Rational Discovery of Antimicrobial Peptides by Means of Artificial Intelligence

Abstract

Antibiotic resistance is a worldwide public health problem due to the costs and mortality rates it generates. However, the large pharmaceutical industries have stopped searching for new antibiotics because of their low profitability, given the rapid replacement rates imposed by the increasingly observed resistance acquired by microorganisms. Alternatively, antimicrobial peptides (AMPs) have emerged as potent molecules with a much lower rate of resistance generation. The discovery of these peptides is carried out through extensive in vitro screenings of either rational or non-rational libraries. These processes are tedious and expensive and generate only a few AMP candidates, most of which fail to show the required activity and physicochemical properties for practical applications. This work proposes implementing an artificial intelligence algorithm to reduce the required experimentation and increase the efficiency of high-activity AMP discovery. Our deep learning (DL) model, called AMPs-Net, outperforms the state-of-the-art method by 8.8% in average precision. Furthermore, it is highly accurate to predict the antibacterial and antiviral capacity of a large number of AMPs. Our search led to identifying two unreported antimicrobial motifs and two novel antimicrobial peptides related to them. Moreover, by coupling DL with molecular dynamics (MD) simulations, we were able to find a multifunctional peptide with promising therapeutic effects. Our work validates our previously proposed pipeline for a more efficient rational discovery of novel AMPs.

Recent Progress in the Discovery and Design of Antimicrobial Peptides Using Traditional Machine Learning and Deep Learning

“…The early study by Fjell et al [ 114 ] used QSAR descriptors and fed them to an artificial neural network that predicted peptide activity against P. aeruginosa , and achieved 94% accuracy in identifying highly active peptides. Since then, many DL methods have been proposed for predicting AMps [ 80 , 89 , 100 , 114 , 115 , 116 , 117 , 118 , 119 ].…”

Deep Learning and Antibiotic Resistance

“…In vitro evaluation of these AMPs concluded that three out of four peptides presented antimicrobial activity. RD-10, a peptide consisting of ten amino acids, showed promising bacteriostatic activity and it was considered a potential candidate for further preclinical examinations [ 19 ].…”

“…Unlike most conventional antibiotics, which have specific functional or structural targets, AMPs act directly on the microorganisms, often causing cell lysis, or modulate the host immunity to enhance defense against microorganisms [ 39 ]. Moreover, they act faster than conventional antibiotics [ 19 ], have a narrower active concentration window for killing [ 40 ], and do not typically damage the DNA of their targets [ 20 , 41 ], though there have been studies describing how AMPs inhibit critical intracellular functions by binding to DNA, RNA, or intracellular proteins [ 42 ]. As a result, they do not induce resistance to the extent that is observed with conventional antibiotics [ 21 ].…”

“…For using recurrent neural networks (RNNs), representation methods together with architecture for each representation must be used to analyze the sequence of amino acids. Vectors are commonly used to describe essential properties of AMPs [ 19 ]. They are used to represent the order of amino acids in AMPs, for example by representing peptides as sequences of vectors, describing the presence of one of the 20 essential amino acids in the sequence.…”

“…However, sequential representation does not consider the interactions between amino acids and the atoms within each amino acid. Essential properties of AMPs must be included, such as amino acid composition or composition–transition–distribution, to make the vectors more descriptive [ 19 ].…”