Inductive QSAR Descriptors. Distinguishing Compounds with Antibacterial Activity by Artificial Neural Networks

Optimizing Neural Networks for Chemical Reaction Prediction: Insights from Methylene Blue Reduction Reactions

This paper offers a thorough investigation of hyperparameter tuning for neural network architectures using datasets encompassing various combinations of Methylene Blue (MB) Reduction by Ascorbic Acid (AA) reactions with different solvents and concentrations. The aim is to predict coefficients of decay plots for MB absorbance, shedding light on the complex dynamics of chemical reactions. Our findings reveal that the optimal model, determined through our investigation, consists of five hidden layers, each with sixteen neurons and employing the Swish activation function. This model yields an NMSE of 0.05, 0.03, and 0.04 for predicting the coefficients A, B, and C, respectively, in the exponential decay equation A + B · e-x/C. These findings contribute to the realm of drug design based on machine learning, providing valuable insights into optimizing chemical reaction predictions.

Predicting Dose-Range Chemical Toxicity using Novel Hybrid Deep Machine-Learning Method

Humans are exposed to thousands of chemicals, including environmental chemicals. Unfortunately, little is known about their potential toxicity, as determining the toxicity remains challenging due to the substantial resources required to assess a chemical in vivo. Here, we present a novel hybrid neural network (HNN) deep learning method, called HNN-Tox, to predict chemical toxicity at different doses. To develop a hybrid HNN-Tox method, we combined two neural network frameworks, the Convolutional Neural Network (CNN) and the multilayer perceptron (MLP)-type feed-forward neural network (FFNN). Combining the CNN and FCNN in the field of environmental chemical toxicity prediction is a novel approach. We developed several binary and multiclass classification models to assess dose-range chemical toxicity that is trained based on thousands of chemicals with known toxicity. The performance of the HNN-Tox was compared with other machine-learning methods, including Random Forest (RF), Bootstrap Aggregation (Bagging), and Adaptive Boosting (AdaBoost). We also analyzed the model performance dependency on varying features, descriptors, dataset size, route of exposure, and toxic dose. The HNN-Tox model, trained on 59,373 chemicals annotated with known LD50 and routes of exposure, maintained its predictive ability with an accuracy of 84.9% and 84.1%, even after reducing the descriptor size from 318 to 51, and the area under the ROC curve (AUC) was 0.89 and 0.88, respectively. Further, we validated the HNN-Tox with several external toxic chemical datasets on a large scale. The HNN-Tox performed optimally or better than the other machine-learning methods for diverse chemicals. This study is the first to report a large-scale prediction of dose-range chemical toxicity with varying features. The HNN-Tox has broad applicability in predicting toxicity for diverse chemicals and could serve as an alternative methodology approach to animal-based toxicity assessment.

Ligand Binding Prediction Using Protein Structure Graphs and Residual Graph Attention Networks

Computational prediction of ligand–target interactions is a crucial part of modern drug discovery as it helps to bypass high costs and labor demands of in vitro and in vivo screening. As the wealth of bioactivity data accumulates, it provides opportunities for the development of deep learning (DL) models with increasing predictive powers. Conventionally, such models were either limited to the use of very simplified representations of proteins or ineffective voxelization of their 3D structures. Herein, we present the development of the PSG-BAR (Protein Structure Graph-Binding Affinity Regression) approach that utilizes 3D structural information of the proteins along with 2D graph representations of ligands. The method also introduces attention scores to selectively weight protein regions that are most important for ligand binding. Results: The developed approach demonstrates the state-of-the-art performance on several binding affinity benchmarking datasets. The attention-based pooling of protein graphs enables identification of surface residues as critical residues for protein–ligand binding. Finally, we validate our model predictions against an experimental assay on a viral main protease (Mpro)—the hallmark target of SARS-CoV-2 coronavirus.

Nanostructuring and macroscopic behavior of type V deep eutectic solvents based on monoterpenoids

Type V natural deep eutectic solvents based on monoterpenoids (cineole, carvone, menthol, and thymol) are studied using a combined experimental and molecular modeling approach. The reported physicochemical properties showed low viscous fluids whose properties were characterized as a function of temperature. The theoretical study combining quantum chemistry and classical molecular dynamics simulations provided a nanoscopic characterization of the fluids, particularly for the hydrogen bonding network and its relationship with the macroscopic properties. The considered fluids constitute a suitable type of solvents considering their properties, cost, origin, and sustainability in different technological applications and sow the possibility of developing type V NADES from different types of molecules, especially in the terpenoid family of compounds.

Identification of Novel Antibacterials Using Machine Learning Techniques

Many pharmaceutical companies are avoiding the development of novel antibacterials due to a range of rational reasons and the high risk of failure. However, there is an urgent need for novel antibiotics especially against resistant bacterial strains. Available in silico models suffer from many drawbacks and, therefore, are not applicable for scoring novel molecules with high structural diversity by their antibacterial potency. Considering this, the overall aim of this study was to develop an efficient in silico model able to find compounds that have plenty of chances to exhibit antibacterial activity. Based on a proprietary screening campaign, we have accumulated a representative dataset of more than 140,000 molecules with antibacterial activity against Escherichia coli assessed in the same assay and under the same conditions. This intriguing set has no analogue in the scientific literature. We applied six in silico techniques to mine these data. For external validation, we used 5,000 compounds with low similarity towards training samples. The antibacterial activity of the selected molecules against E. coli was assessed using a comprehensive biological study. Kohonen-based nonlinear mapping was used for the first time and provided the best predictive power (av. 75.5%). Several compounds showed an outstanding antibacterial potency and were identified as translation machinery inhibitors in vitro and in vivo. For the best compounds, MIC and CC50 values were determined to allow us to estimate a selectivity index (SI). Many active compounds have a robust IP position.

Comparative Study of Multitask Toxicity Modeling on a Broad Chemical Space

Acute toxicity is one of the most challenging properties to predict purely with computational methods due to its direct relationship to biological interactions. Moreover, toxicity can be represented by different end points: it can be measured for different species using different types of administration, etc., and it is questionable if the knowledge transfer between end points is possible. We performed a comparative study of prediction multitask toxicity for a broad chemical space using different descriptors and modeling algorithms and applied multitask learning for a large toxicity data set extracted from the Registry of Toxic Effects of Chemical Substances (RTECS). We demonstrated that multitask modeling provides significant improvement over single-output models and other machine learning methods. Our research reveals that multitask learning can be very useful to improve the quality of acute toxicity modeling and raises a discussion about the usage of multitask approaches for regulation purposes. Our MultiTox models are freely available in OCHEM platform ( ochem.eu/multitox ) under CC-BY-NC license.

Antimicrobial cell penetrating peptides with bacterial cell specificity: pharmacophore modelling, quantitative structure activity relationship and molecular dynamics simulation

Current research has shown cell-penetrating peptides and antimicrobial peptides (AMPs) as probable vectors for use in drug delivery and as novel antibiotics. It has been reported that the higher the therapeutic index (TI) the higher would be the bacterial cell penetrating ability. To the best of our knowledge, no in-silico study has been performed to determine bacterial cell specificity of the antimicrobial cell penetrating peptides (aCPP's) based on their TI. The aim of this study was to develop a quantitative structure activity relationship (QSAR) model, which can estimate antimicrobial potential and cell-penetrating ability of aCPPs against S. aureus, to confirm the relationship between the TI and aCPPs and to identify specific descriptors responsible for aCPPs penetrating ability. Molecular dynamics (MD) simulation was also performed to confirm the membrane insertion of the most active aCPPs obtained from the QSAR study. The most appropriate pharmacophore was identified to predict the aCPP's activity. The statistical results confirmed the validity of the model. The QSAR model was successful in identifying the optimal aCPP with high activity prediction and provided insights into the structural requirements to correlate their TI to cell penetrating ability. MD simulation of the best aCPP with 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayer confirmed its interaction with the membrane and the C-terminal residues of the aCPP played a key role in membrane penetration. The strategy of combining QSAR and molecular dynamics, allowed for optimal estimation of ligand-target interaction and confirmed the importance of Trp and Lys in interacting with the POPC bilayer. Communicated by Ramaswamy H. Sarma.

Machine Learning Based Toxicity Prediction: From Chemical Structural Description to Transcriptome Analysis

Toxicity prediction is very important to public health. Among its many applications, toxicity prediction is essential to reduce the cost and labor of a drug's preclinical and clinical trials, because a lot of drug evaluations (cellular, animal, and clinical) can be spared due to the predicted toxicity. In the era of Big Data and artificial intelligence, toxicity prediction can benefit from machine learning, which has been widely used in many fields such as natural language processing, speech recognition, image recognition, computational chemistry, and bioinformatics, with excellent performance. In this article, we review machine learning methods that have been applied to toxicity prediction, including deep learning, random forests, k-nearest neighbors, and support vector machines. We also discuss the input parameter to the machine learning algorithm, especially its shift from chemical structural description only to that combined with human transcriptome data analysis, which can greatly enhance prediction accuracy.

Comparison of global and mode of action-based models for aquatic toxicity

The ability to estimate aquatic toxicity is a critical need for ecological risk assessment and chemical regulation. The consensus in the literature is that mode of action (MOA) based toxicity models yield the most toxicologically meaningful and, theoretically, the most accurate results. In this study, a two-step prediction methodology was developed to estimate acute aquatic toxicity from molecular structure. In the first step, one-against-the-rest linear discriminant analysis (LDA) models were used to predict the MOA. The LDA models were able to predict the MOA with 85.8-88.8% accuracy for broad and specific MOAs, respectively. In the second step, a multiple linear regression (MLR) model corresponding to the predicted MOA was used to predict the acute aquatic toxicity value. The MOA-based approach was found to yield similar external prediction accuracy (r(2) = 0.529-0.632) to a single global MLR model (r(2) = 0.551-0.562) fit to the entire training set. Overall, the global hierarchical clustering approach yielded a higher combination of accuracy and prediction coverage (r(2) = 0.572, coverage = 99.3%) than the other approaches. Utilizing multiple two-dimensional chemical descriptors in MLR models yielded comparable results to using only the octanol-water partition coefficient (log K(ow)).

QSAR models for thiophene and imidazopyridine derivatives inhibitors of the Polo-Like Kinase 1

The inhibitory activity of 103 thiophene and 33 imidazopyridine derivatives against Polo-Like Kinase 1 (PLK1) expressed as pIC50 (-logIC50) was predicted by QSAR modeling. Multivariate linear regression (MLR) was employed to model the relationship between 0D and 3D molecular descriptors and biological activities of molecules using the replacement method (MR) as variable selection tool. The 136 compounds were separated into several training and test sets. Two splitting approaches, distribution of biological data and structural diversity, and the statistical experimental design procedure D-optimal distance were applied to the dataset. The significance of the training set models was confirmed by statistically higher values of the internal leave one out cross-validated coefficient of determination (Q2) and external predictive coefficient of determination for the test set (Rtest2). The model developed from a training set, obtained with the D-optimal distance protocol and using 3D descriptor space along with activity values, separated chemical features that allowed to distinguish high and low pIC50 values reasonably well. Then, we verified that such model was sufficient to reliably and accurately predict the activity of external diverse structures. The model robustness was properly characterized by means of standard procedures and their applicability domain (AD) was analyzed by leverage method.

Peptide design for antimicrobial and immunomodulatory applications

The increasing threat of antibiotic resistance in pathogenic bacteria and the dwindling supply of antibiotics available to combat these infections poses a significant threat to human health throughout the world. Antimicrobial peptides (AMPs) have long been touted as the next generation of antibiotics capable of filling the anti-infective void. Unfortunately, peptide-based antibiotics have yet to realize their potential as novel pharmaceuticals, in spite of the immense number of known AMP sequences and our improved understanding of their antibacterial mechanism of action. Recently, the immunomodulatory properties of certain AMPs have become appreciated. The ability of small synthetic peptides to protect against infection in vivo has demonstrated that modulation of the innate immune response is an effective strategy to further develop peptides as novel anti-infectives. This review focuses on the screening methods that have been used to assess novel peptide sequences for their antibacterial and immunomodulatory properties. It will also examine how we have progressed in our ability to identify and optimize peptides with desired biological characteristics and enhanced therapeutic potential. In addition, the current challenges to the development of peptides as anti-infectives are examined and the strategies being used to overcome these issues are discussed.

Predicting antitumor activity of peptides by consensus of regression models trained on a small data sample

Predicting antitumor activity of compounds using regression models trained on a small number of compounds with measured biological activity is an ill-posed inverse problem. Yet, it occurs very often within the academic community. To counteract, up to some extent, overfitting problems caused by a small training data, we propose to use consensus of six regression models for prediction of biological activity of virtual library of compounds. The QSAR descriptors of 22 compounds related to the opioid growth factor (OGF, Tyr-Gly-Gly-Phe-Met) with known antitumor activity were used to train regression models: the feed-forward artificial neural network, the k-nearest neighbor, sparseness constrained linear regression, the linear and nonlinear (with polynomial and Gaussian kernel) support vector machine. Regression models were applied on a virtual library of 429 compounds that resulted in six lists with candidate compounds ranked by predicted antitumor activity. The highly ranked candidate compounds were synthesized, characterized and tested for an antiproliferative activity. Some of prepared peptides showed more pronounced activity compared with the native OGF; however, they were less active than highly ranked compounds selected previously by the radial basis function support vector machine (RBF SVM) regression model. The ill-posedness of the related inverse problem causes unstable behavior of trained regression models on test data. These results point to high complexity of prediction based on the regression models trained on a small data sample.

Quantitative structure-activity relationships and docking studies of calcitonin gene-related peptide antagonists

Defining the role of calcitonin gene-related peptide in migraine pathogenesis could lead to the application of calcitonin gene-related peptide antagonists as novel migraine therapeutics. In this work, quantitative structure-activity relationship modeling of biological activities of a large range of calcitonin gene-related peptide antagonists was performed using a panel of physicochemical descriptors. The computational studies evaluated different variable selection techniques and demonstrated shuffling stepwise multiple linear regression to be superior over genetic algorithm-multiple linear regression. The linear quantitative structure-activity relationship model revealed better statistical parameters of cross-validation in comparison with the non-linear support vector regression technique. Implementing only five peptide descriptors into this linear quantitative structure-activity relationship model resulted in an extremely robust and highly predictive model with calibration, leave-one-out and leave-20-out validation R(2) of 0.9194, 0.9103, and 0.9214, respectively. We performed docking of the most potent calcitonin gene-related peptide antagonists with the calcitonin gene-related peptide receptor and demonstrated that peptide antagonists act by blocking access to the peptide-binding cleft. We also demonstrated the direct contact of residues 28-37 of the calcitonin gene-related peptide antagonists with the receptor. These results are in agreement with the conclusions drawn from the quantitative structure-activity relationship model, indicating that both electrostatic and steric factors should be taken into account when designing novel calcitonin gene-related peptide antagonists.

Descriptors for antimicrobial peptides

INTRODUCTION

A frightening increase in the number of isolated multidrug resistant bacterial strains linked to the decline in novel antimicrobial drugs entering the market is a great cause for concern. Cationic antimicrobial peptides (AMPs) have lately been introduced as a potential new class of antimicrobial drugs, and computational methods utilizing molecular descriptors can significantly accelerate the development of new peptide drug candidates.

AREAS COVERED

This paper gives a broad overview of peptide and amino-acid scale descriptors available for AMP modeling and highlights which of these are currently being used in quantitative structure-activity relationship (QSAR) studies for AMP optimization. Additionally, some key commercial computational tools are discussed, and both successful and less successful studies are referenced, illustrating some of the challenges facing AMP scientists. Through examples of different peptide QSAR studies, this review highlights some of the missing links and illuminates some of the questions that would be interesting to challenge in a more systematic fashion.

EXPERT OPINION

Computer-aided peptide QSAR using molecular descriptors may provide the necessary edge to peptide drug discovery, enabling successful design of a new generation anti-infective drug molecules. However, if this wonderful scenario is to play out, computational chemists and peptide microbiologists would need to start playing together and not just side by side.

Optimization of antibacterial peptides by genetic algorithms and cheminformatics

Pathogens resistant to available drug therapies are a pressing global health problem. Short, cationic peptides represent a novel class of agents that have lower rates of drug resistance than derivatives of current antibiotics. Previously, we created a software system utilizing artificial neural networks that were trained on quantitative structure-activity relationship descriptors calculated for a total of 1400 synthetic peptides for which antibacterial activity was determined. Using the trained system, we correctly identified additional peptides with activity of 94% accuracy; active peptides were 47 of the top rated 50 peptides chosen from an in silico library of nearly 100,000 sequences. Here, we report a method of generating candidate peptide sequences using the heuristic evolutionary programming method of genetic algorithms (GA), which provided a large (19-fold) improvement in identification of novel antibacterial peptides. Approximately 0.50% of peptides evaluated during the GA method were classified as highly active, while only 0.026% of the nearly 100,000 sequences we previously screened were classified as highly active. A selection of these peptides was tested in vitro and activities reported here. While GA significantly improves the possibility of identifying candidate peptides, we encountered important pitfalls to this method that should be considered when using GA.

Identification of novel antibacterial peptides by chemoinformatics and machine learning

The rise of antibiotic resistant pathogens is one of the most pressing global health issues. Discovery of new classes of antibiotics has not kept pace; new agents often suffer from cross-resistance to existing agents of similar structure. Short, cationic peptides with antimicrobial activity are essential to the host defenses of many organisms and represent a promising new class of antimicrobials. This paper reports the successful in silico screening for potent antibiotic peptides using a combination of QSAR and machine learning techniques. On the basis of initial high-throughput measurements of activity of over 1400 random peptides, artificial neural network models were built using QSAR descriptors and subsequently used to screen an in silico library of approximately 100,000 peptides. In vitro validation of the modeling showed 94% accuracy in identifying highly active peptides. The best peptides identified through screening were found to have activities comparable or superior to those of four conventional antibiotics and superior to the peptide most advanced in clinical development against a broad array of multiresistant human pathogens.

Use of artificial intelligence in the design of small peptide antibiotics effective against a broad spectrum of highly antibiotic-resistant superbugs

Increased multiple antibiotic resistance in the face of declining antibiotic discovery is one of society's most pressing health issues. Antimicrobial peptides represent a promising new class of antibiotics. Here we ask whether it is possible to make small broad spectrum peptides employing minimal assumptions, by capitalizing on accumulating chemical biology information. Using peptide array technology, two large random 9-amino-acid peptide libraries were iteratively created using the amino acid composition of the most active peptides. The resultant data was used together with Artificial Neural Networks, a powerful machine learning technique, to create quantitative in silico models of antibiotic activity. On the basis of random testing, these models proved remarkably effective in predicting the activity of 100,000 virtual peptides. The best peptides, representing the top quartile of predicted activities, were effective against a broad array of multidrug-resistant "Superbugs" with activities that were equal to or better than four highly used conventional antibiotics, more effective than the most advanced clinical candidate antimicrobial peptide, and protective against Staphylococcus aureus infections in animal models.

QSAR modeling and computer-aided design of antimicrobial peptides

The drastic increase in multi-drug-resistant bacteria has created an urgent need for new therapeutic interventions, including antimicrobial peptides, an interesting template for novel drug development. However, the process of optimizing peptide antimicrobial activity and specificity using large peptide libraries is both tedious and expensive. Here we confirm the use of a mathematical model for prediction, prior to synthesis, of peptide antibacterial activity toward the antibiotic resistant pathogen Pseudomonas aeruginosa. By the use of novel descriptors quantifying the contact energy between neighboring amino acids, as well as a set of inductive and conventional QSAR descriptors, we were able to model the antibacterial activity of peptides. Cross-correlation and optimization of the implemented descriptor values enabled us to build two models, using very limited sets of peptides, which were able to correctly predict the activity of 85 or 71% of the tested peptides, within a twofold deviation window of the corresponding previously assessed IC(50) values, measured earlier. Though these two models were significantly different in size, they demonstrated no significant difference in their predictive power, implying that it is possible to build powerful predictive models using even small sets of structurally different peptides, when using contact-energy descriptors and inductive and conventional QSAR descriptors in the model design.

Short linear cationic antimicrobial peptides: screening, optimizing, and prediction

The problem of pathogenic antibiotic-resistant bacteria such as Staphylococcus aureus and Pseudomonas aeruginosa is worsening, demonstrating the urgent need for new therapeutics that are effective against multidrug-resistant bacteria. One potential class of substances is cationic antimicrobial peptides. More than 1000 natural occurring peptides have been described so far. These peptides are short (less than 50 amino acids long), cationic, amphiphilic, demonstrate different three-dimensional structures, and appear to have different modes of action. A new screening assay was developed to characterize and optimize short antimicrobial peptides. This assay is based on peptides synthesized on cellulose, combined with a bacterium, where a luminescence gene cassette was introduced. With help of this method tens of thousands of peptides can be screened per year. Information gained by this high-throughput screening can be used in quantitative structure-activity relationships (QSAR) analysis. QSAR analysis attempts to correlate chemical structure to measurement of biological activity using statistical methods. QSAR modeling of antimicrobial peptides to date has been based on predicting differences between peptides that are highly similar. The studies have largely addressed differences in lactoferricin and protegrin derivatives or similar de novo peptides. The mathematical models used to relate the QSAR descriptors to biological activity have been linear models such as principle component analysis or multivariate linear regression. However, with the development of high-throughput peptide synthesis and an antibacterial activity assay, the numbers of peptides and sequence diversity able to be studied have increased dramatically. Also, "inductive" QSAR descriptors have been recently developed to accurately distinguish active from inactive drug-like activity in small compounds. "Inductive" QSAR in combination with more complex mathematical modeling algorithms such as artificial neural networks (ANNs) may yield powerful new methods for in silico identification of novel antimicrobial peptides.

Evaluating different descriptors for model design of antimicrobial peptides with enhanced activity toward P. aeruginosa

The number of isolated drug-resistant pathogenic microbes has increased drastically over the past decades, demonstrating an urgent need for new therapeutic interventions. Antimicrobial peptides have for a long time been looked upon as an interesting template for drug optimization. However, the process of optimizing peptide antimicrobial activity and specificity, using large peptide libraries is both tedious and expensive. Here, we describe the construction of a mathematical model for prediction, prior to synthesis, of peptide antibacterial activity toward Pseudomonas aeruginosa. By use of novel descriptors quantifying the contact energy between neighboring amino acids in addition to a set of inductive and conventional quantitative structure-activity relationship descriptors, we are able to model the peptides antibacterial activity. Cross-correlation and optimization of the implemented descriptor values have enabled us to build a model (Bac2a- #2) that was able to correctly predict the activity of 84% of the tested peptides, within a twofold deviation window of the corresponding IC50 values, measured earlier. The predictive power, is an average of 10 submodels, each predicting the activity of 20 randomly excluded peptides, with a predictive success of 16.7 +/- 1.6 peptides. The model has also been proven significantly more accurate than a simpler model (Bac2a- #1), where the inductive and conventional quantitative structure-activity relationship descriptors were excluded.

Progressive Docking: A Hybrid QSAR/Docking Approach for Accelerating In Silico High Throughput Screening

A combination of protein-ligand docking and ligand-based QSAR approaches has been elaborated, aiming to speed-up the process of virtual screening. In particular, this approach utilizes docking scores generated for already processed compounds to build predictive QSAR models that, in turn, assess hypothetical target binding affinities for yet undocked entries. The "progressive docking" has been tested on drug-like substances from the NCI database that have been docked into several unrelated targets, including human sex hormone binding globulin (SHBG), carbonic anhydrase, corticosteroid-binding globulin, SARS 3C-like protease, and HIV1 reverse transcriptase. We demonstrate that progressive docking can reduce the amount of computations 1.2- to 2.6-fold (when compared to traditional docking), while maintaining 80-99% hit recovery rates. This progressive-docking procedure, therefore, substantially accelerates high throughput screening, especially when using high accuracy (slower) docking approaches and large-sized datasets, and has allowed us to identify several novel potent nonsteroidal SHBG ligands.