Research On Antimicrobial Peptide Prediction Method Based On Amino Acid Embedding And Feature Fusion

Antimicrobial resistance is one of our most serious health threats.Antimicrobial peptides(AMPs),effecter molecules of innate immune system,can defend host organisms against microbes and most have shown a lowered likelihood for bacteria to form resistance compared to many conventional drugs.Thus,AMPs are gaining popularity as better substitute to antibiotics.To aid researchers in novel AMPs discovery,we design computational approaches to screen for promising candidates.In this work,we mainly carried out the following research:The recognition of AMPs is important for the design of molecular drugs.In this paper,we have constructed and collated four AMP datasets,including the frog AMP dataset,the AMP structure dataset,the AMP function dataset,and the comprehensive AMP dataset.And for different datasets and different goals,we designed spatial autocorrelation function,enhanced learning model,deep learning model and amino acid embedding,respectively,to achieve AMP identification and function and structure prediction.In view of the characteristics of biological sequences,we proposed a method using amino acid embedding technology and PSSM to encode sequence,so that the model can automatically capture the similarity between amino acids and retain the evolutionary information of the sequence.We have also integrated an automatic feature fusion module into the deep learning model,so that the model can select different features for different sequences,and automatically fuse heterogeneous information.Results show that the proposed model outperforms state-of-the-art methods on recognition of AMPs.By visualizing the data in each layer of the model,we overcome the black-box nature of deep learning,explained the knowledge learned by each part of the model,and verified the effectiveness of the model.The ACEP model can capture the similar relationship between amino acids,calculate the attention scores for different parts of a peptide sequence,so that spot those important parts that significantly contribute to the final predictions and automatically fuse a variety of heterogeneous information and features.For high-throughput AMPs recognition,open source software and datasets are made freely available at https://github.com/Fuhaoyi/ACEP.