Antifungal peptide, screening method therefor and use thereof
By using machine learning to screen antifungal peptides with amino acid chain lengths of 17-50 from the gut microbiome, the problems of drug resistance and low development efficiency of existing antifungal drugs have been solved, achieving effective inhibition of a variety of fungi and improving the development efficiency of antifungal drugs.
Patent Information
- Application Number
- PCT/CN2026/109887
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-14
- Filing Date
- 2026-07-10
- Publication Date
- 2026-11-19
AI Technical Summary
Existing antifungal drugs face the problem of drug resistance. Traditional antibiotics have problems such as high toxicity and high drug resistance when treating fungal infections. Moreover, the development efficiency of existing antimicrobial peptide drugs is low, making it difficult to effectively deal with the increasingly serious threat of fungal infections.
Using machine learning methods, 36 antifungal peptides with amino acid chain lengths of 17-50 were screened from the gut microbiome of colorectal cancer patients. Their antifungal effects against various pathogenic fungi were verified by in vitro antibacterial activity experiments, providing peptides for the preparation of antifungal drugs.
The screened antifungal peptides can significantly inhibit Candida auris, Candida tropicalis, Candida albicans, Cryptococcus neoformans, and Candida glabrata at extremely low concentrations, improving the efficiency of antifungal drug development and reducing discovery costs.
Smart Images

Figure CN2026109887_19112026_PF_FP_ABST
Abstract
Description
Antifungal peptides, their screening methods and applications Technical Field
[0001] This invention belongs to the field of antibacterial drug technology, specifically relating to an antifungal peptide, its screening method, and its application. Background Technology
[0002] Infectious diseases are a leading cause of death and disability worldwide. The World Health Organization estimates that drug-resistant bacterial infections directly cause 1.27 million deaths annually and indirectly cause approximately 4.95 million more, imposing a tremendous social and economic burden in resource-constrained environments. Currently, antibiotics are the primary treatment for infectious diseases; however, the overuse of antibiotics in recent years has led to the emergence of numerous multidrug-resistant organisms (MDROs), which can spread rapidly and cause increasingly difficult-to-treat infectious diseases, posing a serious threat to global public health security.
[0003] Compared to bacterial infections, fungal infections have always been a clinical challenge. Deep fungal and invasive fungal infections have high ICU mortality rates, with approximately 13.5 million people worldwide suffering from life-threatening Candida and mold infections each year, and the number of deaths increasing annually. Superficial fungal infections, on the other hand, tend to recur and are generally difficult to cure with medication.
[0004] Currently, among all antifungal drugs, only polyenes, flucytosine, azoles, and echinocandins are widely used in the treatment of fungal infections. However, all of these antifungal drugs have significant drawbacks: for example, amphotericin B has high nephrotoxicity; the main disadvantage of pyrimidine antifungals is the rapid development of resistance, with high toxicity including hepatotoxicity and hematologic toxicity; azoles have a high rate of resistance; and echinocandins are gradually losing their sensitivity to certain fungi. The problem of resistance to traditional antifungal drugs is becoming increasingly serious, and there is a lack of available preventative vaccines for fungal infections, posing a significant challenge to the future prevention and treatment of fungal infections.
[0005] Fungal infections pose a significant challenge to humans and global ecosystems. Developing novel antibiotics, probiotics, prebiotics, immunomodulators, and peptides are effective strategies for combating fungal infections. Some of these, including novel compounds synthesized through chemical synthesis, new structures targeting known targets, and modified versions of existing drugs, are currently undergoing preclinical and clinical trials as antifungal drug candidates. However, these novel antifungal drugs still face limitations in target discovery, clinical development, and safety assessment, requiring further in-depth research and clinical validation. Therefore, the continuous development of novel antifungal drugs to address the increasingly serious threat of fungal infections is crucial for global public health.
[0006] Short peptides have attracted widespread attention due to their high diversity, convenient synthesis, biodegradability, low toxicity, broad-spectrum antibacterial activity, and low likelihood of inducing drug resistance. In particular, a large number of antimicrobial peptides (AMPs) derived from bacteria can be used to treat bacterial, fungal, and viral infections, and even cancer, showing promise as a replacement for traditional antibiotics. However, the development of existing antimicrobial peptide drugs has certain limitations, such as long development time and low efficiency.
[0007] The development of artificial intelligence (AI) technology has accelerated drug discovery to some extent. With the advancement of computer technology and gene sequencing technology, machine learning is playing an increasingly important role in drug discovery. The gut microbiome possesses a high abundance and diversity of genes, making it one of the largest reservoirs of antibiotic resistance genes. Studies have shown extensive cross- and intra-domain interactions among the four kingdoms of the human gut microbiome, and the association between fungi and bacteria is enhanced in colorectal cancer patients. Therefore, in-depth analysis of the gut microbiome using AI technology holds promise for discovering potential antimicrobial drugs. Summary of the Invention
[0008] Therefore, this invention utilizes machine learning methods to identify antifungal peptides (AFPs) from the gut microbiome of colorectal cancer patients, and then detects and verifies the antifungal activity of these peptides through in vitro antibacterial activity experiments. Based on this method, this invention screened 36 antifungal peptides effective against Candida species. These antifungal peptides have amino acid chain lengths concentrated between 17 and 50, small molecular weights, are easy to synthesize, and exhibit high antifungal activity, achieving significant antibacterial effects even at extremely low concentrations of 2 μg / ml. They show significant inhibitory effects against common clinical pathogenic fungi, such as Candida auris, Candida tropicalis, Candida albicans, Cryptococcus neoformans (H99), and Candida glabrata. The antifungal peptides provided by this invention can be used to prepare antifungal drugs or other antibacterial products to exert antifungal infection effects.
[0009] This invention provides the following technical solution:
[0010] In a first aspect, the present invention provides an antifungal peptide, characterized in that the antifungal peptide is selected from at least one of the polypeptides shown in SEQ ID NO.1-SEQ ID NO.36:
[0011] Table 1. Amino acid sequence information of antifungal peptides
[0012] The fungus is selected from at least one of Candida auris, Candida tropicalis, Candida albicans, Cryptococcus neoformans (H99), and Candida glabrata.
[0013] In some embodiments of the present invention, the fungus is Candida auris, and the antifungal peptide against Candida auris has at least one of the following structures:
[0014] a1)X aa1 MKX aa2 RPSVKX aa3 X aa4 CX aa5 KCX aa6 X aa7 IX aa8 RX aa9 GX aa10 VMVICENPKHKQX aa11 QGX aa12 ;
[0015] a2)MRVKITLACTECKQRNYNTX bb1 KNKKNX bb2 PDRLEMNKYCX bb3 FCX bb4 KHTX bb5 HX bb6 ETK;
[0016] a3)MGKKLKKCLVELLKIFLL;(SY6)
[0017] a4)MKKAKEKEGGKYGGLKNHTALLGKRRVCHYTDKRKIWKNNGAARLSIAV; (SY11)
[0018] a5)MNEELVAAKKELFNLRFQNATNQLDNTSRIKEVRKNIARIQTVIAQKNA; (SY29)
[0019] a6)MPGVRRMPGGGGTDGGGPEGEKCSGERVFDGEAETGYFGSGIAFGVAGCL; (SY35)
[0020] a7)MPVHRNTNRFVIDILLKNLKKYKNTQFYILGCKKGCKSRKCLIYTLYCGE; (SY36)
[0021] a8)MTRRRSLKTLSVYHKTDTFGRKKRHFAQKCAISTLSAREG; (SY56)
[0022] a9)MRFLRHPCFTIKNLRLWI; (SY62)
[0023] a10)MKGALVSIICCIFANGFGYCHFQDIVGKQKHSLLFRVQPKASETYWK; (SY66)
[0024] a11)MKVLKIPNSGTTDGLNKQKQGLKVVLKANDHGRFGKKDRSVED; (SY70)
[0025] a12)MLFRKRNRVRGRKTTAPLYCGRSGNGNGRWNCKGYSDTGSGCWRWKRL; (SY71)
[0026] a13)MMNRKKIGSKRRQASRNISLGNPKKEAVKSLLLPLTRIIVPGKSTI. (SY72)
[0027] In structure a1), X aa1 Existence or non-existence, when X aa1 When it exists, X aa1 Selected from MEVNS; X aa2 Selected from I or V; X aa3 Selected from P or K; X aa4 Selected from M or I; X aa5 Selected from E or D; X aa6 Selected from K or R; X aa7 Selected from I or V; X aa8 Selected from K or R; X aa9 Selected from K or H; X aa10 Selected from R, K, or H; X aa11 Selected from R or K; X aa12 Existence or non-existence, when X aa12 When it exists, X aa12 Selected from N.
[0028] In a specific embodiment of the present invention, structure a1) is selected from the following structures:
[0029] a1-1)MEVNSMKVRPSVKPMCEKCKIIKRKGRVMVICENPKHKQRQG; (SY4)
[0030] a1-2)MKVRPSVKKMCEKCKIIKRKGKVMVICENPKHKQRQG; (SY18)
[0031] a1-3)MKVRPSVKPICEKCKVIKRKGHVMVICENPKHKQKQG; (SY22)
[0032] a1-4)MKVRPSVKPMCDKCRIIRRKGRVMVICENPKHKQRQGN; (SY23)
[0033] a1-5)MKVRPSVKPMCEKCKIIKRHGKVMVICENPKHKQRQG; (SY25)
[0034] a1-6)MKIRPSVKPMCEKCRIIKRKGRVMVICENPKHKQKQG; (SY10)
[0035] In structure a2), X bb1 Selected from K, M, or V, X bb2 Selected from D, S, or T, X bb3 Selected from K or R, X bb4 Selected from K or R, X bb5 Selected from L or V, X bb6 Selected from K or R.
[0036] In a specific embodiment of the present invention, structure a2) is selected from the following structures:
[0037] a2-1)MRVKITLACTECKQRNYNTKKNKKNDPDRLEMNKYCKFCKKHTLHKETK; (SY45)
[0038] a2-2)MRVKITLACTECKQRNYNTMKNKKNDPDRLEMNKYCRFCRKHTVHRETK; (SY47)
[0039] a2-3)MRVKITLACTECKQRNYNTMKNKKNSPDLEMNKYCRFCKKHTLHKETK; (SY73)
[0040] a2-4)MRVKITLACTECKQRNYNTVKNKKNTPDRLEMNKYCRFCRKHTLHKETK; (SY74)
[0041] In a specific embodiment of the present invention, the antifungal peptide against Candida auris is selected from at least one of SY4, SY6, SY10, SY11, SY18, SY22, SY23, SY25, SY29, SY35, SY36, SY45, SY47, SY56, SY62, SY66, SY70, SY71, SY72, SY73, and SY74.
[0042] The amino acid sequence of the antifungal peptide against Candida auris is as follows:
[0043] Table 2. Amino acid sequence information of antifungal peptides against Candida auris.
[0044] In some embodiments of the present invention, the fungus is Candida tropicalis, and the antifungal peptide against Candida tropicalis has at least one of the following structures:
[0045] b1)X aa1 MKVRPSVKX aa2 X aa3 CX aa4 KCX aa5 X aa6 IX aa7 RX aa8 GX aa9 VMVICENPKHKQRQG;
[0046] b2)MAVVGKKRRSIRNCRTGKKTGRYTNHIGSKGISGTRKNIR; (SY2)
[0047] b3)MTVKITLACTECKQRNYNTTKNKKNNPDRLEMKKYCRFCKKHTVHRETK; (SY58)
[0048] b4)MCSDRGAHTKAALIGHIKNGKKSYNNRVKTWSFRLSKLQHLSKEKL; (SY3)
[0049] b5)MSPAGWERWKSKYLVNLPGKNRIGRKMCYNNGNAHFPFCCLTGFAAA. (SY53)
[0050] In structure b1), X aa1 Existence or non-existence, when X aa1 When it exists, X aa1Selected from MEVNS; X aa2 Selected from P or K; X aa3 Selected from M or I; X aa4 Selected from E or D; X aa5 Selected from K or R; X aa6 Selected from I or V; X aa7 Selected from K or R; X aa8 Selected from K or N; X aa9 Selected from R or K.
[0051] In a specific embodiment of the present invention, structure b1) is selected from the following structures:
[0052] b1-1)MEVNSMKVRPSVKPMCEKCKIIKRKGRVMVICENPKHKQRQG; (SY4)
[0053] b1-2)MKVRPSVKKMCDKCKIIKRKGRVMVICENPKHKQRQG; (SY17)
[0054] b1-3)MKVRPSVKPMCEKCKIIRRKGRVMVICENPKHKQRQG; (SY26)
[0055] b1-4)MKVRPSVKKICDKCKVIKRNGKVMVICENPKHKQRQG; (SY15)
[0056] b1-5)MKVRPSVKPICDKCRVIKRKGRVMVICENPKHKQRQG; (SY19)
[0057] In a specific embodiment of the present invention, the antifungal peptide against Candida tropicalis is selected from at least one of SY2, SY17, SY26, SY58, SY3, SY4, SY15, and SY19.
[0058] The amino acid sequence of the antifungal peptide against Candida tropicalis is as follows:
[0059] Table 3. Amino acid sequence information of antifungal peptides against Candida tropicalis.
[0060] In some embodiments of the present invention, the fungus is Candida albicans, and the antifungal peptide against Candida albicans has at least one of the following structures:
[0061] c1)MX aa1 Xaa2 X aa3 ITLACTECKQRNYX aa4 TX aa5 KX aa6 KKX aa7 TX aa8 PDRX aa9 EX aa10 X aa11 KYCX aa12 FCX aa13 KHTX aa14 HRETK;
[0062] c2)MKVLKIPNSGTTDGLNKQKQGLKVVLKANDHGRFGKKDRSVED; (SY70)
[0063] c3)MCPWHNCRGTLPHGRLKKKSKKEQYKCISNGITNRLRPIGNVRPRSWVRS; (SY61)
[0064] c4)MKGALVSIICCIFANGFGYCHFQDIVGKQKHSLLFRVQPKASETYWK; (SY66)
[0065] c5)MKKRYGFIYVDKDDQGNGSLKRYRKKSFYWYKKVIETNGQTKFE; (SY67)
[0066] c6)MKNILIMLLELLILLILVKKI; (SY69)
[0067] c7)MYKQQIKRLLFHMLMHLLKMKILKIKKYLFI. (SY77)
[0068] In structure c1), X aa1 Selected from R or T; X aa2 Selected from T or V; X aa3 Selected from R or K; X aa4 Selected from D or N; X aa5 Selected from M, T, or V; X aa6 Selected from D or N; X aa7 Selected from A or N; X aa8 Selected from H, N, or T; X aa9 Selected from M or L; X aa10 Selected from T or M; X aa11 Selected from S, K, or N; X aa12 Selected from K or R; X aa13 Selected from G, K, or R; X aa14 Selected from L or V.
[0069] In a specific embodiment of the present invention, structure c1) is selected from the following structures:
[0070] c1-1)MRTRITLACTECKQRNYDTMKDKKAHPDRMETSKYCKFCGKHTLHRETK; (SY40)
[0071] c1-2)MTVKITLACTECKQRNYNTTKNKKNNPDRLEMKKYCRFCKKHTVHRETK; (SY58)
[0072] c1-3)MRVKITLACTECKQRNYNTVKNKKNTPDRLEMNKYCRFCRKHTLHKETK; (SY74)
[0073] In a specific embodiment of the present invention, the antifungal peptide against Candida albicans is selected from at least one of SY70, SY40, SY58, SY61, SY66, SY67, SY69, SY74, and SY77.
[0074] The amino acid sequence of the antifungal peptide against Candida albicans is as follows:
[0075] Table 4. Amino acid sequence information of antifungal peptides against Candida albicans.
[0076] In some embodiments of the present invention, the fungus is Cryptococcus neoformans (H99), and the antifungal peptide against Cryptococcus neoformans (H99) has at least one of the following structures:
[0077] d1)MKVRPSVKPX aa1 CEKCKX aa2 IX aa3 RKGX aa4 VMVICENPKHKQX aa5 QG;
[0078] d2)MRTRITLACTECKQRNYSTTKDKKVMPDRMEIKKYCKFCKSHTLHKETK; (SY42)
[0079] d3)MTRRRSLKTLSVYHKTDTFGRKKRHFAQKCAISTLSAREG; (SY56)
[0080] d4)MNEELVAAKKELFNLRFQNATNQLDNTSRIKEVRKNIARIQTVIAQKNA. (SY29)
[0081] In structure d1), X aa1 Selected from I or M; X aa2 Selected from V or I; X aa3 Selected from K or R; X aa4 Selected from H or R; X aa5 Selected from K or R.
[0082] In a specific embodiment of the present invention, the structure d1) is selected from the following structures:
[0083] d1-1)MKVRPSVKPICEKCKVIKRKGHVMVICENPKHKQKQG; (SY22)
[0084] d1-2)MKVRPSVKPMCEKCKIIRRKGRVMVICENPKHKQRQG; (SY26)
[0085] In a specific embodiment of the present invention, the antifungal peptide against Cryptococcus neoformans (H99) is selected from at least one of SY22, SY26, SY42, SY56, and SY29.
[0086] The amino acid sequence of the antifungal peptide against Cryptococcus neoformans (H99) is as follows:
[0087] Table 5. Amino acid sequence information of antifungal peptides against Cryptococcus neoformans.
[0088] In some embodiments of the present invention, the fungus is Candida glabrata, and the antifungal peptide against Candida glabrata has the following structure:
[0089] MKVRPSVKPMCDKCRVIKRKGRVMVICENPKHKQRQG(SY24).
[0090] In a second aspect, the present invention provides a pharmaceutical composition comprising an effective dose of the antifungal peptide described in the first aspect of the present invention.
[0091] Furthermore, the pharmaceutical composition also includes pharmaceutically acceptable excipients.
[0092] In a preferred embodiment of the invention, the pharmaceutical composition is suitable for non-gastrointestinal administration, such as via intravenous, intramuscular, intradermal, and subcutaneous routes. Therefore, depending on the applicable dosage form, the pharmaceutically acceptable excipient is selected from antioxidants, buffers, antibacterial agents, suspending agents, solubilizers, or solutes that make the formulation isotonic with the blood of the subject.
[0093] In one specific embodiment of the present invention, the pharmaceutical composition is an injection administered to the subject via intravenous injection.
[0094] Thirdly, the present invention provides a method for screening the antifungal peptides described in the first aspect of the present invention based on machine learning methods, the method comprising the following steps:
[0095] S1) Construct an open reading frame (ORF) prediction framework and use the ORF Finder to predict metagenomic data from colorectal cancer cohorts and healthy individuals to obtain the encoding ORF amino acid sequences;
[0096] S2) Construct a machine learning model for accurate identification of antifungal peptides. Input the amino acid sequence obtained in step S1) into the model for prediction, screen amino acid sequences with high scores, and perform species annotation and peptide abundance assessment on the gut metagenomics.
[0097] S3) Correlation analysis was performed on fungi and peptides present in more than 5% of the samples;
[0098] S4) Cross-validate the initially screened amino acid sequences with the metaproteome dataset to screen for sequences with protein expression potential;
[0099] S5) Synthesize peptides and use growth curves or plate colony counting methods to detect their antibacterial activity, and screen out peptides with significant antibacterial effects as candidate antifungal peptides.
[0100] In step S1), the present invention uses BWA v0.7.17, samtools v1.17, bedtools v2.29.2, trim_galore v0.6.10, MEGAHIT v1.2.9, and ORF Finder to build an ORF prediction framework.
[0101] The model for accurately identifying antifungal peptides described in step S2) was constructed using the following method:
[0102] S2.1) Collect antifungal peptide and non-antifungal peptide sequences as a dataset;
[0103] S2.2) Clean and deduplicatize the antifungal peptide and non-antifungal peptide data, and construct balanced and unbalanced datasets with various ratios based on sequence length (6-50 and 51-100), such as 1:1, 1:most class equal, and 1:2.
[0104] S2.3) The training set consists of 80% of the samples randomly selected from the antifungal peptides (labeled 1) and the corresponding proportion of non-antifungal peptides (labeled 0), and the test set consists of the remaining 20% of the samples.
[0105] S2.4) The model was fine-tuned within the Protein Bert framework to adapt to the antifungal peptide classification task, resulting in the best-performing model for different length ranges.
[0106] This invention found that the training set constructed at a 1:1 ratio exhibits the best performance and prediction results.
[0107] The method for screening antifungal peptides provided by this invention can shorten the antifungal peptide development cycle, reduce discovery costs, and improve the efficiency of antifungal drug development.
[0108] Fourthly, the present invention provides the use of any of the following in inhibiting fungal growth and reproduction or in the preparation of antifungal products:
[0109] (I) The antifungal peptide described in the first aspect of the present invention; and / or
[0110] (II) The pharmaceutical composition described in the second aspect of the present invention.
[0111] The fungus is selected from at least one of Candida auris, Candida tropicalis, Candida albicans, Cryptococcus neoformans (H99), and Candida glabrata.
[0112] The products are selected from pharmaceuticals, cosmetics, personal care products, food, food packaging, and antibacterial coatings.
[0113] In a specific embodiment of the present invention, the drugs include human medicines, veterinary drugs, and pesticides.
[0114] This invention overcomes the problems of traditional antimicrobial peptide design and synthesis, which rely on laboratories, are experimentally driven, time-consuming, labor-intensive, and inefficient. Based on an artificial intelligence drug screening system using natural language processing and deep learning, it identifies a group of antifungal peptides with significant inhibitory activity against Candida and Cryptococcus fungi, significantly shortening the antifungal peptide identification cycle, reducing discovery costs, and improving antibiotic development efficiency. Furthermore, the AI-based drug screening system provided by this invention will also discover more potential antifungal peptides in a shorter time. Attached Figure Description
[0115] Figure 1. Flowchart of antifungal peptide screening based on deep learning.
[0116] Figure 2 shows the global distribution of the 11 population cohorts collected in this invention.
[0117] Figure 3 shows the distribution of the training dataset by length.
[0118] Figure 4. Distribution of antifungal peptide sequence lengths predicted by ORF.
[0119] Figure 5. Length distribution of candidate antifungal peptides.
[0120] Figure 6. Similarity comparison between candidate antifungal peptides and the model training set.
[0121] Figure 7 Comparison of the amino acid composition of candidate antifungal peptides and model training set.
[0122] Figure 8. Comparison of the physicochemical properties of the candidate antifungal peptides and the amino acid physicochemical properties of the model training set.
[0123] Figure 9 shows the antifungal activity of candidate antifungal peptides against Candida auris. The horizontal axis represents the concentration of candidate antifungal peptides. "*" represents P≤0.05; "**" represents P≤0.01; "***" represents P≤0.001 (two-tailed T-test).
[0124] Figure 10 shows the antimicrobial activity of candidate antimicrobial peptides against Candida tropicalis. The horizontal axis represents the concentration of candidate antifungal peptides. "*" represents P≤0.05; "**" represents P≤0.01; "***" represents P≤0.001 (two-tailed T-test).
[0125] Figure 11 shows the antifungal activity of candidate antifungal peptides against Candida albicans. The horizontal axis represents the concentration of candidate antifungal peptides. "*" represents P≤0.05; "**" represents P≤0.01; "***" represents P≤0.001 (two-tailed T-test).
[0126] Figure 12 shows the antimicrobial activity of candidate antimicrobial peptides against Candida glabrata. The horizontal axis represents the concentration of candidate antifungal peptides. "*" represents P≤0.05; "**" represents P≤0.01; "***" represents P≤0.001 (two-tailed T-test).
[0127] Figure 13 shows the antifungal activity of candidate antifungal peptides against Cryptococcus neoformans. The horizontal axis represents the concentration of candidate antifungal peptides. "*" represents P≤0.05; "**" represents P≤0.01; "***" represents P≤0.001 (two-tailed T-test). Detailed Implementation
[0128] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0129] 1. Research Methods
[0130] This invention mines antifungal peptides (AFPs) from the gut microbiome using deep learning. The process, as shown in Figure 1, is divided into three parts: screening, identification, and validation. Specifically, it includes: establishing a multi-center population cohort gut microbiome metagenomic database and predicting microbiome open reading frames (ORFs) (module 1); establishing a model for accurately identifying antifungal peptides (module 2); evaluating the activity potential of candidate peptides and screening them based on their species-negative correlation (modules 3-4); an in vitro screening system (module 5); and an in vivo screening system (module 6).
[0131] 1.1 Establishment of a Gut Metagenomic Database
[0132] This invention collected metagenomic data from eight colorectal cancer cohorts (PRJNA731589; PRJEB10878; PRJEB27928; PRJEB6070; PRJNA447983; PRJEB7774; PRJEB12449; PRJDB4176) and three healthy cohorts (PRJEB11532; PRJNA319574; SRP080787) worldwide, including 684 colorectal cancer patient samples and 2065 healthy human samples, for a total of 2749 intestinal metagenomic samples.
[0133] 1.2 Prediction of Open Reading Frames
[0134] An open reading frame (ORF) is a DNA or RNA sequence that encodes a protein, located between a start codon (usually AUG) and a stop codon (such as UAA, UAG, or UGA). To ensure the reproducibility and representativeness of our experimental results, the datasets included in this study exhibit broad geographical heterogeneity and genetic background (Figure 2). This invention utilizes BWA v0.7.17, samtools v1.17, bedtools v2.29.2, trim_galore v0.6.10, MEGAHIT v1.2.9, and ORF Finder to construct an ORF prediction framework (Figure 3). Starting from metagenomic data of the human gut microbiome, after removing host DNA and adapters, the data is assembled de novo using MEGAHIT and predicted using ORF Finder to obtain the amino acid sequence.
[0135] 1.3 Model Construction for Precise Identification of Antifungal Peptides
[0136] Existing antifungal peptides (AFPs) and non-antifungal peptides (Non-AFPs) were collected and merged into a fine-tuning dataset. The dataset had an imbalanced sequence length distribution, as shown in Figure 3. The data was trained for prediction by grouping by length to ensure that the model did not predict whether a sequence was an antifungal peptide based on length characteristics. After grouping the dataset by different length intervals, multiple balanced and imbalanced datasets were constructed. These datasets were then fed into a pre-trained model (Protein BERT) for fine-tuning to obtain the best-performing model for different length intervals. The score represents the model's evaluation of the prediction results; a higher score indicates higher reliability. This invention sets the model cutoff value to 0.95. Finally, the sequences were sorted according to their scores, with higher-scoring sequences being prioritized.
[0137] This invention establishes balanced and unbalanced data models according to different proportions of data. The training set constructed with a 1:1 ratio exhibits the best performance and prediction effect.
[0138] The model for identifying antifungal peptides provided by this invention is optimized to eliminate the influence of sequence length on the model and enriches the prediction results. A higher cutoff value is set for the model's prediction results, selecting candidate antifungal peptides with higher scores, thus reducing computational cost. Table 6 shows a performance comparison between the model for identifying antifungal peptides provided by this invention and existing antimicrobial peptide models. The comparison results show that the model provided by this invention has higher accuracy and F1 score in identifying antifungal peptides.
[0139] Table 6 Performance comparison between the model of this invention and existing tools
[0140] 1.LEE H,LEE S,LEE I,et al.AMP-BERT:Prediction of antimicrobial peptide function based on a BERT model[J].Protein Science,2022,32(1).
[0141] 2.SU X,XU J,YIN Y,et al.Antimicrobial peptide identification using multi-scale convolutional network[J].BMC Bioinformatics,2019,20(1):730.
[0142] 3.SANTOS-JúNIOR C D,PAN S,ZHAO X-M,et al.Macrel:antimicrobial peptide screening in genomes and metagenomes[J].PeerJ,2020,8:e10555.
[0143] 4.BHADRA P,YAN J,LI J,et al.AmPEP:Sequence-basedprediction ofantimicrobial peptides using distribution patterns of amino acid properties and random forest[J].Scientific Reports,2018,8(1):1697.
[0144] 5.LI C,SUTHERLAND D,HAMMOND S A,et al.AMPlify:attentive deep learning model for discovery of novel antimicrobial peptides effective against WHO priority pathogens[J].BMC Genomics,2022,23(1):77.
[0145] 6.FINGERHUT LCHW,MILLER DJ,STRUGNELL JM,et al.ampir:an Rpackage for fast genome-wide prediction ofantimicrobialpeptides[J].Bioinformatics,2021,36(21):5262-3.
[0146] 1.4 Gut Microbiome Species Annotation
[0147] A custom annotation database was built using Jellyfish (v2.3.0) software, including the NCBI RefSeq database, FungiDB, and Ensemble database, and fungal species were annotated for 11 population cohorts using Kraken2 v2.1.3 software.
[0148] 1.5 Prediction of peptide abundance in the gut microbiome
[0149] To obtain the abundance information of candidate antifungal peptides in each metagenomic sample, this invention uses Paladin software to perform antifungal peptide abundance analysis on 11 population cohorts.
[0150] 1.6 Correlation Analysis
[0151] Spearman correlation analysis was performed on fungi and candidate peptides present in more than 5% of samples in each population cohort. The correlation was calculated using the CorAndPvalue(Spearman) function of the 'WGCNA' package in R and visualized using the Pheatmap package. Candidate AFPs were further screened based on correlation < 0 and P < 0.05. The analysis was performed in R (V4.2.3).
[0152] 1.7 Amino acid sequence evaluation
[0153] The amino acid content of the peptide was calculated using the 'ProteinAnalysis' function of the 'Bio.SeqUtils.ProtParam' module in Biopython, and the analysis was performed in Python (V3.11.0). The physicochemical properties of the peptide were evaluated using the 'Peptides' package in R, and violin plots were drawn using 'ggplot2' in R.
[0154] 1.8 Peptide Synthesis
[0155] The peptides used in this study were synthesized by Taupe and Gir Biochemicals, and their precise molecular weights were determined by mass spectrometry. The purity of all peptides was determined by high-performance liquid chromatography (HPLC), and all purity levels were greater than 95%.
[0156] 1.9 Antibacterial activity assay
[0157] ①Strain preservation: The strain is frozen in 30% glycerol and stored at -80℃ for later use;
[0158] ②Preparation of YPD (Yeast Extract Peptone Dextrose) culture medium;
[0159] ③ Bacterial strain resuscitation: Take 5 ml of 1X YPD liquid culture medium into a 15 ml centrifuge tube; take a small amount of cells and seed them into YPD liquid culture medium, and incubate overnight at 30℃ with a shaker at 180-200 rpm for 15 hours.
[0160] ④ Determine strain density: After strain resuscitation, centrifuge at 3000 rpm for 2 minutes and discard the supernatant; wash twice with 1X PBS buffer and discard the supernatant; prepare a bacterial suspension using 2X YPD liquid medium to achieve an initial OD600 value of 0.04; prepare candidate antifungal peptide solutions of a certain concentration according to the experimental plan; add 75 μl of each peptide solution and 75 μl of bacterial suspension to a flat-bottomed 96-well plate; continuously monitor the plate at 30℃ for 24 hours, measuring OD every 15 minutes. 600 ;
[0161] ⑤ Determine the number of strains: After continuous detection for 24 hours, perform culture omics. Determine the dilution factor for culture based on the strain density. Dilute to a suitable gradient using 1X PBS buffer, take 50 μl and plate it. After incubating the plate at 30℃ for 24 hours, perform statistical counting and plot the data using GraphPad Prism 8.
[0162] 2. Research Findings
[0163] 2.1 Screening of candidate antifungal peptides
[0164] This study completed ORF predictions for metagenomics in 11 population cohorts. The colorectal cancer group yielded 3,755,508,707 amino acid sequences, while the healthy control group yielded 8,922,395,898 amino acid sequences. The predicted peptide sequence lengths were mainly distributed in the 6-50 amino acid range (Figure 4). Therefore, this study set the model prediction range to 6-50 and the cutoff value to 0.99, resulting in 10,819,929 candidate antifungal peptides (C_AFPs) in the colorectal cancer group and 21,884,138 candidate antifungal peptides in the healthy control group, for a total of 27,533,632.
[0165] 2.2 Identification of candidate antifungal peptides
[0166] While performing model predictions, this study also conducted species annotation and peptide abundance assessment of the gut metagenomics. After model predictions were completed, we performed correlation analysis on fungi and peptides present in more than 5% of samples from each population cohort. Considering that some peptides may exhibit different or even opposite correlations under different host states, we divided the population cohorts into a colorectal cancer group and a healthy group during the correlation analysis. We screened out peptides in each cohort that showed a statistically significant negative correlation with most fungi. The results showed that 4367 peptides were present only in the colorectal cancer group and 7082 peptides were present only in the healthy group. After removing completely repetitive sequences, a total of 16493 candidate antifungal peptides were obtained for further analysis.
[0167] Next, to ensure that the predicted ORF expression in bioinformatics represents protein / peptide, we performed cross-validation with a metaproteomics dataset. Using the PXD009056 and PXD038906 metaproteomics datasets, we cross-validated 16493 candidate antifungal peptides with 420 metaproteomics datasets using the Andromeda (MaxQuant) method to screen for candidate antifungal peptides with protein expression potential. The results showed that a total of 77 candidate antifungal peptides had expression potential. The peptide length distribution is shown in Figure 5. The lengths of these 77 peptides ranged from 17 to 50 AAs, with 37 AAs and 49 AAs being the most common.
[0168] Furthermore, this study evaluated the similarity between candidate antifungal peptides and the model's positive training set (which consisted of previously collected known AFP data). As shown in Figures 6-8, the candidate antifungal peptides screened in this study showed mostly less than 60% similarity to the known datasets, and their amino acid compositions were not entirely similar. They also exhibited low sequence similarity to the antifungal peptides in the training set. These results not only demonstrate the efficiency and innovation of the designed process in generating antimicrobial peptides (AFPs) but also showcase its significant advantage in terms of diversity.
[0169] 2.3 Validation of candidate antifungal peptides
[0170] This invention synthesized 70 candidate antifungal peptides and evaluated the antifungal activity of the 70 candidate antifungal peptides against 5 pathogenic fungi: Candida albicans, Candida glabrata, Candida auris, Candida tropicalis, and Cryptococcus neoformans (H99).
[0171] The five fungal strains were treated with a high concentration (500 μg / ml) of candidate antifungal peptides, and the peptides were screened in the first round based on the antifungal results. The initial screening results showed that the candidate antifungal peptides exhibited inhibitory effects on the five fungal strains, with varying inhibitory effects and antifungal activities against different fungi. Next, the inventors tested the antifungal activity of the candidate antifungal peptides with inhibitory effects at a low concentration (2 μg / ml). A total of 36 peptides with antifungal activity were obtained from 44 candidate peptides, and the amino acid sequences of these 36 peptides are shown in SEQ ID NO.1-SEQ ID NO.36.
[0172] The antibacterial data results of the 36 antifungal peptides are shown in Table 7.
[0173] Table 7 Antibacterial activity of antifungal peptides
[0174] Based on the antibacterial activity data in Table 7, the following conclusions can be drawn:
[0175] (1) 22 candidate antifungal peptides showed strong activity against Candida auris, namely SY4, SY6, SY10, SY11, SY18, SY22, SY23, SY25, SY29, SY35, SY36, SY45, SY47, SY56, SY62, SY66, SY70, SY71, SY72, SY73, and SY74 (Figure 9).
[0176] (2) Nine candidate antifungal peptides showed strong antifungal activity against Candida tropicalis, namely SY2, SY17, SY26, SY58, SY3, SY4, SY15 and SY19 (Figure 10).
[0177] (3) Nine candidate antifungal peptides showed strong antibacterial activity against Candida albicans, namely SY70, SY40, SY58, SY61, SY66, SY67, SY69, SY74 and SY77 (Figure 11).
[0178] (4) One candidate antifungal peptide, SY24, showed strong antifungal activity against Candida glabrata (Figure 12).
[0179] (5) Five candidate antifungal peptides showed strong antibacterial activity against Cryptococcus neoformans (H99), namely SY22, SY26, SY42, SY56 and SY29 (Figure 13).
[0180] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. An antifungal peptide, characterized in that, The antifungal peptide is selected from at least one of the polypeptides shown in SEQ ID NO.1-SEQ ID NO.
36.
2. The antifungal peptide according to claim 1, characterized in that, The fungus is selected from at least one of Candida auris, Candida tropicalis, Candida albicans, Cryptococcus neoformans (H99), and Candida glabrata.
3. A pharmaceutical composition comprising an effective dose of the antifungal peptide of claim 1 or 2.
4. The pharmaceutical composition of claim 3, wherein, The pharmaceutical composition also includes pharmaceutically acceptable excipients.
5. A method for screening the antifungal peptides of claim 1 or 2 based on machine learning methods, the method comprising the following steps: S1) Construct an ORF prediction framework and use ORF Finder to predict the amino acid sequences from metagenomic data of colorectal cancer cohorts and healthy individuals; S2) Construct a model for accurate identification of antifungal peptides. Input the amino acid sequence obtained in step S1) into the model for prediction, screen amino acid sequences with high scores, and perform species annotation and peptide abundance assessment on the gut metagenomics. S3) Perform correlation analysis on fungi and peptides present in more than 5% of the samples included in the cohort; S4) Cross-validate the initially screened amino acid sequences with the metaproteomics dataset to screen for sequences with protein expression potential; S5) Synthesize peptides and use fungi to test their antibacterial activity, and screen out peptides with significant antibacterial effects as antifungal peptides.
6. The method of claim 5, wherein, In step S1), an ORF prediction framework is built using BWA v0.7.17, samtools v1.17, bedtools v2.29.2, trim_galore v0.6.10, MEGAHIT v1.2.9, and ORF Finder.
7. The method of claim 5, wherein, The model for accurately identifying antifungal peptides described in step S2) was constructed using the following method: S2.1) Collect antifungal peptide and non-antifungal peptide sequences as a dataset; S2.2) Clean and deduplicatize the antifungal peptide and non-antifungal peptide data, and construct balanced and unbalanced datasets with various ratios based on sequence length (6-50 and 51-100), such as 1:1, 1:most class equal, and 1:
2. S2.3) The training set consists of 80% of the samples randomly selected from the antifungal peptides (labeled 1) and the corresponding proportion of non-antifungal peptides (labeled 0), and the test set consists of the remaining 20% of the samples; S2.4) The model was fine-tuned within the Protein Bert framework to adapt to the antifungal peptide classification task, resulting in the best-performing model for different length ranges.
8. Any of the following applications in inhibiting fungal growth and reproduction or in the preparation of antifungal products: (I) The antifungal peptide according to claim 1 or 2; and / or (II) The pharmaceutical composition according to claim 3 or 4; The fungus is selected from at least one of Candida auris, Candida tropicalis, Candida albicans, Cryptococcus neoformans (H99), and Candida glabrata.
9. Use according to claim 8, characterized in that, The product is selected from a pharmaceutical product, a cosmetic product, a hygiene product, a food product, a food packaging, an antibacterial coating.
10. Use according to claim 9, characterized in that, The pharmaceutical product comprises a human pharmaceutical product, a veterinary drug, or a pesticide.
