Computational antimicrobial peptide design and evaluation against multidrug-resistant clinical isolates of bacteria

Abstract

There is a pressing need for new therapeutics to combat multi-drug and carbapenem-resistant bacterial pathogens. This challenge prompted us to use a long short-term memory (LSTM) language model to understand the underlying grammar, i.e. the arrangement and frequencies of amino acid residues, in known antimicrobial peptide sequences. According to the output of our LSTM network, we synthesized 10 peptides and tested them against known bacterial pathogens. All of these peptides displayed broad-spectrum antimicrobial activity, validating our LSTM-based peptide design approach. Our two most effective antimicrobial peptides displayed activity against multidrugresistant (MDR) clinical isolates of Escherichia coli, Acinetobacter baumannii, Klebsiella pneumoniae, Pseudomonas aeruginosa, Staphylococcus aureus, and coagulase-negative staphylococci (CoNS) strains. High activity against extended-spectrum beta-lactamase (ESBL), meticillin-resistant S. aureus (MRSA), and carbapenem-resistant strains was also observed. Our peptides selectively interacted with and disrupted bacterial cell membranes and caused secondary gene-regulatory effects. Initial structural characterization revealed that our most effective peptide appeared to be well folded. We conclude that our LSTM-based peptide design approach appears to have correctly deciphered the underlying grammar of antimicrobial peptide sequences, as demonstrated by the experimentally observed efficacy of our designed peptides.Antibiotic resistance is an ever-increasing threat which is gradually rendering our current repertoire of antibiotics obsolete. If no new drugs are developed, deaths due to antimicrobial resistance are expected to exceed 10 million annually by 2050 (1). Antimicrobial peptides (AMPs) are one potential solution to this problem. Naturally occurring AMPs continue to remain an important component of the innate immune system despite their ancient evolutionary origin and widespread prevalence across many forms of life (2). Some derivatives of these such as pexiganan (3), omiganan (4), and OP-145 (5) are currently undergoing late-stage clinical trials (for diabetic foot ulcers, rosacea, and ear infections respectively (6)). Other peptides such as Novexatin (7) and Lytixar (8) are currently undergoing early-stage clinical trials (for the treatment of toenail fungal infections and MRSA respectively (6)). Currently, over 2000 natural and designed antimicrobial peptides are curated in various databases (9-11), and have displayed broad-spectrum activity against Gram positive, Gram negative, fungal, mycobacterial, and protozoal pathogens (9). Designed AMPs vs. MDR isolates.positive charge are attracted and incorporated into negatively charged bacterial membranes. Once inside the membrane, they are believed to cause disruption through three possible mechanisms: toroidal pore formation (12), carpet formation (13), and barrel stave formation (14). Although the specifics of each mechanism differ, all propose peptide-induced membrane rupture, allowing cytoplasmic leak...

Deep learning regression model for antimicrobial peptide design

“…Beyond identifying and optimizing existing AMPs, several groups have used variational autoencoders (Das et al, 35 2018) or generative recurrent neural network (RNN)-based models (Müller et al, 2018;Nagarajan et al, 2018) to 36 generate new AMP sequences. These models generate sequences without an associated prediction of activity, 37 although Nagarajan et al further added a regression model (performance unspecified) to filter the designed 38 sequences by predicted activity.…”

Designing Anticancer Peptides by Constructive Machine Learning

“…As imilar computational concept has recently been pursued to yield membranolytic antimicrobial peptides, relying on extensive machine learning for activity prediction. [18] In contrast, after trainingt he computer model on genericp eptide features, fine-tuning by transfer learning [19] with merely a small number of ACPs proved sufficient to obtain novel active ACPs in this present study.T hese computer-generated amino acid sequences were maximally 27 %i dentical (56 %s imilarity) to the peptides from the fine-tuning set, as determined by LALIGN. [20] Of course, the presented example is only one instance of generatived esign,a nd it will be important to train this model on other peptidec lasses of interest.…”

“…With 10 out of the 12 peptides showing the desired biological activity, the results of this study validate the constructive deep learning approach for prospective de novo design of bioactive peptides. A similar computational concept has recently been pursued to yield membranolytic antimicrobial peptides, relying on extensive machine learning for activity prediction . In contrast, after training the computer model on generic peptide features, fine‐tuning by transfer learning with merely a small number of ACPs proved sufficient to obtain novel active ACPs in this present study.…”

A uniformin vitroefficacy dataset to guide antimicrobial peptide design

“…We synthesized and experimentally characterized 20 peptides. 10 sequences (NN2_0018 → NN2_0055) posed good antimicrobial activity and were described in our previous work 10 . Another 10 sequences (NN2_0000 → NN2_0009) possessed poor antimicrobial activity.…”

“…We identifed effective, broad-spectrum peptides, using a relative scoring scheme 10 . Simply described, for a given peptide, its peptide score was calculated by counting number of cultures it inhibited with the lowest MIC(in comparison to the MICs of all other peptides for a given culture).…”