Deep Learning-Based Discovery And Antimicrobial Activity Study Of Antimicrobial Peptides

Overuse of antibiotics has resulted in the emergence of drug-resistant bacteria.Nowadays,almost all antibiotics are varying degrees resistant in clinical,which greatly reduces the therapeutic effect of antibiotics.Therefore,developing new antibacterials or antimicrobial technology is an urgent need in the industry and academia.Antimicrobial peptides are considered as the most promising new generation of antiagents.However,the R&D process of new antimicrobial peptides is complex and costly.In recent years,emerging artificial intelligence can accelerate the discovery of antimicrobial peptides,and greatly saved R&D costs.In this study,three prediction models,including the multi-classification model,GRU-CNN model,and filtering model,were established through deep learning,and 12possible active peptides were filtered out from 30,000 random peptides.The activity of the 12 peptides were verified by in vitro antibacterial experiments.The results showed that P2,P11,and P12 with good antibacterial activities（MIC≤60μg/m L）,which means that our virtual model has a better precision.By measuring the circular dichroism spectra of 12 Peptides,we found that the secondary structures of the peptides includeα-helix,β-sheet,and disordered structure,indicating that the peptides filtered by the deep learning model we established manifested structural diversity.Stability is the main obstacle restricting the druggability of peptides.we assayed the Na Cl and serum sensitivity of P2,P11,and P12.The experimental results indicated that P2,P11,and P12 are able to maintain antibacterial activity in Na Cl solution.However,three peptides behaved differently satisfactorily in rat serum,the protein binding rates of P2,P11,and P12 were 3.22%,76.95%,and 31.72%,respectively.In addition,P12 had the longest serum half-life,which exceeded 2 hours.We further evaluated the biocompatibility of our three peptides.The experimental results indicated that the hemolytic rate of P2,P11,and P12 to rat erythrocytes were extremely low,and the 5%hemolytic concentration（HC₅）of all three peptides were greater than 400μg/m L.However,all the three peptides exhibited some cytotoxicity to human liver cells（HL-7702）and human embryonic kidney cells（HEK-293T）:the IC₅₀ of P2,P11,and P12 to HL-7702 were 43.67,77.26,and 25.91μg/m L,respectively.And for HEK-293T,the IC₅₀of P2,P11,and P12 were 59.52,>100,and 26.01μg/m L,respectively.Further,we selected P12 as the research object according to the above results,then we explored the antimicrobial mechanism,potential for inducing bacterial resistance,and in vivo antibacterial activity.Time-killing kinetics showed that S.aureus or E.coli were killed by P12 at a concentration of 4×MIC within 2 hours.Antibacterial mechanism,including cytoplasmic membrane depolarization assay,outer/inner membrane permeation assay,propidium iodide uptake assay,and scanning electron microscope observation,indicated that P12 mainly exerts its antibacterial effect by breaking the cell membrane.In addition,a high concentration of P12（8×MIC）can effectively inhibit the formation of S.aureus biofilm.The development of drug resistance in S.aureus treated with P12 and ciprofloxacin indicated that the MICs of ciprofloxacin increased 8-fold within 21 passages against S.aureus while the MICs of P12 only increased 1-fold,which indicated that P12 is hard to induce bacterial drug resistance.In the wound infected with S.aureus,P12 was effective in clearing the colony load,which indicated that P12 has excellent antibacterial activity in vivo.Moreover,Acute toxicity experiments proved that P12 shows low toxicity in mice.In conclusion,this study successfully constructed three predictive models through deep learning,which can be used for the discovery of antimicrobial peptides.The P12filtered by this model exhibited excellent in vivo and in vitro antibacterial activity,high stability,low hemolytic activity,and a certain degree of cytotoxicity,a large dose of P12（400mg/kg）did not cause obvious pathological changes in the major organs of mice.In addition,P12 exerts an antibacterial effect through the membrane destruction mechanism,and it did not develop resistance rapidly.The study provides a very effective strategy for filter new antibacterials and a new perspective for the high-throughput filtering of other antibacterials.