Advances in Antimicrobial Peptide Discovery via Machine Learning and Delivery via Nanotechnology

Abstract

Antimicrobial peptides (AMPs) have been investigated for their potential use as an alternative to antibiotics due to the increased demand for new antimicrobial agents. AMPs, widely found in nature and obtained from microorganisms, have a broad range of antimicrobial protection, allowing them to be applied in the treatment of infections caused by various pathogenic microorganisms. Since these peptides are primarily cationic, they prefer anionic bacterial membranes due to electrostatic interactions. However, the applications of AMPs are currently limited owing to their hemolytic activity, poor bioavailability, degradation from proteolytic enzymes, and high-cost production. To overcome these limitations, nanotechnology has been used to improve AMP bioavailability, permeation across barriers, and/or protection against degradation. In addition, machine learning has been investigated due to its time-saving and cost-effective algorithms to predict AMPs. There are numerous databases available to train machine learning models. In this review, we focus on nanotechnology approaches for AMP delivery and advances in AMP design via machine learning. The AMP sources, classification, structures, antimicrobial mechanisms, their role in diseases, peptide engineering technologies, currently available databases, and machine learning techniques used to predict AMPs with minimal toxicity are discussed in detail.

Keywords: LL-37; antibiotic resistance; antimicrobial peptide; drug delivery; machine learning; nanotechnology; peptide database; peptide design; peptide engineering.

Figure 1

AMPs interact with microbial membranes and their associated mechanisms of action. The mechanisms for membrane permeation begin with (a) AMP interaction with microbial membranes, allowing AMPs to insert themselves. (b) The barrel-stave model results when AMPs insert perpendicularly into microbial membranes and self-assemble to form pores [83,84]. (c) The carpet model works by absorbing AMPs, resulting in disrupting the microbial membranes through micellization [86]. (d) The toroidal-pore model results from incorporating AMPs to form a continuous interface with the microbial membrane [28,84]. AMPs can also translocate across microbial membranes and bind to intracellular targets, resulting in cell death. These intracellular targets include (I) inhibiting DNA and RNA synthesis, (II) inhibiting protein synthesis, and (III) inhibition of cell wall synthesis [22]. Adapted from (Bahar et al., 2013) CC BY 3.0 [103]. Created with BioRender.com (accessed on 15 December 2022).

Figure 2

Two-dimensional structures of multiple FDA-approved AMPs. These AMPs include bacitracin [105], colistin [109], daptomycin [106], gramicidin A [110], vancomycin [108], teicoplanin [107], dalbavancin [111], oritavancin [112], telavancin [113], boceprevir [116], telaprevir [117], and enfuvirtide [118]. Created with BioRender.com (accessed on 15 December 2022).

Figure 3

Schematic representation of machine learning algorithms used to develop prediction models. Machine learning algorithms are divided into two main categories. (a) Supervised machine learning uses a training set of known AMPs and the desired outcome, such as antimicrobial or non-antimicrobial properties. This model uses raw data of known AMPs to train the algorithm, resulting in a trained model for analyzing new data. Then, when inputting new data into the model, it can predict if that new sequence has antimicrobial or non-antimicrobial properties. (b) Unsupervised machine learning does not have a training data set. Instead, unsupervised learning predicts if new sequences are antimicrobial or non-antimicrobial based on any patterns or trends associated with the data. Adapted from (Ma, Y., 2018) CC BY 4.0 [163]. Created with BioRender.com (accessed on 15 December 2022).

Figure 5

Machine learning uses various models to classify and predict outcomes based on data. (a) k-nearest neighbor (KNN) classifies data using the nearest neighbor, which is the data point closest to the desired point. When the nearest items are determined, the algorithm recommends using these items based on the previous majority vote of nearest neighbors [204]. (b) Support vector machine (SVM) is a machine learning algorithm that classifies data by finding a hyperplane to separate the data. A hyperplane creates a line that is the maximum distance from the nearest training points, resulting in a strong separation between classes [203,205]. (c) Artificial neural network (ANN) uses hyper-parameters to approximate relationships between input and output values. ANNs are composed of computational units, which receive signals and transform them to make a prediction using the network formed by layers [203,206]. (d) Random forests build multiple decision trees to classify an event and avoid issues such as overfitting data due to outliers [204]. Adapted from CC BY-SA 3.0 [209], CC BY 4.0 [210], CC BY-SA 3.0 [211], and CC BY-SA 4.0 [212].

Figure 7

Overall scheme of the advantages of AMPs and different approaches to improve challenges associated with therapeutic applications. AMPs have advantages over current treatments due to their broad range protection and other properties not found in antibiotics. However, these peptides have obstacles when being applied to therapeutic use due to complications such as toxicity and instability. Delivery approaches are used to overcome some of these issues with AMPs. Machine learning has shown to be a promising approach for using models to predict and design AMPs with less therapeutic complications. Databases help train machine learning models to enhance prediction methods for AMPs. Created with BioRender.com (accessed on 21 December 2022).

Figure 4

A general machine learning workflow of AMP discovery and design, including a summary of the major techniques in each stage of the workflow. Redistributed from (Yan et al., 2022) CC BY 4.0 [193].

Figure 6

Nanoparticle delivery vehicles for AMPs to enhance biological stability. These delivery vehicles have different purposes based on the properties desired for delivery. Metal nanoparticles have demonstrated the ability to enhance the antimicrobial activity of AMPs but have limitations at high concentrations due to potential toxicity [235]. Lipid nanoparticles are biocompatible, which gives them the advantage of being used to avoid toxicity in tissues [236]. Polymeric nanoparticles contain a coating of molecules with specific properties such as antimicrobial activity and good bioavailability [237]. Forming other nanostructures helps combat complications, such as different structures, allowing biofilm inhibition [238]. Modified with permission from CC-BY 4.0 [235]. Created with BioRender.com (accessed on 21 December 2022).