Machine learning-guided discovery and design of non-hemolytic peptides

VIPER: Virus Inhibition Via Peptide Engineering and Receptor Mimicry

A key step in most viral infections is the binding of a viral protein to a host receptor, leading to the virus entering the host cell. Disrupting this protein-protein interaction is an effective strategy for preventing infection and subsequent disease. Building on recent advances in computational tools for structural biology, we introduce Virus Inhibition via Peptide Engineering and Receptor Mimicry (VIPER), a novel approach for the automatic derivation and optimization of biomimetic decoy peptides that mimic binding sites of human proteins. VIPER leverages structural data from human-pathogen protein complexes, yielding peptides that can competitively inhibit viral entry by mimicking the natural receptor. We computationally validated VIPER using molecular dynamics simulations and showcased its applicability on three clinically relevant viruses, highlighting its potential to accelerate therapeutic development. With a focus on reproducibility and extensibility, VIPER can facilitate the rapid development of antiviral inhibitors by automating the design and optimization of biomimetic compounds.

Enhancing Antimicrobial Peptides from Frog Skin: A Rational Approach

Antimicrobial resistance is a global health threat, which has been worsened by the slow development of new antibiotics. The rational design of natural-derived antimicrobial peptides (AMPs) offers a promising alternative for enhancing the efficacy of AMPs and accelerating drug discovery. This paper describes the rational design of improved peptide derivatives starting from hylin-Pul3, a peptide previously isolated from the frog Boana pulchella, by optimizing its hydrophobicity, cationicity, and amphipathicity. In silico screening identified six promising candidates: dHP3-31, dHP3-50, dHP3-50.137, dHP3-50.190, dHP3-84, and dHP3-84.39. These derivatives exhibited enhanced activity against Gram-negative bacteria, emphasizing the role of cationicity and the strategic arginine incorporation. Hemolytic assays revealed the derivatives' improved selectivity, particularly for the derivatives with "imperfect amphipathicity". In fibroblast assays, dHP3-84 was well-tolerated, while dHP3-84.39 promoted cell proliferation. Antioxidant assays (ABTS assays) highlighted the Trp-containing derivatives' (dHP3-50.137, dHP3-31) significant activity. The lipid membrane interaction studies showed that hylin-Pul3 disrupts membranes directly, while dHP3-84.39, dHP3-50, and dHP3-50.137 promote vesicle aggregation. Conversely, dHP3-84 did not induce membrane disruption or aggregation, suggesting an intracellular mode of action. Machine learning models were effective in predicting bioactivity, as these predicted AMPs showed enhanced selectivity and potency. Among them, dHP3-84 demonstrated broad-spectrum potential. These findings highlight the value of rational design, in silico screening, and structure-activity studies in optimizing AMPs for therapeutic applications.

Prediction of hemolytic peptides and their hemolytic concentration

Peptide-based drugs often fail in clinical trials due to their toxicity or hemolytic activity against red blood cells (RBCs). Existing methods predict hemolytic peptides but not the concentration (HC50) required to lyse 50% of RBCs. This study develops classification and regression models to identify and quantify hemolytic activity. These models train on 1926 peptides with experimentally determined HC50 against mammalian RBCs. Analysis indicates that hydrophobic and positively charged residues were associated with higher hemolytic activity. Among classification models, including machine learning (ML), quantum ML, and protein language models, a hybrid model combining random forest (RF) and a motif-based approach achieves the highest area under the receiver operating characteristic curve (AUROC) of 0.921. Regression models achieve a Pearson correlation coefficient (R) of 0.739 and a coefficient of determination (R²) of 0.543. These models outperform existing methods and are implemented in HemoPI2, a web-based platform and standalone software for designing peptides with desired HC50 values ( http://webs.iiitd.edu.in/raghava/hemopi2/ ).

Design, synthesis, and biological evaluation of novel pleuromutilin derivatives with methicillin-resistant Staphylococcus aureus -targeting phenol linker groups

Methicillin-resistant Staphylococcus aureus (MRSA) remains a significant global health threat, necessitating the development of new therapeutic agents. Pleuromutilin derivatives offer a promising solution due to their potent antibacterial activity, particularly against Gram-positive bacteria such as MRSA. In this study, we synthesized a series of pleuromutilin derivatives with phenol linker containing C14 side chains and evaluated in vitro and in vivo antibacterial activities. Several compounds showed potent activity against MRSA and Staphylococcus aureus with minimal inhibitory concentrations (MICs) as low as 0.03125 μg/mL. In particular, compounds a4 and b4 showed rapid bactericidal activity, significantly reducing MRSA loads in time-kill kinetics and exhibiting slower resistance development compared to tiamulin. In vivo, compound a4 showed superior efficacy in reducing MRSA-induced lung damage in a mouse model at a lower effective dose (ED50 = 6.48 mg/kg) compared to tiamulin (ED50 = 11.38 mg/kg). Molecular docking and molecular dynamics studies also showed that compound a4 binds strongly to the ribosomal peptidyl transferase center (PTC), a key target for pleuromutilin derivatives. These results suggest that compound a4, with its enhanced antibacterial activity and low resistance potential, is a promising candidate for further development as an effective treatment for MRSA infections.

Unraveling the hemolytic toxicity tapestry of peptides using chemical space complex networks

Peptides have emerged as promising therapeutic agents. However, their potential is hindered by hemotoxicity. Understanding the hemotoxicity of peptides is crucial for developing safe and effective peptide-based therapeutics. Here, we employed chemical space complex networks (CSNs) to unravel the hemotoxicity tapestry of peptides. CSNs are powerful tools for visualizing and analyzing the relationships between peptides based on their physicochemical properties and structural features. We constructed CSNs from the StarPepDB database, encompassing 2,004 hemolytic peptides, and explored the impact of seven different (dis)similarity measures on network topology and cluster (communities) distribution. Our findings revealed that each CSN extracts orthogonal information, enhancing the motif discovery and enrichment process. We identified 12 consensus hemolytic motifs, whose amino acid composition unveiled a high abundance of lysine, leucine, and valine residues, whereas aspartic acid, methionine, histidine, asparagine, and glutamine were depleted. Additionally, physicochemical properties were used to characterize clusters/communities of hemolytic peptides. To predict hemolytic activity directly from peptide sequences, we constructed multi-query similarity searching models, which outperformed cutting-edge machine learning-based models, demonstrating robust hemotoxicity prediction capabilities. Overall, this novel in silico approach uses complex network science as its central strategy to develop robust model classifiers, characterize the chemical space, and discover new motifs from hemolytic peptides. This will help to enhance the design/selection of peptides with potential therapeutic activity and low toxicity.

Molecular similarity: Theory, applications, and perspectives

Molecular similarity pervades much of our understanding and rationalization of chemistry. This has become particularly evident in the current data-intensive era of chemical research, with similarity measures serving as the backbone of many Machine Learning (ML) supervised and unsupervised procedures. Here, we present a discussion on the role of molecular similarity in drug design, chemical space exploration, chemical "art" generation, molecular representations, and many more. We also discuss more recent topics in molecular similarity, like the ability to efficiently compare large molecular libraries.

Therapeutic peptide development revolutionized: Harnessing the power of artificial intelligence for drug discovery

Due to the spread of antibiotic resistance, global attention is focused on its inhibition and the expansion of effective medicinal compounds. The novel functional properties of peptides have opened up new horizons in personalized medicine. With artificial intelligence methods combined with therapeutic peptide products, pharmaceuticals and biotechnology advance drug development rapidly and reduce costs. Short-chain peptides inhibit a wide range of pathogens and have great potential for targeting diseases. To address the challenges of synthesis and sustainability, artificial intelligence methods, namely machine learning, must be integrated into their production. Learning methods can use complicated computations to select the active and toxic compounds of the drug and its metabolic activity. Through this comprehensive review, we investigated the artificial intelligence method as a potential tool for finding peptide-based drugs and providing a more accurate analysis of peptides through the introduction of predictable databases for effective selection and development.

Enhanced prediction of hemolytic activity in antimicrobial peptides using deep learning-based sequence analysis

Antimicrobial peptides (AMPs) are a promising class of antimicrobial drugs due to their broad-spectrum activity against microorganisms. However, their clinical application is limited by their potential to cause hemolysis, the destruction of red blood cells. To address this issue, we propose a deep learning model based on convolutional neural networks (CNNs) for predicting the hemolytic activity of AMPs. Peptide sequences are represented using one-hot encoding, and the CNN architecture consists of multiple convolutional and fully connected layers. The model was trained on six different datasets: HemoPI-1, HemoPI-2, HemoPI-3, RNN-Hem, Hlppredfuse, and AMP-Combined, achieving Matthew's correlation coefficients of 0.9274, 0.5614, 0.6051, 0.6142, 0.8799, and 0.7484, respectively. Our model outperforms previously reported methods and can facilitate the development of novel AMPs with reduced hemolytic activity, which is crucial for their therapeutic use in treating bacterial infections.

Deciphering optimal molecular determinants of non-hemolytic, cell-penetrating antimicrobial peptides through bioinformatics and Random Forest

Antimicrobial peptides (AMPs) are promising molecules for combating resistant pathogens, offering several advantages like broad-spectrum effectiveness and multi-targeted action. While most AMPs exhibit membranolytic activity similar to hemolytic peptides (HPs), some act by entering cells like cell-penetrating peptides (CPPs). The toxicity of AMPs towards the host is the major hurdle in their development and application. Given the peptides' function and toxicity largely depend on their molecular properties, identifying and fine-tuning these factors is imperative for developing effective and safe AMPs. To address these knowledge gaps, we present a study that employs a holistic strategy by investigating the molecular descriptors of AMPs, CPPs, HPs, and non-functional equivalents. The prediction of functional properties categorized datasets of 3697 experimentally validated peptides into six groups and three clusters. Predictive and statistical analyses of physicochemical and structural parameters revealed that AMPs have a mean hydrophobic moment of 1.2, a net charge of 4.5, and a lower isoelectric point of 10.9, with balanced hydrophobicity. For cluster AC-nHPs containing peptides with antimicrobial, cell-penetrating, and non-hemolytic properties, disordered conformation and aggregation propensities, followed by amphiphilicity index, isoelectric point, and net charge were identified as the most critical properties. In addition, this work also explains why most AMPs and HPs are membrane-disruptive, while CPPs are non-membranolytic. In conclusion, the study identifies optimal molecular descriptors and offers valuable insights for designing effective, non-toxic AMPs for therapeutic use.

Generative Adversarial Network-Based Augmentation With Noval 2-Step Authentication for Anti-Coronavirus Peptide Prediction

The virus poses a longstanding and enduring danger to various forms of life. Despite the ongoing endeavors to combat viral diseases, there exists a necessity to explore and develop novel therapeutic options. Antiviral peptides are bioactive molecules with a favorable toxicity profile, making them promising alternatives for viral infection treatment. Therefore, this article employed a generative adversarial network for antiviral peptide augmentation and a novel two-step authentication process for augmented synthetic peptides to enhance antiviral activity prediction. Additionally, five widely utilized deep learning models were employed for classification purposes. Initially, a GAN was used to augment the antiviral peptide. In a two-step authentication process, the NCBI-BLAST was utilized to identify the antiviral activity resemblance between the synthetic and real peptide. Subsequently, the hydrophobicity, hydrophilicity, hydroxylic nature, positive charge, and negative charge of synthetic and authentic antiviral peptides were compared before their utilization. Later, to examine the impact of authenticated peptide augmentation in the prediction of antiviral peptides, a comparison is conducted with the outcomes of non-peptide augmented prediction. The study demonstrates that the 1-D convolution neural network with augmented peptide exhibits superior performance compared to other employed classifiers and state-of-the-art models. The network attains a mean classification accuracy of 95.41%, an AUC value of 0.95, and an MCC value of 0.90 on the benchmark antiviral and anti-corona peptides dataset. Thus, the performance of the proposed model indicates its efficacy in predicting the antiviral activity of peptides.

Challenges and opportunities of poly(amino acid) nanomedicines in cancer therapy

Poly(amino acid) nanomedicines hold significant promise for cancer therapy. However, their clinical translation has not matched the extensive efforts of scientists or the burgeoning body of research. The therapeutic outcomes with most nanomedicines often fall short of the promising results observed in animal experiments. This review explores the challenges faced in cancer therapy using poly(amino acid) nanomedicines, particularly addressing the controversies surrounding the enhanced permeability and retention effect and the lack of methods for controlled and reproducible mass production of poly(amino acid) nanomedicines. Furthermore, this review examines the opportunities emerging in this field due to the rapid advancements in artificial intelligence.

Peptide hemolytic activity analysis using visual data mining of similarity-based complex networks

Peptides are promising drug development frameworks that have been hindered by intrinsic undesired properties including hemolytic activity. We aim to get a better insight into the chemical space of hemolytic peptides using a novel approach based on network science and data mining. Metadata networks (METNs) were useful to characterize and find general patterns associated with hemolytic peptides, whereas Half-Space Proximal Networks (HSPNs), represented the hemolytic peptide space. The best candidate HSPNs were used to extract various subsets of hemolytic peptides (scaffolds) considering network centrality and peptide similarity. These scaffolds have been proved to be useful in developing robust similarity-based model classifiers. Finally, using an alignment-free approach, we reported 47 putative hemolytic motifs, which can be used as toxic signatures when developing novel peptide-based drugs. We provided evidence that the number of hemolytic motifs in a sequence might be related to the likelihood of being hemolytic.

Selector of amino-acid scales set

Experimental and theoretical properties of amino acids as building blocks of peptides and proteins have been extensively researched. Each such method assigns a number to each amino acid, and one such assignment is called amino-acid scale. Their usage in bioinformatics to explain and predict behaviour of peptides and proteins is of essential value. The number of such scales is very large. There are more than a hundred scales related just to hydrophobicity. A large number of scales can be a computational burden for algorithms that try to define peptide descriptors combining several of these scales. Hence, it is of interest to construct a smaller, but still representative set of scales. Here, we present software that does this. We test it on the set of scales using a database constructed by Kawashima and collaborators and show that it is possible to significantly reduce the number of scales observed without losing much of the information. An algorithm is implemented in C#. As a result, we provide a smaller database that might be a very useful tool for the analyses and construction of new peptides. Another interesting application of this database would be to compare the artificial intelligence construction of peptides having as an input the complete Kawashima database and this reduced one. Obtaining in both cases similar results would give much credibility to the constructs of such AI algorithms.

Hybrid transformer-CNN model for accurate prediction of peptide hemolytic potential

Hemolysis is a crucial factor in various biomedical and pharmaceutical contexts, driving our interest in developing advanced computational techniques for precise prediction. Our proposed approach takes advantage of the unique capabilities of convolutional neural networks (CNNs) and transformers to detect complex patterns inherent in the data. The integration of CNN and transformers' attention mechanisms allows for the extraction of relevant information, leading to accurate predictions of hemolytic potential. The proposed method was trained on three distinct data sets of peptide sequences known as recurrent neural network-hemolytic (RNN-Hem), Hlppredfuse, and Combined. Our computational results demonstrated the superior efficacy of our models compared to existing methods. The proposed approach demonstrated impressive Matthews correlation coefficients of 0.5962, 0.9111, and 0.7788 respectively, indicating its effectiveness in predicting hemolytic activity. With its potential to guide experimental efforts in peptide design and drug development, this method holds great promise for practical applications. Integrating CNNs and transformers proves to be a powerful tool in the fields of bioinformatics and therapeutic research, highlighting their potential to drive advancement in this area.

Can large language models predict antimicrobial peptide activity and toxicity?

Antimicrobial peptides (AMPs) are naturally occurring or designed peptides up to a few tens of amino acids which may help address the antimicrobial resistance crisis. However, their clinical development is limited by toxicity to human cells, a parameter which is very difficult to control. Given the similarity between peptide sequences and words, large language models (LLMs) might be able to predict AMP activity and toxicity. To test this hypothesis, we fine-tuned LLMs using data from the Database of Antimicrobial Activity and Structure of Peptides (DBAASP). GPT-3 performed well but not reproducibly for activity prediction and hemolysis, taken as a proxy for toxicity. The later GPT-3.5 performed more poorly and was surpassed by recurrent neural networks (RNN) trained on sequence-activity data or support vector machines (SVM) trained on MAP4C molecular fingerprint-activity data. These simpler models are therefore recommended, although the rapid evolution of LLMs warrants future re-evaluation of their prediction abilities.

A novel framework based on explainable AI and genetic algorithms for designing neurological medicines

The advent of the fourth industrial revolution, characterized by artificial intelligence (AI) as its central component, has resulted in the mechanization of numerous previously labor-intensive activities. The use of in silico tools has become prevalent in the design of biopharmaceuticals. Upon conducting a comprehensive analysis of the genomes of many organisms, it has been discovered that their tissues can generate specific peptides that confer protection against certain diseases. This study aims to identify a selected group of neuropeptides (NPs) possessing favorable characteristics that render them ideal for production as neurological biopharmaceuticals. Until now, the construction of NP classifiers has been the primary focus, neglecting to optimize these characteristics. Therefore, in this study, the task of creating ideal NPs has been formulated as a multi-objective optimization problem. The proposed framework, NPpred, comprises two distinct components: NSGA-NeuroPred and BERT-NeuroPred. The former employs the NSGA-II algorithm to explore and change a population of NPs, while the latter is an interpretable deep learning-based model. The utilization of explainable AI and motifs has led to the proposal of two novel operators, namely p-crossover and p-mutation. An online application has been deployed at https://neuropred.anvil.app for designing an ideal collection of synthesizable NPs from protein sequences.

Therapeutic peptides for coronary artery diseases: in silico methods and current perspectives

Many drug formulations containing small active molecules are used for the treatment of coronary artery disease, which affects a significant part of the world's population. However, the inadequate profile of these molecules in terms of therapeutic efficacy has led to the therapeutic use of protein and peptide-based biomolecules with superior properties, such as target-specific affinity and low immunogenicity, in critical diseases. Protein‒protein interactions, as a consequence of advances in molecular techniques with strategies involving the combined use of in silico methods, have enabled the design of therapeutic peptides to reach an advanced dimension. In particular, with the advantages provided by protein/peptide structural modeling, molecular docking for the study of their interactions, molecular dynamics simulations for their interactions under physiological conditions and machine learning techniques that can work in combination with all these, significant progress has been made in approaches to developing therapeutic peptides that can modulate the development and progression of coronary artery diseases. In this scope, this review discusses in silico methods for the development of peptide therapeutics for the treatment of coronary artery disease and strategies for identifying the molecular mechanisms that can be modulated by these designs and provides a comprehensive perspective for future studies.

Structure-aware machine learning strategies for antimicrobial peptide discovery

Machine learning models are revolutionizing our approaches to discovering and designing bioactive peptides. These models often need protein structure awareness, as they heavily rely on sequential data. The models excel at identifying sequences of a particular biological nature or activity, but they frequently fail to comprehend their intricate mechanism(s) of action. To solve two problems at once, we studied the mechanisms of action and structural landscape of antimicrobial peptides as (i) membrane-disrupting peptides, (ii) membrane-penetrating peptides, and (iii) protein-binding peptides. By analyzing critical features such as dipeptides and physicochemical descriptors, we developed models with high accuracy (86-88%) in predicting these categories. However, our initial models (1.0 and 2.0) exhibited a bias towards α-helical and coiled structures, influencing predictions. To address this structural bias, we implemented subset selection and data reduction strategies. The former gave three structure-specific models for peptides likely to fold into α-helices (models 1.1 and 2.1), coils (1.3 and 2.3), or mixed structures (1.4 and 2.4). The latter depleted over-represented structures, leading to structure-agnostic predictors 1.5 and 2.5. Additionally, our research highlights the sensitivity of important features to different structure classes across models.

Machine learning for antimicrobial peptide identification and design

Artificial intelligence (AI) and machine learning (ML) models are being deployed in many domains of society and have recently reached the field of drug discovery. Given the increasing prevalence of antimicrobial resistance, as well as the challenges intrinsic to antibiotic development, there is an urgent need to accelerate the design of new antimicrobial therapies. Antimicrobial peptides (AMPs) are therapeutic agents for treating bacterial infections, but their translation into the clinic has been slow owing to toxicity, poor stability, limited cellular penetration and high cost, among other issues. Recent advances in AI and ML have led to breakthroughs in our abilities to predict biomolecular properties and structures and to generate new molecules. The ML-based modelling of peptides may overcome some of the disadvantages associated with traditional drug discovery and aid the rapid development and translation of AMPs. Here, we provide an introduction to this emerging field and survey ML approaches that can be used to address issues currently hindering AMP development. We also outline important limitations that can be addressed for the broader adoption of AMPs in clinical practice, as well as new opportunities in data-driven peptide design.

Multiquery Similarity Searching Models: An Alternative Approach for Predicting Hemolytic Activity from Peptide Sequence

The desirable pharmacological properties and a broad number of therapeutic activities have made peptides promising drugs over small organic molecules and antibody drugs. Nevertheless, toxic effects, such as hemolysis, have hampered the development of such promising drugs. Hence, a reliable computational tool to predict peptide hemolytic toxicity is enormously useful before synthesis and experimental evaluation. Currently, four web servers that predict hemolytic activity using machine learning (ML) algorithms are available; however, they exhibit some limitations, such as the need for a reliable negative set and limited application domain. Hence, we developed a robust model based on a novel theoretical approach that combines network science and a multiquery similarity searching (MQSS) method. A total of 1152 initial models were constructed from 144 scaffolds generated in a previous report. These were evaluated on external data sets, and the best models were fused and improved. Our best MQSS model I1 outperformed all state-of-the-art ML-based models and was used to characterize the prevalence of hemolytic toxicity on therapeutic peptides. Based on our model's estimation, the number of hemolytic peptides might be 3.9-fold higher than the reported.

Deconvoluting low yield from weak potency in direct-to-biology workflows with machine learning

High throughput and rapid biological evaluation of small molecules is an essential factor in drug discovery and development. Direct-to-biology (D2B), whereby compound purification is foregone, has emerged as a viable technique in time efficient screening, specifically for PROTAC design and biological evaluation. However, one notable limitation is the prerequisite of high yielding reactions to ensure the desired compound is indeed the compound responsible for biological activity. Herein, we report a machine learning based yield-assay deconfounder capable of deconvoluting low yield from low potency to identify false negatives. We validated this approach by identifying promising SARS-CoV-2 main protease inhibitors with nanomolar activity that rivaled potency observed from the standard D2B workflow. Furthermore, we show how our framework can be utilized in a broad, in silico screen to produce compounds of similar potency as a D2B assay.

Protein-based nanoparticles for therapeutic nucleic acid delivery

To realize the full potential of emerging nucleic acid therapies, there is a need for effective delivery agents to transport cargo to cells of interest. Protein materials exhibit several unique properties, including biodegradability, biocompatibility, ease of functionalization via recombinant and chemical modifications, among other features, which establish a promising basis for therapeutic nucleic acid delivery systems. In this review, we highlight progress made in the use of non-viral protein-based nanoparticles for nucleic acid delivery in vitro and in vivo, while elaborating on key physicochemical properties that have enabled the use of these materials for nanoparticle formulation and drug delivery. To conclude, we comment on the prospects and unresolved challenges associated with the translation of protein-based nucleic acid delivery systems for therapeutic applications.

Engraulisin: A novel marine derived cell penetrating peptide with activity against drug resistant bacteria

Cell penetrating peptides (CPP) with their intrinsic ability to penetrate plasma membranes facilitate intracellular uptake of various macromolecules. Although a substantial number of CPPs have been reported over the last three decades, the number is still inadequate when compared to the theoretically feasible peptides with similar physicochemical composition. Marine organisms, due to their hostile environment, are an immense source of several high-valued therapeutically relevant peptides. Various marine derived antibacterial, antimycotic and anticancer peptides have demonstrated improved activity in comparison to peptides of terrestrial origin. While a significant number of marine bioactive peptides exist, cell penetrating peptides from marine organisms remain unravelled. In this study, we report Engraulisin from Engraulis japonicus, a computationally derived novel cell penetrating peptide of marine origin. Engraulisin manifest successful uptake in mammalian cells at 5 μM concentration with negligible cytotoxicity observed through MTT assay. Analysis of its cellular uptake mechanism revealed significant inhibition at 4 °C suggesting endocytosis as the major route of cellular entry. Interestingly, the novel peptide also demonstrated selective antimicrobial activity against Methicillin-resistant Staphylococcus aureus (MRSA). Additionally, molecular dynamics simulation with POPC and POPG bilayer system unveiled significance of positively charged residues in forming a stable membrane interaction. Engraulisin represents a novel marine-derived cell penetrating peptide which can be explored for cellular delivery of pharmaceutically relevant molecules.

Prediction of the synergistic effect of antimicrobial peptides and antimicrobial agents via supervised machine learning

BACKGROUND

Infectious diseases not only cause severe health problems but also burden the healthcare system. Therefore, the effective treatment of those diseases is crucial. Both conventional approaches, such as antimicrobial agents, and novel approaches, like antimicrobial peptides (AMPs), are used to treat infections. However, due to the drawbacks of current approaches, new solutions are still being investigated. One recent approach is the use of AMPs and antimicrobial agents in combination, but determining synergism is with a huge variety of AMPs time-consuming and requires multiple experimental studies. Machine learning (ML) algorithms are widely used to predict biological outcomes, particularly in the field of AMPs, but no previous research reported on predicting the synergistic effects of AMPs and antimicrobial agents.

RESULTS

Several supervised ML models were implemented to accurately predict the synergistic effect of AMPs and antimicrobial agents. The results demonstrated that the hyperparameter-optimized Light Gradient Boosted Machine Classifier (oLGBMC) yielded the best test accuracy of 76.92% for predicting the synergistic effect. Besides, the feature importance analysis reveals that the target microbial species, the minimum inhibitory concentrations (MICs) of the AMP and the antimicrobial agents, and the used antimicrobial agent were the most important features for the prediction of synergistic effect, which aligns with recent experimental studies in the literature.

CONCLUSION

This study reveals that ML algorithms can predict the synergistic activity of two different antimicrobial agents without the need for complex and time-consuming experimental procedures. The implications support that the ML models may not only reduce the experimental cost but also provide validation of experimental procedures.

Machine learning assisted rational design of antimicrobial peptides based on human endogenous proteins and their applications for cosmetic preservative system optimization

Preservatives are essential components in cosmetic products, but their safety issues have attracted widespread attention. There is an urgent need for safe and effective alternatives. Antimicrobial peptides (AMPs) are part of the innate immune system and have potent antimicrobial properties. Using machine learning-assisted rational design, we obtained a novel antibacterial peptide, IK-16-1, with significant antibacterial activity and maintaining safety based on β-defensins. IK-16-1 has broad-spectrum antimicrobial properties against Escherichia coli, Staphylococcus aureus, Pseudomonas aeruginosa, and Candida albicans, and has no haemolytic activity. The use of IK-16-1 holds promise in the cosmetics industry, since it can serve as a preservative synergist to reduce the amount of other preservatives in cosmetics. This study verified the feasibility of combining computational design with artificial intelligence prediction to design AMPs, achieving rapid screening and reducing development costs.

Accelerating the Discovery and Design of Antimicrobial Peptides with Artificial Intelligence

Peptides modulate many processes of human physiology targeting ion channels, protein receptors, or enzymes. They represent valuable starting points for the development of new biologics against communicable and non-communicable disorders. However, turning native peptide ligands into druggable materials requires high selectivity and efficacy, predictable metabolism, and good safety profiles. Machine learning models have gradually emerged as cost-effective and time-saving solutions to predict and generate new proteins with optimal properties. In this chapter, we will discuss the evolution and applications of predictive modeling and generative modeling to discover and design safe and effective antimicrobial peptides. We will also present their current limitations and suggest future research directions, applicable to peptide drug design campaigns.

Efficient Computational Framework for Target-Specific Active Peptide Discovery: A Case Study on IL-17C Targeting Cyclic Peptides

The development of potentially active peptides for specific targets is critical for the modern pharmaceutical industry's growth. In this study, we present an efficient computational framework for the discovery of active peptides targeting a specific pharmacological target, which combines a conditional variational autoencoder (CVAE) and a classifier named TCPP based on the Transformer and convolutional neural network. In our example scenario, we constructed an active cyclic peptide library targeting interleukin-17C (IL-17C) through a library-based in vitro selection strategy. The CVAE model is trained on the preprocessed peptide data sets to generate potentially active peptides and the TCPP further screens the generated peptides. Ultimately, six candidate peptides predicted by the model were synthesized and assayed for their activity, and four of them exhibited promising binding affinity to IL-17C. Our study provides a one-stop-shop for target-specific active peptide discovery, which is expected to boost up the process of peptide drug development.

Proteins released into the plant apoplast by the obligate parasitic protist Albugo selectively repress phyllosphere-associated bacteria

Biotic and abiotic interactions shape natural microbial communities. The mechanisms behind microbe-microbe interactions, particularly those protein based, are not well understood. We hypothesize that released proteins with antimicrobial activity are a powerful and highly specific toolset to shape and defend plant niches. We have studied Albugo candida, an obligate plant parasite from the protist Oomycota phylum, for its potential to modulate the growth of bacteria through release of antimicrobial proteins into the apoplast. Amplicon sequencing and network analysis of Albugo-infected and uninfected wild Arabidopsis thaliana samples revealed an abundance of negative correlations between Albugo and other phyllosphere microbes. Analysis of the apoplastic proteome of Albugo-colonized leaves combined with machine learning predictors enabled the selection of antimicrobial candidates for heterologous expression and study of their inhibitory function. We found for three candidate proteins selective antimicrobial activity against Gram-positive bacteria isolated from A. thaliana and demonstrate that these inhibited bacteria are precisely important for the stability of the community structure. We could ascribe the antibacterial activity of the candidates to intrinsically disordered regions and positively correlate it with their net charge. This is the first report of protist proteins with antimicrobial activity under apoplastic conditions that therefore are potential biocontrol tools for targeted manipulation of the microbiome.

Accelerating the Screening of Small Peptide Ligands by Combining Peptide-Protein Docking and Machine Learning

This research introduces a novel pipeline that couples machine learning (ML), and molecular docking for accelerating the process of small peptide ligand screening through the prediction of peptide-protein docking. Eight ML algorithms were analyzed for their potential. Notably, Light Gradient Boosting Machine (LightGBM), despite having comparable F1-score and accuracy to its counterparts, showcased superior computational efficiency. LightGBM was used to classify peptide-protein docking performance of the entire tetrapeptide library of 160,000 peptide ligands against four viral envelope proteins. The library was classified into two groups, 'better performers' and 'worse performers'. By training the LightGBM algorithm on just 1% of the tetrapeptide library, we successfully classified the remaining 99%with an accuracy range of 0.81-0.85 and an F1-score between 0.58-0.67. Three different molecular docking software were used to prove that the process is not software dependent. With an adjustable probability threshold (from 0.5 to 0.95), the process could be accelerated by a factor of at least 10-fold and still get 90-95% concurrence with the method without ML. This study validates the efficiency of machine learning coupled to molecular docking in rapidly identifying top peptides without relying on high-performance computing power, making it an effective tool for screening potential bioactive compounds.

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

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.

Synthetic extracellular matrices with function-encoding peptides

The communication of cells with their surroundings is mostly encoded in the epitopes of structural and signalling proteins present in the extracellular matrix (ECM). These peptide epitopes can be incorporated in biomaterials to serve as function-encoding molecules to modulate cell-cell and cell-ECM interactions. In this Review, we discuss natural and synthetic peptide epitopes as molecular tools to bioengineer bioactive hydrogel materials. We present a library of functional peptide sequences that selectively communicate with cells and the ECM to coordinate biological processes, including epitopes that directly signal to cells, that bind ECM components that subsequently signal to cells, and that regulate ECM turnover. We highlight how these epitopes can be incorporated in different biomaterials as individual or multiple signals, working synergistically or additively. This molecular toolbox can be applied in the design of biomaterials aimed at regulating or controlling cellular and tissue function, repair and regeneration.

Addressing antibiotic resistance: computational answers to a biological problem?

The increasing prevalence of infections caused by antibiotic-resistant bacteria is a global healthcare crisis. Understanding the spread of resistance is predicated on the surveillance of antibiotic resistance genes within an environment. Bioinformatics and artificial intelligence (AI) methods applied to metagenomic sequencing data offer the capacity to detect known and infer yet-unknown resistance mechanisms, and predict future outbreaks of antibiotic-resistant infections. Machine learning methods, in particular, could revive the waning antibiotic discovery pipeline by helping to predict the molecular structure and function of antibiotic resistance compounds, and optimising their interactions with target proteins. Consequently, AI has the capacity to play a central role in guiding antibiotic stewardship and future clinical decision-making around antibiotic resistance.

Discovering highly potent antimicrobial peptides with deep generative model HydrAMP

Antimicrobial peptides emerge as compounds that can alleviate the global health hazard of antimicrobial resistance, prompting a need for novel computational approaches to peptide generation. Here, we propose HydrAMP, a conditional variational autoencoder that learns lower-dimensional, continuous representation of peptides and captures their antimicrobial properties. The model disentangles the learnt representation of a peptide from its antimicrobial conditions and leverages parameter-controlled creativity. HydrAMP is the first model that is directly optimized for diverse tasks, including unconstrained and analogue generation and outperforms other approaches in these tasks. An additional preselection procedure based on ranking of generated peptides and molecular dynamics simulations increases experimental validation rate. Wet-lab experiments on five bacterial strains confirm high activity of nine peptides generated as analogues of clinically relevant prototypes, as well as six analogues of an inactive peptide. HydrAMP enables generation of diverse and potent peptides, making a step towards resolving the antimicrobial resistance crisis.

Thinking on the Construction of Antimicrobial Peptide Databases: Powerful Tools for the Molecular Design and Screening

With the accelerating growth of antimicrobial resistance (AMR), there is an urgent need for new antimicrobial agents with low or no AMR. Antimicrobial peptides (AMPs) have been extensively studied as alternatives to antibiotics (ATAs). Coupled with the new generation of high-throughput technology for AMP mining, the number of derivatives has increased dramatically, but manual running is time-consuming and laborious. Therefore, it is necessary to establish databases that combine computer algorithms to summarize, analyze, and design new AMPs. A number of AMP databases have already been established, such as the Antimicrobial Peptides Database (APD), the Collection of Antimicrobial Peptides (CAMP), the Database of Antimicrobial Activity and Structure of Peptides (DBAASP), and the Database of Antimicrobial Peptides (dbAMPs). These four AMP databases are comprehensive and are widely used. This review aims to cover the construction, evolution, characteristic function, prediction, and design of these four AMP databases. It also offers ideas for the improvement and application of these databases based on merging the various advantages of these four peptide libraries. This review promotes research and development into new AMPs and lays their foundation in the fields of druggability and clinical precision treatment.

Machine learning and molecular simulation ascertain antimicrobial peptide against Klebsiella pneumoniae from public database

Antimicrobial peptides (AMPs) are short peptides with a broad spectrum of antimicrobial activity. They play a key role in the host innate immunity of many organisms. The growing threat of microorganisms resistant to antimicrobial agents and the lack of new commercially available antibiotics have made in silico discovery of AMPs increasingly important. Machine learning (ML) has improved the speed and efficiency of AMP discovery while reducing the cost of experimental approaches. Despite various ML platforms developed, there is still a lack of integrative use of ML platforms for AMP discovery from publicly available protein databases. Therefore, our study aims to screen potential AMPs with antibiofilm properties from databases using ML platforms, followed by protein-peptide molecular docking analysis and molecular dynamics (MD) simulations. A total of 5850 peptides classified as non-AMP were screened from UniProtKB and analyzed using various online ML platforms (e.g., CAMPr3, DBAASP, dPABBs, Hemopred, and ToxinPred). Eight potential AMP peptides against Klebsiella pneumoniae with antibiofilm, non-toxic and non-hemolytic properties were then docked to MrkH, a transcriptional regulator of type 3 fimbriae involved in biofilm formation. Five of eight peptides bound more strongly than the native MrkH ligand when analyzed using HADDOCK and HPEPDOCK. Following the docking studies, our MD simulated that a Neuropeptide B (Peptide 3) bind strongly to the MrkH active sites. The discovery of putative AMPs that exceed the binding energies of the native ligand underscores the utility of the combined ML and molecular simulation strategies for discovering novel AMPs with antibiofilm properties.

Integration of Machine Learning and Structural Analysis for Predicting Peptide Antibiofilm Effects: Advancements in Drug Discovery for Biofilm-Related Infections

Background

The rise of antibiotic resistance has become a major concern, signaling the end of the golden age of antibiotics. Bacterial biofilms, which exhibit high resistance to antibiotics, significantly contribute to the emergence of antibiotic resistance. Therefore, there is an urgent need to discover new therapeutic agents with specific characteristics to effectively combat biofilm-related infections. Studies have shown the promising potential of peptides as antimicrobial agents.

Objectives

This study aimed to establish a cost-effective and streamlined computational method for predicting the antibiofilm effects of peptides. This method can assist in addressing the intricate challenge of designing peptides with strong antibiofilm properties, a task that can be both challenging and costly.

Methods

A positive library, consisting of peptide sequences with antibiofilm activity exceeding 50%, was assembled, along with a negative library containing quorum-sensing peptides. For each peptide sequence, feature vectors were calculated, while considering the primary structure, the order of amino acids, their physicochemical properties, and their distributions. Multiple supervised learning algorithms were used to classify peptides with significant antibiofilm effects for subsequent experimental evaluations.

Results

The computational approach exhibited high accuracy in predicting the antibiofilm effects of peptides, with accuracy, precision, Matthew's correlation coefficient (MCC), and F1 score of 99%, 99%, 0.97, and 0.99, respectively. The performance level of this computational approach was comparable to that of previous methods. This study introduced a novel approach by combining the feature space with high antibiofilm activity.

Conclusions

In this study, a reliable and cost-effective method was developed for predicting the antibiofilm effects of peptides using a computational approach. This approach allows for the identification of peptide sequences with substantial antibiofilm activities for further experimental investigations. Accessible source codes and raw data of this study can be found online (hiABF), providing easy access and enabling future updates.

Hemolytic-Pred: A machine learning-based predictor for hemolytic proteins using position and composition-based features

Objective

The objective of this study is to propose a novel in-silico method called Hemolytic-Pred for identifying hemolytic proteins based on their sequences, using statistical moment-based features, along with position-relative and frequency-relative information.

Methods

Primary sequences were transformed into feature vectors using statistical and position-relative moment-based features. Varying machine learning algorithms were employed for classification. Computational models were rigorously evaluated using four different validation. The Hemolytic-Pred webserver is available for further analysis at http://ec2-54-160-229-10.compute-1.amazonaws.com/.

Results

XGBoost outperformed the other six classifiers with an accuracy value of 0.99, 0.98, 0.97, and 0.98 for self-consistency test, 10-fold cross-validation, Jackknife test, and independent set test, respectively. The proposed method with the XGBoost classifier is a workable and robust solution for predicting hemolytic proteins efficiently and accurately.

Conclusions

The proposed method of Hemolytic-Pred with XGBoost classifier is a reliable tool for the timely identification of hemolytic cells and diagnosis of various related severe disorders. The application of Hemolytic-Pred can yield profound benefits in the medical field.

Peptide Utility (PU) search server: A new tool for peptide sequence search from multiple databases

Proteins are essential building blocks in humans that have garnered huge attention from researchers worldwide due to their numerous therapeutic applications. To date, different computational tools have been developed to extract pre-existing information on these biological molecules, but most of these tools suffer from limitations such as non-user friendly interface, redundancy of data, etc. To overcome these limitations, a user-friendly interface, the Peptide Utility (PU) webserver (https://chain-searching.herokuapp.com/) has been developed for searching and analyzing homologous and identical protein/peptide sequences that can be searched from approximately 0.4 million sequences (structural and sequence information) in both online and offline modes. The PU web server can also be used to study different types of interactions in PDBSum, identifying the most dominating interface residues, the most prevalent interactions, and the interaction preferences of different residues. The webserver would also pave way for the design of novel therapeutic peptides and folds by identifying conserved residues in the three-dimensional structure space of proteins.

Chemoinformatics and artificial intelligence colloquium: progress and challenges in developing bioactive compounds

We report the main conclusions of the first Chemoinformatics and Artificial Intelligence Colloquium, Mexico City, June 15-17, 2022. Fifteen lectures were presented during a virtual public event with speakers from industry, academia, and non-for-profit organizations. Twelve hundred and ninety students and academics from more than 60 countries. During the meeting, applications, challenges, and opportunities in drug discovery, de novo drug design, ADME-Tox (absorption, distribution, metabolism, excretion and toxicity) property predictions, organic chemistry, peptides, and antibiotic resistance were discussed. The program along with the recordings of all sessions are freely available at https://www.difacquim.com/english/events/2022-colloquium/ .

Pulled fly balls are harder to catch: a game analysis with a machine learning approach

Two hypotheses were tested: (1) the deflecting motion of fly balls caused by aerodynamic effects varies between the pull side and opposite side of the fair territory, and (2) the probability of flyout is lower on the pull side than the opposite side in Japan’s professional baseball games. From all radar-tracking outputs of official games in 2018–2019, fly balls that resulted in outs or base hits were selected for analysis (N = 25,413), and indices representing horizontal and vertical deflecting motions of fly balls were computed and compared between the pull side and opposite side. A machine learning algorithm was used to construct a model to predict the probability of flyout from the kinematic characteristics of fly balls. Flyout zones where the probability of flyout was > 0.6 were computed for a systematically constructed set of fly balls having identical distribution between the pull side and opposite side. The results showed that: (1) most fly balls landing on the opposite side deflected in the same direction whereas the pulled fly balls deflected to either direction, (2) the pulled low fly balls had greater variability in the deflecting motions than the opposite side counterpart, (3) overall probability of flyout of the low fly balls was lower in the pull side (0.41) than the opposite side (0.49), and (4) the flyout zone of an outfielder in the pull side (mean = 698 m2) for low fly balls was smaller than that of the others (≥ 779 m2). The hypotheses were supported. The pulled low fly balls had substantial variations in the direction and magnitude of deflections, which might have reduced the flyout zone on the pull side.

Development of Peptide-based Metallo-β-lactamase Inhibitors as a New Strategy to Combat Antimicrobial Resistance: A Mini-review

Global dissemination of antimicrobial resistance (AMR) not only poses a significant threat to human health, food security, and social development but also results in millions of deaths each year. In Gram-negative bacteria, the primary mechanism of resistance to β-lactam antibiotics is the production of β-lactamases, one of which is carbapenem-hydrolyzing β-lactamases known as carbapenemases. As a general scheme, these enzymes are divided into Ambler class A, B, C, and D based on their protein sequence homology. Class B β-lactamases are also known as metallo-β-lactamases (MBLs). The incidence of recovery of bacteria expressing metallo-β- lactamases (MBLs) has increased dramatically in recent years, almost reaching a pandemic proportion. MBLs can be further divided into three subclasses (B1, B2, and B3) based on the homology of protein sequences as well as the differences in zinc coordination. The development of inhibitors is one effective strategy to suppress the activities of MBLs and restore the activity of β-lactam antibiotics. Although thousands of MBL inhibitors have been reported, none have been approved for clinical use. This review describes the clinical application potential of peptide-based drugs that exhibit inhibitory activity against MBLs identified in past decades. In this report, peptide-based inhibitors of MBLs are divided into several groups based on the mode of action, highlighting compounds of promising properties that are suitable for further advancement. We discuss how traditional computational tools, such as in silico screening and molecular docking, along with new methods, such as deep learning and machine learning, enable a more accurate and efficient design of peptide-based inhibitors of MBLs.

Function Prediction of Peptide Toxins with Sequence-Based Multi-Tasking PU Learning Method

Peptide toxins generally have extreme pharmacological activities and provide a rich source for the discovery of drug leads. However, determining the optimal activity of a new peptide can be a long and expensive process. In this study, peptide toxins were retrieved from Uniprot; three positive-unlabeled (PU) learning schemes, adaptive basis classifier, two-step method, and PU bagging were adopted to develop models for predicting the biological function of new peptide toxins. All three schemes were embedded with 14 machine learning classifiers. The prediction results of the adaptive base classifier and the two-step method were highly consistent. The models with top comprehensive performances were further optimized by feature selection and hyperparameter tuning, and the models were validated by making predictions for 61 three-finger toxins or the external HemoPI dataset. Biological functions that can be identified by these models include cardiotoxicity, vasoactivity, lipid binding, hemolysis, neurotoxicity, postsynaptic neurotoxicity, hypotension, and cytolysis, with relatively weak predictions for hemostasis and presynaptic neurotoxicity. These models are discovery-prediction tools for active peptide toxins and are expected to accelerate the development of peptide toxins as drugs.

Machine Learning Models to Predict Protein-Protein Interaction Inhibitors

Protein-protein interaction (PPI) inhibitors have an increasing role in drug discovery. It is hypothesized that machine learning (ML) algorithms can classify or identify PPI inhibitors. This work describes the performance of different algorithms and molecular fingerprints used in chemoinformatics to develop a classification model to identify PPI inhibitors making the codes freely available to the community, particularly the medicinal chemistry research groups working with PPI inhibitors. We found that classification algorithms have different performances according to various features employed in the training process. Random forest (RF) models with the extended connectivity fingerprint radius 2 (ECFP4) had the best classification abilities compared to those models trained with ECFP6 o MACCS keys (166-bits). In general, logistic regression (LR) models had lower performance metrics than RF models, but ECFP4 was the representation most appropriate for LR. ECFP4 also generated models with high-performance metrics with support vector machines (SVM). We also constructed ensemble models based on the top-performing models. As part of this work and to help non-computational experts, we developed a pipeline code freely available.

High-Throughput Prediction and Design of Novel Conopeptides for Biomedical Research and Development

Cone snail venoms have been considered a valuable treasure for international scientists and businessmen, mainly due to their pharmacological applications in development of marine drugs for treatment of various human diseases. To date, around 800 Conus species are recorded, and each of them produces over 1,000 venom peptides (termed as conopeptides or conotoxins). This reflects the high diversity and complexity of cone snails, although most of their venoms are still uncharacterized. Advanced multiomics (such as genomics, transcriptomics, and proteomics) approaches have been recently developed to mine diverse Conus venom samples, with the main aim to predict and identify potentially interesting conopeptides in an efficient way. Some bioinformatics techniques have been applied to predict and design novel conopeptide sequences, related targets, and their binding modes. This review provides an overview of current knowledge on the high diversity of conopeptides and multiomics advances in high-throughput prediction of novel conopeptide sequences, as well as molecular modeling and design of potential drugs based on the predicted or validated interactions between these toxins and their molecular targets.

IntegralVac: A Machine Learning-Based Comprehensive Multivalent Epitope Vaccine Design Method

In the growing field of vaccine design for COVID and cancer research, it is essential to predict accurate peptide binding affinity and immunogenicity. We developed a comprehensive machine learning method, ‘IntegralVac,’ by integrating three existing deep learning tools: DeepVacPred, MHCSeqNet, and HemoPI. IntegralVac makes predictions for single and multivalent cancer and COVID-19 epitopes without manually selecting epitope prediction possibilities. We performed several rounds of optimization before integration, then re-trained IntegralVac for multiple datasets. We validated the IntegralVac with 4500 human cancer MHC I peptides obtained from the Immune Epitope Database (IEDB) and with cancer and COVID epitopes previously selected in our laboratory. The other data referenced from existing deep learning tools served as a positive control to ensure successful prediction was possible. As evidenced by increased accuracy and AUC, IntegralVac improved the prediction rate of top-ranked epitopes. We also examined the compatibility between other servers’ clinical checkpoint filters and IntegralVac. This was to ensure that the other servers had a means for predicting additional checkpoint filters that we wanted to implement in IntegralVac. The clinical checkpoint filters, including allergenicity, antigenicity, and toxicity, were used as additional predictors to improve IntegralVac’s prediction accuracy. We generated immunogenicity scores by cross-comparing sequence inputs with each other and determining the overlap between each individual peptide sequence. The IntegralVac increased the immunogenicity prediction accuracy to 90.1% AUC and the binding affinity accuracy to 95.4% compared to the control NetMHCPan server. The IntegralVac opens new avenues for future in silico methods, by building upon established models for continued prediction accuracy improvement.

AMPDeep: hemolytic activity prediction of antimicrobial peptides using transfer learning

Background

Deep learning’s automatic feature extraction has proven to give superior performance in many sequence classification tasks. However, deep learning models generally require a massive amount of data to train, which in the case of Hemolytic Activity Prediction of Antimicrobial Peptides creates a challenge due to the small amount of available data.

Results

Three different datasets for hemolysis activity prediction of therapeutic and antimicrobial peptides are gathered and the AMPDeep pipeline is implemented for each. The result demonstrate that AMPDeep outperforms the previous works on all three datasets, including works that use physicochemical features to represent the peptides or those who solely rely on the sequence and use deep learning to learn representation for the peptides. Moreover, a combined dataset is introduced for hemolytic activity prediction to address the problem of sequence similarity in this domain. AMPDeep fine-tunes a large transformer based model on a small amount of peptides and successfully leverages the patterns learned from other protein and peptide databases to assist hemolysis activity prediction modeling.

Conclusions

In this work transfer learning is leveraged to overcome the challenge of small data and a deep learning based model is successfully adopted for hemolysis activity classification of antimicrobial peptides. This model is first initialized as a protein language model which is pre-trained on masked amino acid prediction on many unlabeled protein sequences in a self-supervised manner. Having done so, the model is fine-tuned on an aggregated dataset of labeled peptides in a supervised manner to predict secretion. Through transfer learning, hyper-parameter optimization and selective fine-tuning, AMPDeep is able to achieve state-of-the-art performance on three hemolysis datasets using only the sequence of the peptides. This work assists the adoption of large sequence-based models for peptide classification and modeling tasks in a practical manner.

Machine Learning Guided Discovery of Non-Hemolytic Membrane Disruptive Anticancer Peptides

Most antimicrobial peptides (AMPs) and anticancer peptides (ACPs) fold into membrane disruptive cationic amphiphilic α-helices, many of which are however also unpredictably hemolytic and toxic. Here we exploited the ability of recurrent neural networks (RNN) to distinguish active from inactive and non-hemolytic from hemolytic AMPs and ACPs to discover new non-hemolytic ACPs. Our discovery pipeline involved: 1) sequence generation using either a generative RNN or a genetic algorithm, 2) RNN classification for activity and hemolysis, 3) selection for sequence novelty, helicity and amphiphilicity, and 4) synthesis and testing. Experimental evaluation of thirty-three peptides resulted in eleven active ACPs, four of which were non-hemolytic, with properties resembling those of the natural ACP lasioglossin III. These experiments show the first example of direct machine learning guided discovery of non-hemolytic ACPs.

Profiling hymenopteran venom toxins: Protein families, structural landscape, biological activities, and pharmacological benefits

Hymenopterans are an untapped source of venom secretions. Their recent proteo-transcriptomic studies have revealed an extraordinary pool of toxins that participate in various biological processes, including pain, paralysis, allergic reactions, and antimicrobial activities. Comprehensive and clade-specific campaigns to collect hymenopteran venoms are therefore needed. We consider that data-driven bioprospecting may help prioritise sampling and alleviate associated costs. This work established the current protein landscape from hymenopteran venoms to evaluate possible sample bias by studying their origins, sequence diversity, known structures, and biological functions. We collected all 282 reported hymenopteran toxins (peptides and proteins) from the UniProt database that we clustered into 21 protein families from the three studied clades - wasps, bees, and ants. We identified 119 biological targets of hymenopteran toxins ranging from pathogen membranes to eukaryotic proteases, ion channels and protein receptors. Our systematic study further extended to hymenopteran toxins' therapeutic and biotechnological values, where we revealed promising applications in crop pests, human infections, autoimmune diseases, and neurodegenerative disorders.

•

The hymenopteran toxin diversity includes 21 protein families from 81 species.

•

Some toxins are shared across wasps, bees and ants, others are clade-specific.

•

Their venoms contain membrane-active peptides, neurotoxins, allergens and enzymes.

•

Hymenopteran toxins have been tested against a total of 119 biological targets.

•

Hymenopteran toxins were predominantly evaluated as anti-infective agents.

Improved Database Filtering Technology Enables More Efficient Ab Initio Design of Potent Peptides against Ebola Viruses

The rapid mutations of viruses such as SARS-CoV-2 require vaccine updates and the development of novel antiviral drugs. This article presents an improved database filtering technology for a more effective design of novel antiviral agents. Different from the previous approach, where the most probable parameters were obtained stepwise from the antimicrobial peptide database, we found it possible to accelerate the design process by deriving multiple parameters in a single step during the peptide amino acid analysis. The resulting peptide DFTavP1 displays the ability to inhibit Ebola virus. A deviation from the most probable peptide parameters reduces antiviral activity. The designed peptides appear to block viral entry. In addition, the amino acid signature provides a clue to peptide engineering to gain cell selectivity. Like human cathelicidin LL-37, our engineered peptide DDIP1 inhibits both Ebola and SARS-CoV-2 viruses. These peptides, with broad antiviral activity, may selectively disrupt viral envelopes and offer the lasting efficacy required to treat various RNA viruses, including their emerging mutants.