ACE inhibition and antioxidant functions, and clarifying their contribution in the enzymatic hydrolysate. Compared to traditional empirical screening pathways, this invention achieves a high-throughput bioactive peptide development process with structure targeting and accurate prediction, providing an efficient and feasible technical path for the functional utilization and product transformation of Hericium erinaceus protein resources.

[0007] To achieve this objective, the present invention adopts the following technical solution:

[0008] Firstly, an artificial intelligence-assisted method for screening functional peptides from Hericium erinaceus includes the following steps:

[0009] (1) Constructing an AI-assisted annotation model: In the DeepSeek platform, with the LLM large language model as the core, one or a combination of "mushroom" and "peptide" as keywords, Paper-scrapper is called to perform high-throughput literature crawling in the PubMed database and output documents; secondly, inputting instructions in DeepSeek, asking it to identify the documents and extract peptide information from the documents, the peptide information including: peptide sequence, origin, structure, function, hydrophilic and hydrophobic residues, and establishing a data knowledge base of structure-function relationship; thirdly, the BioChemia database and InterPro database are connected to the data knowledge base through the external database calling mechanism to achieve comprehensive identification and functional annotation of short chain peptides and long chain peptides, thereby obtaining an AI-assisted annotation model that has completed training and closed-loop optimization;

[0010] (2) Extraction of crude protein from Hericium erinaceus: The fruiting bodies of Hericium erinaceus are dried, cut into pieces, crushed, and sieved to obtain dry powder; the dry powder and water are mixed at a material-to-liquid ratio of 1:(20~25) (w / v), the pH is adjusted to 13, and the mixture is extracted for 8~10 hours in a water bath at 50~70℃. The supernatant of the extract is collected by centrifugation, and the residue is extracted again by water extraction according to the above operation, so that the total number of water extraction operations is at least 2. The supernatants are combined, the pH of the supernatant is adjusted to 4.0, and the protein is allowed to stand to completely precipitate. The protein is collected by centrifugation to obtain crude protein; the crude protein is dissolved in NaOH solution and adjusted to neutral, and then concentrated and enriched by rotary evaporation at 55~65℃ to obtain crude protein extract of Hericium erinaceus.

[0011] (3) Preparation of Hericium erinaceus protein hydrolysate (H) by compound enzymatic hydrolysis erinaceus Protein Hydrolysate (HEph): The crude protein extract of Hericium erinaceus obtained in step (2) was centrifuged to obtain the supernatant, which was then enzymatically hydrolyzed with trypsin and alkaline protease in sequence. The resulting hydrolysate was filtered through an ultrafiltration membrane with a permeation pore size between 10 and 100 kDa to obtain Hericium erinaceus protein hydrolysate.

[0012] (4) Peptide identification: The peptides in the Hericium erinaceus protein hydrolysate obtained in step (3) were identified by liquid chromatography-tandem mass spectrometry (LC-MS / MS) to obtain peptide sequence results;

[0013] (5) AI-assisted annotation and peptide screening: The peptide sequence obtained in step (4) is used as input and imported into the AI-assisted annotation model based on the DeepSeek platform constructed in step (1) for sequence-function matching. The sequence and activity knowledge base of the annotation model is used as the reference library. Peptide sequences that meet any of the following conditions are judged as candidate sequences: When the peptide chain length is 3 to 8 amino acids, if the number of amino acid differences between the candidate sequence and any known sequence in the reference library does not exceed 2, it is judged as a similar sequence; When the peptide chain length is greater than 8 and does not exceed 25 amino acids, if the sequence similarity between the candidate sequence and any known sequence in the reference library is not less than 50%, and the number of amino acid substitutions or deletions does not exceed 50%, it is judged as a similar sequence. If four are found, they are considered similar sequences. Using these similar sequences as candidate sequences, a sequence-activity correspondence is established. To ensure comprehensive identification of both long-chain and short-chain peptides, the built-in analysis functions of the BioChemia (https: / / biochemia.uwm.edu.pl / biopep / peptide_data.php) and InterPro (https: / / www.ebi.ac.uk / interpro / ) databases are used to further confirm the potential biological activities of the candidate sequences and to fully annotate their potential biological activities, thereby obtaining peptides with presumed biological activities.

[0014] The present invention can further use the Fmoc solid-phase method to chemically synthesize the polypeptide with assumed biological activity obtained in step (5), and detect its potential biological activity to verify its actual biological activity, thereby obtaining a polypeptide with actual biological activity.

[0015] The 'high throughput' mentioned in step (1) of this invention refers to automatically processing and extracting peptide-related data of no less than a thousand scales within the same batch.

[0016] Preferably, the keywords include one or any combination of "mushroom", "peptide", "protein", "fungal", and "bioactive peptide".

[0017] In step (1) of this invention, the extracted peptide information is generated into a JSON array according to specified fields. The JSON array must contain the following fields: peptide sequence, origin, structure, function, and hydrophilic / hydrophobic residues. If relevant information is missing, "None" is filled in. If "mushroom" and "peptide" are not included, the data is discarded. This data is used to standardize and parse literature data and import it into the data knowledge base. Further, in step (1), after connecting the BioChemia database and the InterPro database to the data knowledge base through an external database calling mechanism, an instruction is entered in Deepseek to flatten the data based on the identified data and export an Excel file. The fields contained in the Excel file are the same as those in the JSON array. The Excel file is manually proofread and irrelevant data is discarded, thereby completing the training and closed-loop optimization of the AI-assisted annotation model.

[0018] As a preferred option, in step (2), the drying conditions for the fruiting bodies of Hericium erinaceus are: drying at 40-60℃ for 8-10 hours.

[0019] Preferably, in step (2), the sieve aperture is between 50 and 80 mesh to achieve a balance between extraction efficiency and economy.

[0020] Preferably, in step (2), the centrifugation is performed at (8,000~10,000)×g for 10-20 minutes.

[0021] Preferably, in step (2), the pH is adjusted to 13 using sodium hydroxide solution and the pH of the supernatant is adjusted to 4.0 using hydrochloric acid solution. More preferably, the concentration of the NaOH solution is 0.1 M and the concentration of the HCl solution is 0.1 M.

[0022] Preferably, in step (2), the ratio of dry powder to water is 1:20 (w / v).

[0023] As a preferred option, in step (2), under alkaline conditions, the extraction conditions are: extraction for 8 hours in a water bath at 70°C.

[0024] Preferably, in step (2), the extraction operation is performed twice in total. The extraction liquid is centrifuged at a speed of 10,000×g for the first time and at a speed of 12,000×g for the second time.

[0025] Preferably, in step (2), the rotary evaporation temperature is 55°C.

[0026] Preferably, in step (3), the centrifugation speed is (14,000~15,000) × g, more preferably 14,000 × g.

[0027] Preferably, in step (3), the conditions for enzymatic hydrolysis with trypsin and alkaline protease are as follows: adjust the pH of the supernatant to 7.5 with hydrochloric acid, add trypsin solution, and hydrolyze at 37°C for 16-20 h. Terminate the reaction by adjusting the pH to 10 with Tris-Base buffer, then add alkaline protease aqueous solution and continue hydrolysis at 37°C for 8-10 h. More preferably, the trypsin solution is prepared using 25 mM ammonium bicarbonate aqueous solution to ensure enzyme stability and good function. More preferably, the trypsin hydrolysis time is 16 h. More preferably, the alkaline protease hydrolysis time is 8 h. The ratio of the supernatant to trypsin and alkaline protease is preferably with a slight excess of trypsin and alkaline protease to meet the requirements of sufficient reaction in the shortest possible time and minimizing costs.

[0028] Preferably, in step (3), the permeation pore size of the ultrafiltration membrane is 10 kDa.

[0029] Preferably, in step (4), the polypeptide sample of the Hericium erinaceus protease hydrolysate is separated using a high-performance liquid chromatography (HPLC) system and analyzed on a Q-Exactive HF-X mass spectrometer. The HPLC system uses a reversed-phase C18 column for separation, with phase A being a 0.1% (v / v) formic acid aqueous solution and phase B being a 0.1% (v / v) formic acid acetonitrile solution. The mobile phase gradient is set as follows: 0-50 minutes: the volume percentage of phase B increases from 4% to 50%; 50-54 minutes: the volume percentage of phase B increases from 50% to 100%; 54-60 minutes: the volume percentage of phase B is maintained at 100%. The MS analysis of the Q-Exactive HF-X mass spectrometer uses positive ion mode, with a precursor ion scanning range of m / z 300-1800. The primary MS resolution is 70,000 at m / z 200, and the secondary MS resolution is 17,500 at m / z 200. MaxQuant is used for the analysis. Version 1.5.5.1 processes the acquired data to obtain the final results for peptide identification.

[0030] Secondly, the present invention also provides bioactive Hericium erinaceus polypeptides or polypeptide compositions.

[0031] The sequence of the Hericium erinaceus polypeptide is selected from one of the following: DVSELLAPE, LLPH, LVLL, VDYP, APEPVGDIGLIGL, VHTLPDLPYAYDALEPYI, IHPTPESVFASHKHLFPPLP;

[0032] The Hericium erinaceus polypeptide composition is obtained by removing the solvent from the Hericium erinaceus protease hydrolysate obtained in step (3) of the method described in the first aspect.

[0033] The Hericium erinaceus polypeptides or polypeptide compositions obtained by this invention are obtained based on the screening method described in the first aspect. They have significant ACE inhibition and antioxidant activity and are suitable for functional foods, nutritional intervention preparations and related application scenarios.

[0034] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0035] This invention uses Hericium erinaceus as a source of functional peptides. Through AI-assisted annotation of potentially functionally active peptides and in vitro verification of their activity, the specific sequences and functions of the peptides are ultimately determined, thereby achieving high-value utilization of Hericium erinaceus peptides. This invention has the following advantages and beneficial effects:

[0036] 1. Precise Identification of Short Peptide Active Sequences: Traditional methods for identifying functional peptides are limited by the structural complexity and lack of functional tags in short-chain peptides (≤10 amino acids), making accurate annotation and screening difficult. Existing AI models are generally only suitable for identifying and classifying medium- to long-chain peptides (>10 amino acids), failing to cover the numerous short-chain functional peptides present in natural protein hydrolysates. This invention, based on the semantic annotation capabilities of the DeepSeek platform and combined with structural feature parameters and protein databases such as BioChemia and InterPro, achieves efficient functional annotation of short-chain peptides. This method successfully identified seven peptides: DVSELLAPE, LLPH, LVLL, and VDYP, with sequence lengths less than 10. This demonstrates that the AI-assisted annotation model can detect and record short-chain peptides, significantly improving the breadth and accuracy of functional factor identification compared to traditional machine recognition models.

[0037] 2. Novel Low-Cost High-Throughput Screening Method: Traditional functional peptide development relies on the individual in vitro activity verification of each isolated and purified peptide. The process includes protein extraction, hydrolysis, separation, screening, and sequence identification, typically taking 3–6 weeks, involving the parallel testing of at least 800 samples, and requiring 4–6 researchers to complete multiple batches of experiments. Furthermore, most peptides exhibit no activity or only weak activity, indicating a high failure rate and significant resource waste in traditional functional peptide discovery. This invention introduces an AI-assisted annotation model to achieve high-throughput screening from the massive peptide sequences obtained by LC-MS / MS, retaining only candidate peptides with high confidence for the verification stage. Compared to traditional processes, this reduces the number of samples tested by approximately 80%, saves 2–4 weeks of development time and more than half of the manpower required, significantly improving the efficiency and cost-effectiveness of peptide activity identification.

[0038] 3. AI-assisted annotation results are consistent with in vitro validation results: This invention used LC-MS / MS technology to resolve over 200 endogenous polypeptide sequences from Hericium erinaceus protein, and employed LLM for functional annotation, ultimately identifying 7 candidate polypeptides with potential activity: DVSELLAPE, LLPH, LVLL, VDYP, APEPVGDIGLIGL, VHTLPDLPYAYDALEPYI, and IHPTPESVFASHKHLFPPLP. Four of these were annotated as ACE inhibitory peptides, and three as antioxidant peptides. In subsequent in vitro functional models, all 7 polypeptides exhibited biological effects similar to the annotation results at their corresponding functional levels, showing similar or identical functional trends in in vitro antioxidant and ACE inhibitory activities. These results demonstrate that the AI-assisted annotation strategy can accurately identify polypeptides with clear functions from a vast number of sequences, providing a reliable method for the precise discovery of functional factors.

[0039] 4. Identification and Validation Pathway Construction of Antihypertensive / Antioxidant Peptides from Hericium erinaceus: This invention uses Hericium erinaceus protein as raw material. Through LC-MS / MS identification combined with functional annotation using the DeepSeek platform, seven active peptides with clear ACE inhibitory or antioxidant functions were discovered for the first time. These peptides have clear origins and structures, and their significant functions were verified by in vitro models. This invention establishes an efficient development pathway from natural proteins to specific functional peptides, possessing direct protective value and application transformation potential. Furthermore, this method has good versatility, applicable not only to the functional screening and annotation of Hericium erinaceus protein peptides and their extended peptides, but also to the identification and product development of functional factors from other natural protein resources (such as fungi, plants, and animals), demonstrating broad application value. (iv) Description of the attached drawings

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0041] Figure 1 This is an overview of the significant differences and innovations between this invention and traditional technical routes in the peptide screening process.

[0042] Figure 2 This is a flowchart of an AI-assisted annotation model built based on machine learning.

[0043] Figure 3 The results of identifying the bioactivity of Hericium erinaceus peptides based on AI-assisted annotation.

[0044] Figure 4 Examples 1-8 and [Example 9] illustrate the scavenging power of L(+)-ascorbic acid against DPPH free radicals.

[0045] Figure 5 Examples 1-8 and [Example 9] illustrate the scavenging power of L(+)-ascorbic acid against ABTS free radicals.

[0046] Figure 6 Examples 1-8 and [Example 10] illustrate the inhibitory effect of captopril on ACE enzyme activity. (V) Detailed Implementation Methods

[0047] The present invention will be further illustrated below with reference to specific embodiments. The present invention is implemented based on the inventive technical solution, and detailed implementation methods and operating steps are provided; however, the scope of protection of the present invention is not limited to the following embodiments. Experimental methods in the following embodiments that do not specify specific conditions are generally performed under conventional conditions.

[0048] Example 1

[0049] Step 1: Place the fruiting bodies of Hericium erinaceus in a 60℃ hot air drying oven and dry for 8 hours. After taking them out, cut them into pieces, grind them thoroughly with a grinder, and then pass them through a 50-mesh sieve to obtain Hericium erinaceus powder for later use.

[0050] Step 2: Weigh an appropriate amount of dry powder and add deionized water at a material-to-liquid ratio of 1:20 (w / v). Adjust the pH to 13 with sodium hydroxide solution (0.1M) and extract at a constant temperature of 70℃ for 8 hours. Centrifuge the resulting extract at 10,000×g for 20 minutes, collect the supernatant, and repeat the extraction under the same conditions once more. Combine the two supernatants. Then adjust the pH of the combined extract to 4.0 with hydrochloric acid solution (0.1M), let it stand for 30 minutes, and then centrifuge at 12,000×g for 20 minutes. Collect the precipitate, which is the crude protein from Hericium erinaceus.

[0051] Step 3: The precipitate was dissolved in 0.1 mol / L sodium hydroxide solution and the pH was adjusted to neutral. The precipitate was then concentrated and enriched by rotary evaporation at 55°C to obtain crude protein extract of Hericium erinaceus. The extract was then refrigerated at 4°C for later use.

[0052] Step 4: Centrifuge the protein extract obtained in Step 3 at 14,000×g for 10 minutes, collect 40 mL of the supernatant, adjust the pH to 7.5 with sodium hydroxide solution (0.1 M), add 1 mL of trypsin solution (5 mg / mL, prepared with 25 mmol / L ammonium bicarbonate solution), and incubate at 37℃ for 16 hours. After the reaction is complete, adjust the pH to 10 with 1 mol / L Tris-Base solution to terminate trypsin activity.

[0053] Step 5: Add 2 mL of alkaline protease aqueous solution (5 mg / mL) to the enzymatic hydrolysate obtained in Step 4, and continue enzymatic hydrolysis at 37℃ for 8 hours. The reaction is then terminated. The obtained enzymatic hydrolysate is then filtered through a Millipore Labscale TFF system using a 10 kDa ultrafiltration membrane to remove high molecular weight impurities, yielding Hericium erinaceus protease hydrolysate (HEph).

[0054] Examples 2-7

[0055] Step 1: The polypeptide sequences in the Hericium erinaceus proteolytic solution (HEph) obtained in Example 1 were identified using liquid chromatography-tandem mass spectrometry (LC-MS / MS). First, the sample was separated using an Easy-nLC high-performance liquid chromatography system (Thermo Fisher Scientific, USA) and then analyzed using a Q-Exactive HF-X mass spectrometer (Thermo Fisher Scientific, USA). During the separation stage, a reversed-phase C18 column (0.15 mm × 150 mm, Column Technology Inc.) was used. Phase A was 0.1% formic acid aqueous solution, and Phase B was 0.1% formic acid acetonitrile solution. The mobile phase gradient was set as follows: 0-50 minutes, Phase B concentration increased from 4% to 50%; 50-54 minutes, Phase B concentration increased from 50% to 100%; 54-60 minutes, Phase B concentration was maintained at 100%. Mass spectrometry analysis was performed in positive ion mode, with the precursor ion scan range set to m / z 300–1800. MS1 scan (primary mass spectrometry) had a resolution of 70,000 at m / z 200, and MS2 scan (secondary mass spectrometry) had a resolution of 17,500 at m / z 200. The raw data were processed using MaxQuant 1.5.5.1 software to obtain qualitative peptide identification results, as shown below. Figure 3 As shown.

[0056] Step 2: As Figure 2As shown, inputting commands into the DeepSeek platform and setting the search keywords to any single item or combination of "mushroom", "peptide", "protein", "fungal" or "bioactive peptide" will invoke the Paper-scrapper module to perform high-throughput literature scraping from the PubMed database and output the search results as PDF documents. Subsequently, inputting commands into the DeepSeek platform will drive the large language model to perform semantic recognition and information extraction on the PDF documents. In Deepseek, commands are input to identify PDF documents, extract peptide information, and generate a JSON array containing the peptide information according to specified fields. This JSON array must include the following fields: peptide sequence, origin, structure, function, and hydrophilic / hydrophobic residues. If any relevant information is missing, "None" is entered. If "mushroom" and "peptide" are not included, the data set is discarded. This establishes a knowledge base containing over 7000 structure-function relationships. The BioChemia and InterPro databases are connected to this knowledge base via an external database access mechanism to achieve comprehensive identification and functional coverage of both short-chain and long-chain peptides. Then, commands are input into Deepseek to flatten the identified data and export an Excel file containing the same fields as the JSON array. The resulting Excel file is then manually proofread, discarding irrelevant data, thus completing the LLM training and closed-loop optimization.The peptide sequences identified in step 1 are used as input information and imported into the AI-assisted annotation model built and trained based on the large language model. A structure-function relationship matching operation is performed, and peptide sequences meeting the following conditions are considered candidate sequences: when the peptide chain length is 3-8 amino acids, if the number of amino acid differences does not exceed 2, it is considered a similar sequence; when the peptide chain length is >8 and ≤25 amino acids, if the sequence similarity (Identity) is ≥50% and the number of amino acid substitutions / deletions does not exceed 4, it is considered a similar sequence. Similar sequences are used as candidate sequences. The retrieved candidate sequences are manually sorted and automatically aligned to obtain accurate sequence-function relationships. Sex pairing; to ensure comprehensive identification of both long and short chain peptides, the built-in analysis functions of the BioChemia (https: / / biochemia.uwm.edu.pl / biopep / peptide_data.php) and InterPro (https: / / www.ebi.ac.uk / interpro / ) databases were used to further confirm the potential biological activities of candidate sequences and to fully annotate their potential biological activities, thereby obtaining 7 peptides with putative biological activities. Their sequences are shown in Table 1. Table 1 shows the AI-assisted annotation and functional classification of peptide biological activities.

[0057] Step 3: Based on the AI-assisted annotation and screening described in Step 2, seven candidate polypeptide sequences with potential antihypertensive or antioxidant activities were selected. These seven potentially active polypeptides were artificially synthesized using the Fmoc solid-phase peptide synthesis method. The synthesis was completed by Nanjing Jietai Biotechnology Co., Ltd. (Nanjing, China). The synthesized polypeptide products were tested for purity by high-performance liquid chromatography (HPLC), confirming that their purity was above 95%. The obtained polypeptides were designated as: [Example 2] DVSELLAPE, [Example 3] LLPH, [Example 4] LVLL, [Example 5] VDYP, [Example 6] APEPVGDIGLIGL, [Example 7] VHTLPDLPYAYDALEPYI, [Example 8] IHPTPESVFASHKHLFPPLP. These were used for subsequent in vitro functional verification of antioxidant and ACE inhibitory activities. [Example 9] L-(+) levozide and [Example 10] captopril were used as positive controls in the subsequent in vitro functional verification of antioxidant and ACE inhibitory activities, respectively.

[0058] The antioxidant / ACE inhibitory activity of the samples obtained in the above examples was verified using the following test methods:

[0059] (1) Determination of DPPH free radical scavenging rate: First, prepare a 0.05 mg / mL DPPH ethanol solution; dry the Hericium erinaceus protease hydrolysate (HEph) obtained according to the method in Example 1 to remove the solvent and obtain HEph. Dissolve the seven potentially active peptides artificially synthesized by the Fmoc solid-phase peptide synthesis method and HEph in a solvent (made by mixing PBS buffer and DMSO at a volume ratio of 9:1) to obtain peptide solutions with concentrations of 0.0625, 0.125, 0.25, 0.5, and 1 mg / mL. Mix 100 μL of peptide solution and 100 μL of HEph solution with 100 μL of DPPH ethanol solution in a 96-well plate, and then measure the absorbance at a wavelength of 517 nm using an ELISA reader to calculate the DPPH free radical scavenging rate. A high scavenging rate indicates that the sample has good antioxidant properties.

[0060] (2) Determination of ABTS free radical scavenging rate: First, prepare 7.4 mmol / L ABTS diammonium salt stock solution and 2.6 mmol / L potassium persulfate stock solution. Mix the two solutions in equal volumes and place them at room temperature in the dark for 12 hours. Use pH7.4 PBS: DMSO (9:1 (v / v)) at a ratio of 1:20 (v:v) to make the detection value reach about 0.7 at 734 nm using an ELISA reader to obtain ABTS working solution. Mix 100 μL of peptide solutions with concentrations of 0.0625, 0.125, 0.25, 0.5, and 1 mg / mL (prepared as above) and 100 μL of HEph solution (prepared as above) with 100 μL of ABTS working solution in a 96-well plate. Then, measure the absorbance at 734 nm using an ELISA reader and calculate the ABTS free radical scavenging rate. If the scavenging rate is high, it indicates that the obtained eluent has good antioxidant properties.

[0061] (3) In vitro inhibition of angiotensin-converting enzyme (ACE) activity: First, prepare N-[3-(2-furyl)propionyl]-phenylalanylglycine (FAPGG) substrate solution: Weigh a certain amount of FAPPGG, dissolve it in 25 mmol / L ammonium bicarbonate buffer, adjust the pH to 7.5, and make up the volume to 0.5 mmol / L; Second, prepare ACE enzyme working solution: Weigh ACE enzyme powder and dissolve it in the same buffer to prepare an ACE enzyme solution with an activity of 0.1 U / mL for later use. Subsequently, in each well of a 96-well microplate, add 100 μL of FAPGG substrate solution, 10 μL of peptide solution (prepared as above) or HEph solution (prepared as above) at concentrations of 0.0625, 0.125, 0.25, 0.5, or 1 mg / mL, and 10 μL of ACE enzyme solution. After thorough mixing, immediately measure the initial absorbance at 340 nm. The mixture containing substrate, ACE enzyme, and sample is designated A1; the mixture containing only substrate and ACE enzyme is designated A2. 空白1 A mixture containing only substrate and sample, without ACE enzyme, is designated as A1. 空白2 After thorough mixing and shaking for 10 seconds, the absorbance at 340 nm (A1) was measured. 测量 A1 空白1 and A1 空白2 Then it was incubated at 37°C for 30 minutes, and the absorbance at 340 nm was measured again (A2). 测量 A2 空白1 and A2 空白2 The change in absorbance is calculated as ΔA. sample =A1 测量 -A2 测量 ΔA blank =(A1 空白1 -A2 空白1 )-(A1 空白2 -A2 空白2 ACE inhibition rate = (ΔA) blank -ΔA sample ) / ΔA blank Calculated as ×100%. A high inhibition rate indicates that the sample has a strong inhibitory ability on angiotensin-converting enzyme.

[0062] Test results are available Figure 4-6 .

[0063] The DPPH assay results are expressed as antioxidant rate (%). The antioxidant activities of the different examples, measured from strongest to weakest, are as follows: [Example 5] > [Example 7] > [Example 8] > [Example 6] > [Example 1] > [Example 3] > [Example 4] > [Example 2]. Among them, [Example 5] showed the strongest scavenging ability against DPPH free radicals, with an IC50 value of [missing value]. 50 The value was 1.610 mg / mL, significantly better than [Example 1] (14.86 mg / mL) and [Example 2] (119.3 mg / mL), and its effect was close to that of the positive control L-(+)-ascorbic acid (IC50). 50 = 8.306×10 -2 (mg / mL). Linear fit R for all samples 2 All values ​​were above 0.83, indicating a good fit.

[0064] The results of the ABTS method are expressed as antioxidant rate (%), and the antioxidant activity measured from strongest to weakest is as follows: [Example 8] > [Example 5] > [Example 1] > [Example 2] > [Example 3] > [Example 4] > [Example 7] > [Example 6]. Among them, the ABTS free radical scavenging IC50 of [Example 8] is... 50 The value is the lowest, at only 8.505 × 10. - 5 mg / mL, superior to the positive control L-(+)-ascorbic acid (5.876×10 mg / mL). -4 (mg / mL), exhibiting extremely strong scavenging ability. Most examples are superior to [Example 1], demonstrating its functionally dominant sequence in the enzymatic hydrolysate. The fitting correlation coefficient R for all samples is... 2 All values ​​ranged from 0.83 to 0.97. The above experimental results validated the reliability of the AI ​​model annotation results and indicated that the in vitro antioxidant activity of HEph [Example 1] was between that of the individual functional peptides, failing to exhibit optimal performance. This suggests that the series of peptides in the enzymatic hydrolysis product may have a certain degree of antagonistic or neutralizing effect. This phenomenon further demonstrates that the key functional peptides obtained through AI model annotation and screening in this invention can effectively represent the main active components in complex enzymatic hydrolysis systems, thus providing a more accurate structural basis for functional evaluation and mechanism attribution. Based on these results, the enzymatic hydrolysate can be directly used in functional foods and nutritional intervention products without further separation and purification, significantly simplifying the preparation process and reducing industrial application costs.

[0065] ACE inhibition results are expressed as ACE inhibition rate (%). The inhibitory ability of different examples on ACE from strongest to weakest is as follows: [Example 4] > [Example 3] > [Example 1] > [Example 5] > [Example 2] > [Example 8] > [Example 6] > [Example 7]. Among them, [Example 4] has the most significant ACE inhibition effect, IC50... 50 The value was 0.9276 mg / mL, significantly better than [Example 1] (1.170 mg / mL) and other peptides. The IC50 of the positive control captopril was [not specified]. 50 It is 6.791×10 -2 The highest activity was observed at mg / mL. In other examples, [Example 3] and [Example 5] also showed good inhibitory activity. The above experimental results verified the reliability of the AI ​​model annotation results, and also showed that the in vitro anti-ACE activity of HEph [Example 1] was between that of each single functional peptide, and did not show the best effect, suggesting that there may be a certain degree of antagonism or neutralization of the series of peptides in the enzymatic hydrolysis product. This phenomenon further shows that the key functional peptides obtained by AI model annotation and screening of the present invention can effectively represent the main active ingredients in complex enzymatic hydrolysis systems, thus providing a more accurate structural basis for functional evaluation and mechanism attribution. Based on this result, the enzymatic hydrolysate can be directly used in functional foods and nutritional intervention products without further separation and purification, significantly simplifying the preparation process and reducing the cost of industrial application. The above experimental results show that the annotated functional peptides and Hericium erinaceus enzymatic hydrolysate have good development potential and are suitable for functional foods, nutritional interventions and related market scenarios, showing good application prospects and industrial transformation value.

[0066] In summary, the results demonstrate that the AI-assisted annotation and screening model constructed in this invention exhibits significant accuracy and practical value in annotating antioxidant and ACE inhibitory functions. All identified functional peptide sequences are novel discoveries with clear structures and well-defined functions. Furthermore, the enzymatic hydrolysate containing these peptides exhibits stable complex activity, possessing practical application advantages and can be protected as a functional raw material or end product to expand its development potential in food, health products, and pharmaceutical preparations. The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Equivalent changes and modifications made according to the scope of the patent and the description of the present invention should still fall within the scope of the present invention.

[0067] Table 1

[0068]

Claims

1. A method for screening functional peptides of Hericium erinaceus based on artificial intelligence assistance, characterized in that: The method includes the following steps: (1) Constructing an AI-assisted annotation model: In the DeepSeek platform, with the LLM large language model as the core, one or a combination of "mushroom" and "peptide" as keywords, Paper-scrapper is called to perform high-throughput literature crawling in the PubMed database and output documents; secondly, inputting instructions in DeepSeek to identify the documents and extract peptide information from them, including peptide sequence, source, structural features, functional annotation, hydrophilic and hydrophobic residues, and establishing a data knowledge base of structure-function relationship; thirdly, the BioChemia database and InterPro database are connected to the data knowledge base through an external database calling mechanism to achieve comprehensive identification and functional annotation of short-chain peptides and long-chain peptides, thereby obtaining an AI-assisted annotation model that has completed training and closed-loop optimization; (2) Extraction of crude protein from Hericium erinaceus: The fruiting bodies of Hericium erinaceus are dried, cut into pieces, crushed, and sieved to obtain dry powder; the dry powder and water are mixed at a material-to-liquid ratio of 1:(20~25) (w / v), the pH is adjusted to 13, and the mixture is extracted for 8~10 hours in a water bath at 50~70℃. The supernatant of the extract is collected by centrifugation, and the residue is extracted again by water extraction according to the above operation, so that the total number of water extraction operations is at least 2. The supernatants are combined, the pH of the supernatant is adjusted to 4.0, and the protein is allowed to stand to completely precipitate. The protein is collected by centrifugation to obtain crude protein; the crude protein is dissolved in NaOH solution and adjusted to neutral, and then concentrated and enriched by rotary evaporation at 55~65℃ to obtain crude protein extract of Hericium erinaceus. (3) Preparation of Hericium erinaceus protein hydrolysate by compound enzymatic hydrolysis: The crude protein extract of Hericium erinaceus obtained in step (2) is centrifuged to obtain the supernatant, and then enzymatically hydrolyzed with trypsin and alkaline protease in sequence. The resulting hydrolysate is filtered through an ultrafiltration membrane with a permeation pore size between 10 and 100 kDa to obtain Hericium erinaceus protein hydrolysate. (4) Peptide identification: The peptides in the Hericium erinaceus protein hydrolysate obtained in step (3) were identified by liquid chromatography-tandem mass spectrometry to obtain peptide sequence results; (5) AI-assisted annotation and peptide screening: The peptide sequence obtained in step (4) is used as input and imported into the AI-assisted annotation model based on the DeepSeek platform constructed in step (1) for sequence-function matching. The sequence and activity knowledge base of the annotation model is used as a reference library. Peptide sequences that meet any of the following conditions are judged as candidate sequences: When the peptide chain length is 3 to 8 amino acids, if the number of amino acid differences between the candidate sequence and any known sequence in the reference library does not exceed 2, it is judged as a similar sequence; When the peptide chain length is greater than 8 and does not exceed 25 amino acids, if the sequence similarity between the candidate sequence and any known sequence in the reference library is not less than 50% and the number of amino acid substitutions or deletions does not exceed 4, it is judged as a similar sequence; The similar sequences are used as candidate sequences to establish the corresponding sequence-activity correspondence; In order to ensure the comprehensive identification of long chain peptides and short chain peptides, the built-in analysis function of the BioChemia and InterPro databases is used to further confirm the potential biological activity of the candidate sequences and to fully annotate their potential biological activity, thereby obtaining peptides with assumed biological activity.

2. The screening method as described in claim 1, characterized in that: The keywords include one or any combination of "mushroom", "peptide", "protein", "fungal", and "bioactive peptide".

3. The screening method as described in claim 1, characterized in that: In step (1), the extracted peptide information is used to generate a JSON array according to the specified fields. The JSON array must contain the following fields: peptide sequence, source, structural features, functional annotation, hydrophilic and hydrophobic residues. If the relevant information is missing, fill in "None". If "mushroom" and "peptide" are not included, the data set is discarded. This data is used to standardize the analysis of literature data and import it into the data knowledge base.

4. The screening method as described in claim 3, characterized in that: In step (1), after connecting the BioChemia database and the InterPro database to the data knowledge base through the external database calling mechanism, input the command in Deepseek to flatten the data based on the identified data and export an Excel file. The fields contained in the Excel file are the same as those in the JSON array. The obtained Excel file is manually proofread and irrelevant data is discarded, thereby completing the training and closed-loop optimization of the AI-assisted annotation model.

5. The screening method as described in claim 1, characterized in that: In step (2), the sieve aperture is between 50 and 80 mesh.

6. The screening method as described in claim 1, characterized in that: In step (2), the ratio of dry powder to water is 1:20 (w / v), and the extraction conditions are: extraction for 8 hours in a water bath at 70℃.

7. The screening method as described in claim 1, characterized in that: In step (3), the conditions for enzymatic hydrolysis with trypsin and alkaline protease in sequence are as follows: adjust the pH of the supernatant to 7.5 with hydrochloric acid, add trypsin solution (preferably prepared with 25 mM ammonium bicarbonate aqueous solution), and carry out enzymatic hydrolysis at 37°C for 16-20 h (preferably 16 h). Adjust the pH to 10 with Tris-Base buffer to terminate the reaction, and then add alkaline protease aqueous solution, and continue enzymatic hydrolysis at 37°C for 8-10 h (preferably 8 h).

8. The screening method as described in claim 1, characterized in that: In step (3), the permeation pore size of the ultrafiltration membrane is 10 kDa.

9. The screening method as described in claim 1, characterized in that: In step (4), the peptide sample of Hericium erinaceus protease hydrolysate was separated using a high-performance liquid chromatography (HPLC) system and analyzed on a Q-Exactive HF-X mass spectrometer. The HPLC system used a reversed-phase C18 column for separation, with phase A being 0.1% formic acid aqueous solution and phase B being 0.1% formic acid acetonitrile solution. The mobile phase gradient was set as follows: 0-50 minutes: the volume percentage of phase B increased from 4% to 50%; 50-54 minutes: the volume percentage of phase B increased from 50% to 100%; 54-60 minutes: the volume percentage of phase B was maintained at 100%. The MS analysis of the Q-Exactive HF-X mass spectrometer was performed in positive ion mode, with the precursor ion scanning range being m / z 300-1800. The primary MS resolution was 70,000 at m / z 200 and the secondary MS resolution was 17,500 at m / z 200. The acquired data was processed using MaxQuant version 1.5.5.1 to obtain the final results of peptide identification.

10. A bioactive Hericium erinaceus polypeptide or polypeptide composition, wherein the sequence of the Hericium erinaceus polypeptide is selected from one of the following: DVSELLAPE, LLPH, LVLL, VDYP, APEPVGDIGLIGL, VHTLPDLPYAYDALEPYI, IHPTPESVFASHKHLFPPLP; The Hericium erinaceus polypeptide composition is obtained by removing the solvent from the Hericium erinaceus protease hydrolysate obtained in step (3) of the screening method according to any one of claims 1-9.

Citation Information

Patent Citations

  • Anti-fatigue polypeptide in yak blood as well as screening method and application thereof

    CN119775357A

Cited By

  • Anti-fatigue polypeptide in yak blood as well as screening method and application thereof

    CN119775357A

  • Yak blood anti-fatigue polypeptide, screening method and application thereof

    CN119775357B