Antimicrobial peptides, antimicrobial peptide compositions, and their applications in the Qinghai-Tibet Plateau habitat, discovered using machine learning and deep learning methods.

By using machine learning and deep learning to mine antimicrobial peptides from metagenomic data of multiple habitats on the Qinghai-Tibet Plateau, the problem of insufficient antimicrobial peptide mining in existing technologies has been solved. Antimicrobial peptides with broad-spectrum antimicrobial activity and high biosafety have been obtained, which are suitable for the preparation of antimicrobial drugs and biopesticides.

CN123106852APending Publication Date: 2026-11-17LANZHOU UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611627839.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Current technologies lack systematic exploration and research on antimicrobial peptides derived from environmental microorganisms in the Qinghai-Tibet Plateau, and traditional antibiotics face serious threats from drug-resistant bacterial infections and agricultural diseases, making the discovery of novel antimicrobial peptides urgently needed.

Method used

Using machine learning and deep learning methods, potential antimicrobial peptides were mined from metagenomic data of multiple habitats on the Qinghai-Tibet Plateau. Candidate antimicrobial peptide sequences were obtained through a multi-level screening process, and their antimicrobial activity and biosafety were verified by solid-phase chemical synthesis.

Benefits of technology

We have obtained antimicrobial peptides with unique origin, broad-spectrum antimicrobial activity and high biosafety, which are suitable for the preparation of antimicrobial drugs and biopesticides. They can effectively inhibit human opportunistic pathogens and plant pathogenic microorganisms and have industrialization potential.

✦ 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

This invention discloses antimicrobial peptides, antimicrobial peptide compositions, and their applications mined using machine learning and deep learning methods in the Qinghai-Tibet Plateau habitat. By integrating multiple machine learning and deep learning models, a systematic screening of microbial-encoded peptide sets from a multi-habitat metagenomic database of the Qinghai-Tibet Plateau was conducted, yielding 10 high-confidence candidate antimicrobial peptides. Physicochemical property analysis showed that the molecular weights of the candidate antimicrobial peptides ranged from 2718 to 5003 Da, with isoelectric points all above 11, and their sequences exhibited significant novelty compared to known databases. In vitro MIC experiments demonstrated that the candidate antimicrobial peptides exhibited significant broad-spectrum antimicrobial activity against various opportunistic pathogens and plant pathogens. Hemolytic activity assessment showed that all peptides had hemolysis rates below 5% at effective inhibitory concentrations (≤32 μg / mL), with AMP9 and AMP17 showing 0% hemolysis rates at all tested concentrations, demonstrating excellent biosafety. The peptides themselves and their compositions can be used for the development of novel antimicrobial drugs and biopesticides.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of bioinformatics, microbiology and biomedicine, and in particular to antimicrobial peptides, antimicrobial peptide compositions and their applications in the Qinghai-Tibet Plateau environment, which are mined based on machine learning and deep learning methods. Background Technology

[0002] Bacterial and fungal infections pose a significant threat to global public health and agricultural production. In the healthcare field, opportunistic pathogens such as Staphylococcus aureus (Staphylococcus) are also a major concern. Staphylococcus aureus ), Escherichia coli ( Escherichia coli ), Pseudomonas aeruginosa ( Pseudomonas aeruginosa Infections caused by plant pathogens such as *Pseudomonas syringae*, especially the prevalence of drug-resistant strains, have become a significant global public health crisis. In agricultural production, plant pathogens such as *Pseudomonas syringae* (…) are particularly prevalent. Pseudomonas syringae Fusarium oxysporum ( Fusarium oxysporum ) and Fusarium pseudograss ( Fusarium pseudograminearum This causes significant agricultural economic losses every year and seriously threatens food security.

[0003] Antimicrobial peptides (AMPs) are an important component of the organism's innate immune system. They typically consist of 20-100 amino acid residues and possess properties such as positive charge, amphiphilicity, and broad-spectrum antibacterial activity. Compared to traditional antibiotics, antimicrobial peptides have advantages such as a unique mechanism of action (primarily through disruption of bacterial cell membranes), low susceptibility to inducing drug resistance, and good biodegradability, making them considered next-generation candidate molecules for combating drug-resistant bacterial infections and agricultural diseases.

[0004] The Qinghai-Tibet Plateau, as the highest and largest plateau region on Earth, possesses a unique extreme environment (high ultraviolet radiation, low temperature, low oxygen, high salinity, etc.), which endows its microorganisms with special survival strategies and metabolic diversity. Studies have shown that bioactive molecules produced by microorganisms in extreme environments often have unique structural and functional properties. However, systematic research on the discovery of antimicrobial peptides derived from environmental microorganisms on the Qinghai-Tibet Plateau remains very limited.

[0005] With the rapid development of high-throughput sequencing technology and computational biology, metagenomic data is growing exponentially, providing unprecedented opportunities for large-scale mining of functional bioactive molecules. Machine learning and deep learning methods have been successfully applied to the prediction and discovery of antimicrobial peptides, but there is currently a lack of systematic antimicrobial peptide mining and research targeting the metagenomics of microorganisms in the unique habitat of the Qinghai-Tibet Plateau. Summary of the Invention

[0006] Based on the above, the purpose of this invention is to provide antimicrobial peptides from the Qinghai-Tibet Plateau habitat mined using machine learning and deep learning methods. Through large-scale mining methods based on deep learning, potential antimicrobial peptides in the metagenomics of microorganisms in various habitats of the Qinghai-Tibet Plateau are systematically screened, and their broad-spectrum antimicrobial activity and biosafety are verified through in vitro experiments.

[0007] Another object of the present invention is to provide an antimicrobial composition comprising the above-mentioned antimicrobial peptides.

[0008] Another object of the present invention is to provide the use of the above-mentioned antimicrobial peptides or antimicrobial compositions in the preparation of antimicrobial drugs or biopesticides.

[0009] To achieve its purpose, the present invention adopts the following technical solution: The first aspect of the technical solution of the present invention provides antimicrobial peptides of the Qinghai-Tibet Plateau habitat mined based on machine learning and deep learning methods, the amino acid sequences of which are shown in SEQ ID NO.1-10 respectively.

[0010] Preferably, the amino acid sequences of the antimicrobial peptides are as shown in SEQ ID NO.3, SEQ ID NO.4, SEQ ID NO.5, and SEQ ID NO.7, respectively (which have broad-spectrum antimicrobial effects).

[0011] A second aspect of the present invention provides an antibacterial composition comprising antimicrobial peptides from the Qinghai-Tibet Plateau environment mined using machine learning and deep learning methods.

[0012] Preferably, the antimicrobial composition further comprises a pharmaceutically or agriculturally acceptable carrier, adjuvant, diluent, or excipient.

[0013] Preferably, the antibacterial composition includes, but is not limited to, antibacterial drugs, disinfectants, preservatives, feed additives, or daily chemical detergents.

[0014] The third aspect of the technical solution of the present invention provides the application of antimicrobial peptides of the Qinghai-Tibet Plateau habitat mined based on machine learning and deep learning methods in the preparation of antimicrobial compositions, wherein the antimicrobial compositions inhibit human opportunistic pathogens and / or anti-plant pathogenic bacteria and / or anti-plant pathogenic fungi. The opportunistic pathogens in humans are one or more of Staphylococcus aureus, Staphylococcus epidermidis, Escherichia coli, or Pseudomonas aeruginosa. The plant pathogenic bacteria is *Pseudomonas syringae*. The plant pathogenic fungus is one or a mixture of two of Fusarium oxysporum or Fusarium pseudograss.

[0015] The beneficial effects of this invention are as follows: 1. The antimicrobial peptides have unique sources and rich biodiversity: The antimicrobial peptides are derived from microorganisms in various habitats such as lakes, rivers, wetlands, and soils on the Qinghai-Tibet Plateau. The host microorganisms span multiple bacterial phyla and include a variety of novel or uncultured species, representing a unique treasure trove of natural resources.

[0016] 2. Outstanding broad-spectrum antibacterial activity: The antimicrobial peptides exhibit significant inhibitory activity against both opportunistic pathogenic bacteria (Staphylococcus aureus, Staphylococcus epidermidis, Escherichia coli, Pseudomonas aeruginosa) and plant pathogenic microorganisms (Pseudomonas syringae, Fusarium oxysporum, Fusarium graminearum), demonstrating dual application potential. The inhibitory effects of multiple candidate peptides on plant pathogenic fungi are far superior to the reference peptide LL-37.

[0017] 3. Excellent biocompatibility: Multiple candidate antimicrobial peptides (AMP9 and AMP17 have a hemolysis rate of 0% across the entire concentration range, while AMP6, AMP14, and AMP26 all have a hemolysis rate of less than 2%) exhibit extremely low toxicity to eukaryotic erythrocytes at effective inhibitory concentrations, far lower than the known functional peptide LL-37 (which has a hemolysis rate of 10% at 16 μg / mL), demonstrating excellent biocompatibility.

[0018] 4. Simple structure and easy to synthesize artificially: The candidate peptides have small molecular weights (2718-5003 Da) and moderate amino acid sequence lengths (21-42 residues), making them easy to produce on a large scale through solid-phase chemical synthesis methods, and have good industrialization prospects.

[0019] 5. Comprehensive and systematic discovery methods: It integrates multiple mainstream machine learning and deep learning models to form a multi-level candidate antimicrobial peptide screening process from coarse screening to fine screening, which effectively reduces the false positive rate and improves the discovery efficiency. Attached Figure Description

[0020] Figure 1 The mass spectrum of AMP3; Figure 2 This is the mass spectrum of AMP6; Figure 3 This is the mass spectrum of AMP9; Figure 4 The mass spectrum of AMP13; Figure 5 The mass spectrum of AMP14; Figure 6 The mass spectrum of AMP17; Figure 7 The mass spectrum of AMP20; Figure 8 The mass spectrum of AMP25; Figure 9 This is the mass spectrum of AMP26; Figure 10 The mass spectrum of AMP402; Figure 11 The liquid chromatogram of AMP3; Figure 12 The liquid chromatogram of AMP6; Figure 13 The liquid chromatogram of AMP9; Figure 14 The liquid chromatogram of AMP13; Figure 15 The liquid chromatogram of AMP14; Figure 16 The liquid chromatogram of AMP17; Figure 17 The liquid chromatogram of AMP20; Figure 18 The liquid chromatogram of AMP25; Figure 19 The liquid chromatogram of AMP26; Figure 20 This is the liquid chromatogram of AMP402. Detailed Implementation

[0021] The following detailed embodiments illustrate the antimicrobial peptide discovery method, physicochemical properties of candidate antimicrobial peptides, artificial synthesis method, antimicrobial activity, and biosafety of the present invention.

[0022] Example 1: Large-scale mining of candidate antimicrobial peptides based on deep learning This embodiment uses a variety of machine learning and deep learning models to screen candidate antimicrobial peptides from metagenomic data of multiple habitats on the Qinghai-Tibet Plateau, obtaining 10 high-confidence candidate antimicrobial peptide sequences. The specific steps are as follows: (1) Data acquisition: Download the Tibetan Plateau multi-habitat microbial metagenomic dataset (numbers PRJNA305188, PRJNA449471, PRJNA520979, PRJNA521031, PRJNA521032, PRJNA521034, PRJNA554750, PRJNA589661, PRJNA714430, PRJNA752512, PRJNA776043, PRJNA823845, PRJNA835221, PRJNA836635, PRJNA837727, PRJCA017393) from the NCBI SRA database (https: / / www.ncbi.nlm.nih.gov / sra / ) and the National Center for Biotechnology Information (https: / / ngdc.cncb.ac.cn / gsa / ). This dataset contains 798 samples from various habitats on the Qinghai-Tibet Plateau, including lakes, rivers, wetlands, soils, and glaciers, covering a rich array of microbial diversity resources.

[0023] (2) Obtaining medium-to-high quality genomes and species annotation: The trimgalore software was used to quality control the raw metagenomic reads to remove low-quality sequences; then, the MEGAHIT software was used with default parameters to assemble the quality-controlled high-quality reads, retaining contig sequences longer than 500 bp; subsequently, MetaBAT2, MaxBin2, VAMB, and SemiBin2 were used to bin the metagenomic assembly results of each sample; CheckM2 was used to evaluate the genome integrity and contamination of the binning results, and medium-to-high quality genomes with integrity ≥50% and contamination <10% that metagenomically meet the minimum information standard for genome assembly (MIMAG) were retained for subsequent analysis. The GTDB-Tk genome classification database toolkit was used with reference to the GTDB R220 version database (https: / / gtdb.ecogenomic.org / stats / r220) to perform species taxonomic annotation on the above genomes.

[0024] (3) Protein coding sequence prediction: Gene prediction was performed on the above genome using Prodigal software. All predicted protein coding sequences were extracted, and peptides with an amino acid length of 20-100 AA were screened to construct a candidate peptide sequence pool. (4) Initial screening using machine learning: The amPEPpy software based on machine learning was used to predict the antimicrobial peptide activity of candidate peptides, and sequences with an antimicrobial peptide prediction probability greater than or equal to 0.5 were retained as the initial screening candidate set.

[0025] (5) Deep learning re-screening: The initial candidate set was further predicted using deep learning software AMPlify and AMPScanner v2. The prediction results of the two models were integrated to obtain high-confidence candidate antimicrobial peptide sequences as high-confidence candidate sets.

[0026] (6) Comprehensive evaluation: The high-confidence candidate set was comprehensively evaluated using AMPProject (AMPSorter module) based on the protein big language model. Sequences with a predicted probability greater than or equal to 0.5 were retained as the final candidate antimicrobial peptides (cAMP), and a total of 10 candidate antimicrobial peptides were obtained (Table 1).

[0027] (7) Comparison with known databases: The candidate antimicrobial peptide sequences were compared with known experimentally validated antimicrobial peptide databases (APD 3, DBAASP 4, and DRAMP 4) and computationally predicted peptide databases (AMPSphere) using DIAMOND software (blastp subroutine, default threshold E-value 0.001) to confirm the novelty of the sequences. The results showed that no similar sequences were found in any of the candidate antimicrobial peptides in the known databases (Table 3), indicating that the candidate antimicrobial peptides mined in this invention have high novelty.

[0028] Table 1 shows the antimicrobial probabilities of the 10 candidate antimicrobial peptides predicted by combining machine learning and deep learning. The results in Table 1 show that the average predicted probability of the 10 candidate antimicrobial peptides is above 0.73, and the average predicted probability of several peptides, such as AMP13, AMP17, AMP20, AMP25, and AMP402, reaches above 0.90, indicating extremely high confidence in the prediction of antimicrobial activity.

[0029] Example 2: Sources and Host Species Annotations of Candidate Antimicrobial Peptides All 10 candidate antimicrobial peptides were derived from the metagenomics of microorganisms in the diverse habitats of the Qinghai-Tibet Plateau. Specific ecological sources and host microorganism annotations are shown in Table 2.

[0030] Table 2. Notes on the sources and host species of 10 candidate antimicrobial peptides Table 2 shows that the 10 candidate antimicrobial peptides originated from four typical extreme habitats of the Tibetan Plateau: lakes (5), rivers (3), wetlands (1), and soil (1). The host microorganisms belonged to four phyla: Bacteroidota (4), Pseudomonadota (4), Verrucomicrobiota (1), and Nitrospirota (1). Many of the genus-level annotations belonged to uncultured or novel taxa (such as new genera like PNNG01, JACTMZ01, and JACJXF01), indicating that these antimicrobial peptides originated from the unique biodiversity resources of the Tibetan Plateau.

[0031] Example 3: Physicochemical property analysis of candidate antimicrobial peptides and comparison with known functional peptide LL-37 To thoroughly evaluate the biochemical properties of the 10 candidate antimicrobial peptides, the online tool ProtParam (https: / / web.expasy.org / protparam / ) was used to predict the molecular weight, amino acid length, isoelectric point (pI), instability index (II), and gross hydrophilicity (GRAVY) of all candidate peptides. Using the known broad-spectrum antimicrobial peptide LL-37 (LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES, 37 amino acids) as a reference, LL-37, one of the earliest naturally occurring antimicrobial peptides discovered in humans, exhibits antibacterial activity against various pathogens and serves as a landmark reference molecule in the field of antimicrobial peptides. The physicochemical properties of the candidate antimicrobial peptides and their comparison with LL-37 are shown in Table 3.

[0032] Table 3. Physicochemical properties of 10 candidate peptides and reference peptide LL-37, and their similarity to known databases. Note: NA indicates that the peptide has no significant similarity to known antimicrobial peptide databases. DIAMOND's default threshold E-value is 0.001.

[0033] The data analysis in Table 3 shows that: (1) Molecular weight: The molecular weight of the 10 candidate antimicrobial peptides ranges from 2718 to 5003 Da, which is in the same order of magnitude as the reference LL-37 (4493 Da). The molecular weight is moderate, which is conducive to cell membrane penetration and exerting antibacterial effect, and is easy to synthesize in solid phase.

[0034] (2) Isoelectric point: The isoelectric points (pI) of all candidate peptides were between 11.10 and 13.35, which were higher than those of the reference LL-37 (pI 10.61), showing a high positive charge characteristic. The high positive charge helps the candidate peptides interact with the negatively charged bacterial cell membrane or fungal cell wall, which may enhance the antibacterial efficacy.

[0035] (3) Gravy value: Most candidate peptides have negative Gravy values ​​(hydrophilicity), are easily soluble in water, and are convenient for preparing aqueous pesticides or drugs.

[0036] (4) Sequence novelty: No similar sequences were found in the known antimicrobial peptide database for any of the candidate peptides, indicating that the candidate antimicrobial peptide sequences discovered in this invention are highly novel.

[0037] Example 4: Artificial Synthesis of Candidate Antimicrobial Peptides In this embodiment, all candidate antimicrobial peptides and the reference LL-37 were artificially synthesized by Sangon Biotech (Shanghai) Co., Ltd. using the solid-phase chemical synthesis (SPPS) method. An Fmoc / tBu protection strategy was employed, and the peptide products were purified by liquid chromatography to obtain a purity >95%. The specific synthesis steps are as follows: (1) Resin activation and coupling of the first amino acid: Wang resin was selected as the solid support. 1 eq of the first C-terminal protected amino acid and 1.5 eq of DIEA were taken and reacted with the resin for 2 h with 10 mL of DMF as solvent. The resin was dried and washed 3 times with DMF. Methanol and DIEA were added to end the resin for 1 h to ensure that the resin was fully activated.

[0038] (2) Fmoc deprotection: Remove the Fmoc protecting group with 20% piperidine / DMF solution, treat for 10 min / time, repeat 2 times; thoroughly clean the resin with DMF to ensure complete deprotection.

[0039] (3) Stepwise coupling of amino acids: 3 eq Fmoc protected amino acids, 3 eq HOBT (1-hydroxybenzotriazole) and 3 eq DIC (N,N-diisopropylcarbodiimide) were added sequentially, and the reaction was carried out at room temperature for 1.5 h with DMF as solvent. After washing the resin, the deprotection-coupling cycle was repeated until the assembly of all amino acid sequences was completed.

[0040] (4) Side chain deprotection and peptide-resin cleavage: After washing and drying the resin, treat it with a cleavage reagent containing 95% TFA + 2% Tis + 2% EDT + 1% H2O for 2 h to cut the peptide chain and remove the side chain protecting group.

[0041] (5) Crude product collection and purification: Filter to remove resin, wash the filtrate repeatedly with ice-cold ether to precipitate crude peptides, and collect by centrifugation; purify by reversed-phase high-performance liquid chromatography (RP-HPLC) and collect the target peak components; freeze-dry to obtain a white loose powder of pure antimicrobial peptide.

[0042] (6) Quality control: Electrospray ionization mass spectrometry (ESI-MS) analysis verified that the molecular weight was consistent with the theoretical value. Figures 1-10 ); RP-HPLC purity >95% ( Figures 11-20 It meets the requirements for experimental and industrial applications.

[0043] Example 5: Detection of antimicrobial activity of candidate antimicrobial peptides (MIC determination) In this embodiment, the minimum inhibitory concentration (MIC) of each candidate antimicrobial peptide against seven pathogens was determined using the micro-broth dilution method, strictly following the guidelines of the European Committee for Testing Antimicrobial Susceptibility (EUCAST), to evaluate its broad-spectrum antimicrobial activity, with the known functional peptide LL-37 as a reference.

[0044] 1. Experimental bacterial strains: Opportunistic pathogens include Staphylococcus epidermidis (Staphylococcus epidermidis) Staphylococcus epidermidis Staphylococcus aureus ( Staphylococcus aureus ), Escherichia coli ( Escherichia coli ), Pseudomonas aeruginosa ( Pseudomonas aeruginosa The plant pathogenic bacteria is *Pseudomonas syringae* (…). Pseudomonas syringae It can cause various plant diseases such as bacterial spot; the plant pathogenic fungus is Fusarium oxysporum (…). Fusarium oxysporum ) and Fusarium pseudograss ( Fusarium pseudograminearum ).

[0045] 2. MIC determination method: The candidate antimicrobial peptides synthesized in Example 4 and the reference LL-37 were dissolved in sterile ultrapure water to prepare a 512 μg / mL stock solution. A two-fold serial dilution method was used to establish a concentration gradient (1-256 μg / mL). The experiment included a blank control group, a peptide background control group, a positive control group (LL-37), and a peptide experimental group, with each group having three independent biological replicates. MIC was defined as the lowest peptide concentration at which no visible bacterial growth / fungal spore germination was significantly inhibited.

[0046] 3. Experimental results: The MIC results of 10 candidate antimicrobial peptides and LL-37 against 7 pathogens are shown in Table 4.

[0047] Table 4. MIC (μg / mL) of 10 candidate peptides against pathogenic bacteria and fungi. Note: MIC>256 means that there is no significant antibacterial effect at the maximum concentration of the experimental peptide of 256 μg / mL.

[0048] The results analysis in Table 4 show that the above peptides have the following characteristics: (1) Broad-spectrum antibacterial activity: All candidate antimicrobial peptides showed antibacterial activity against at least 5 of the 7 pathogens. Among them, AMP9, AMP13, AMP14 and AMP20 peptides could inhibit all 7 pathogens, showing broad-spectrum antibacterial properties.

[0049] (2) Peptide with the best overall antibacterial effect: AMP9 ranked first with an average MIC of 19.43 μg / mL, and its antibacterial effect against Staphylococcus aureus ( Staphylococcus aureus ATCC 25923), Staphylococcus epidermidis ( Staphylococcus epidermidis ATCC 12228), Fusarium graminearum ( Fusarium pseudograminearum The MICs of zbs1 were all 8 μg / mL, showing excellent performance against both opportunistic and plant pathogens. AMP13 (average 27.43 μg / mL), AMP14 (average 27.43 μg / mL), and AMP20 (average 29.14 μg / mL) also exhibited outstanding overall performance.

[0050] (3) Outstanding inhibitory effect on plant pathogenic fungi: AMP3, AMP9, AMP14, AMP17, AMP20, and AMP25 had MICs as low as 4-8 μg / mL against Fusarium oxysporum, and AMP20 and AMP25 had inhibitory effects against Fusarium oxysporum (… Fusarium oxysporum The MIC of ATCC48112 is only 4 μg / mL, which is far superior to LL-37 (MIC >256 μg / mL), fully demonstrating the potential of antimicrobial peptides derived from microorganisms on the Qinghai-Tibet Plateau in the prevention and control of plant fungal diseases.

[0051] (4) Inhibition of plant pathogenic bacteria: AMP13 and AMP17 inhibit *Pseudomonas syringae* (… Pseudomonas aeruginosa The MIC of ATCC 29853 was 8-16 μg / mL, which was better than that of LL-37 (32 μg / mL), showing better inhibitory activity against Pseudomonas syringae.

[0052] (5) Compared with LL-37: LL-37 has a very poor inhibitory effect on plant pathogenic fungi (MIC>256 μg / mL), while the multiple candidate antimicrobial peptides discovered in this invention show significantly better broad-spectrum antimicrobial activity under the same conditions, and have unique advantages in the control of plant pathogens.

[0053] Example 6: Evaluation of the hemolytic activity of candidate antimicrobial peptides In this embodiment, the hemolytic activity of 10 candidate antimicrobial peptides with the best antibacterial effect and the reference peptide LL-37 was evaluated to determine their biosafety to mammalian cells.

[0054] 1. Preparation of red blood cell suspension: Use 2 mL of fresh defibrinated sheep blood collected within 24 h, centrifuge at 800×g for 10 min to remove serum proteins; add sterile PBS buffer (pH 7.4) to resuspend the cells, repeat centrifugation and washing twice until the supernatant is clear; finally, dilute with PBS to prepare a 2% red blood cell suspension and store at 4°C for later use.

[0055] 2. Sample preparation: Using sterile PBS as solvent, solutions of each candidate antimicrobial peptide with concentration gradients of 8, 16, 32, 64, and 128 μg / mL were prepared; the negative control group was an equal volume of PBS, and the positive control group was a 2% Triton X-100 solution.

[0056] 3. Incubation and detection: Mix 100 μL of red blood cell suspension with 100 μL of antimicrobial peptide solution (or control solution) of various concentrations, and incubate at 37°C for 1 h; after centrifugation at 800×g for 10 min, transfer 100 μL of the supernatant to a 96-well plate, and detect the absorbance value at 540 nm using a microplate reader. Hemolysis rate (%) = (A_sample - A_PBS) / (A_TritonX-100 - A_PBS) × 100%. All experiments were repeated 3 times, and the average value of the results was taken, as shown in Table 5.

[0057] Table 5. Hemolysis rate (%) of sheep blood cells by 10 antimicrobial peptides and LL-37 Table 5 Results Analysis: (1) AMP9 and AMP17 have excellent biocompatibility: the hemolysis rate is 0% in the full concentration range of 8-128 μg / mL, showing no hemolytic activity, which is in stark contrast to the reference LL-37 (the hemolysis rate reached 10.04% at 16 μg / mL).

[0058] (2) AMP14, AMP6, AMP26 and AMP402 have excellent biosafety: the hemolysis rate is less than 2% at the highest test concentration of 128 μg / mL, and the hemolysis rate of AMP14, AMP26 and AMP402 is less than 1% at concentrations greater than or equal to 32 μg / mL. They have no obvious toxicity to red blood cells throughout the entire test concentration range.

[0059] (3) At the conventional effective antibacterial concentration (≤32 μg / mL): the hemolysis rate of all 10 peptides was less than 5%; at concentrations below 16 μg / mL, the hemolysis rate of all peptides was less than 2%. The above candidate peptides showed significantly lower toxicity to eukaryotic erythrocytes than LL-37 at effective bactericidal concentrations, demonstrating important safety advantages for clinical and agricultural applications.

[0060] (4) AMP13 has high hemolytic activity at high concentrations: AMP13 has a hemolytic rate of 48.12% at 128 μg / mL, which is similar to that of LL-37 (45.31%); however, the hemolytic rate is only 1.79% at 16 μg / mL. The concentration should be controlled in practical applications.

Claims

1. Antimicrobial peptides from the Qinghai-Tibet Plateau environment, mined using machine learning and deep learning methods, characterized in that... The amino acid sequences of the antimicrobial peptides are shown in SEQ ID NO.1-10, respectively.

2. The antimicrobial peptides of the Qinghai-Tibet Plateau habitat mined using machine learning and deep learning methods as described in claim 1, characterized in that, The amino acid sequences of the antimicrobial peptides are shown in SEQ ID NO.3, SEQ ID NO.4, SEQ ID NO.5, and SEQ ID NO.7, respectively.

3. An antibacterial composition comprising antimicrobial peptides from the Qinghai-Tibet Plateau environment as described in claim 1 or 2, which are mined using machine learning and deep learning methods.

4. The antibacterial composition comprising antimicrobial peptides from the Qinghai-Tibet Plateau environment mined using machine learning and deep learning methods as described in claim 3, characterized in that, The antimicrobial composition also includes pharmaceutically or agriculturally acceptable carriers, adjuvants, diluents, or excipients.

5. The antibacterial composition comprising antimicrobial peptides from the Qinghai-Tibet Plateau environment mined using machine learning and deep learning methods as described in claim 4, characterized in that, The antibacterial composition includes, but is not limited to, antibacterial drugs, disinfectants, preservatives, feed additives, or daily chemical detergents.

6. The application of the antimicrobial peptides from the Qinghai-Tibet Plateau environment mined using machine learning and deep learning methods as described in claim 1 in the preparation of antimicrobial compositions, characterized in that, The antimicrobial composition inhibits human opportunistic pathogens and / or plant pathogenic bacteria and / or plant pathogenic fungi; The opportunistic pathogens in humans are one or more of Staphylococcus aureus, Staphylococcus epidermidis, Escherichia coli, or Pseudomonas aeruginosa. The plant pathogenic bacteria is *Pseudomonas syringae*. The plant pathogenic fungus is one or a mixture of two of Fusarium oxysporum or Fusarium pseudograss.