Maxsmi: Maximizing molecular property prediction performance with confidence estimation using SMILES augmentation and deep learning

Hydrophobicity, as measured by Log P, is an important molecular property in assessing toxicity and carcinogenicity of disinfection by-products (DBPs). With increasing public health concerns of DPBs, there are considerable benefits in developing Quantitative Structure Activity Relationship (QSAR) models, capable of accurately predicting Log P, which could be used in health risk assessment of DBPs. In this research, Log P values of 46 halogenated alkanes, as one class of DBP compounds, were used to develop QSAR models for this class. Three molecular descriptors, namely, Energy of the Lowest Unoccupied Molecular Orbital (ELUMO), Number of Chlorine (NCl) and Number of Carbon (NC) were used in Multiple Linear Regression (MLR) analysis. The QSAR models developed were validated according to the principles set up by the Organization for Economic Co-operation and Development (OECD). Model Applicability Domain (AD) of the developed QSAR models was defined and mechanisms were interpreted. Considering large number of halogenated alkane compounds, the established QSAR models performed very well in terms of goodness-of-fit, robustness and predictability. The predicted values of Log P of DBPs by the QSAR models have correlation coefficient R2 from 81 % to 98 % with the observed Log P. The leverage approach by Williams plot was applied to detect and remove outliers. As a result, the correlation coefficient, R2, of the QSAR models increased by approximately 2 % to 13 %, before and after removing the outliers, respectively. The developed QSAR model was statistically validated for its predictive power of Log P by the Leave-One-Out (LOO) and Leave-Many-Out (LMO) cross validation methods.

BackgroundNew dipeptidyl peptidase-4 (DPP-4) inhibitors need to be developed to be used as agents with low adverse effects for the treatment of type 2 diabetes mellitus. This study aims to build quantitative structure-activity relationship (QSAR) models using the artificial intelligence paradigm. Rotation Forest and Deep Neural Network (DNN) are used to predict QSAR models. We compared principal component analysis (PCA) with sparse PCA (SPCA) as methods for transforming Rotation Forest. K-modes clustering with Levenshtein distance was used for the selection method of molecules, and CatBoost was used for the feature selection method.ResultsThe amount of the DPP-4 inhibitor molecules resulting from the selection process of molecules using K-Modes clustering algorithm is 1020 with logP range value of -1.6693 to 4.99044. Several fingerprint methods such as extended connectivity fingerprint and functional class fingerprint with diameters of 4 and 6 were used to construct four fingerprint datasets, ECFP_4, ECFP_6, FCFP_4, and FCFP_6. There are 1024 features from the four fingerprint datasets that are then selected using the CatBoost method. CatBoost can represent QSAR models with good performance for machine learning and deep learning methods respectively with evaluation metrics, such as Sensitivity, Specificity, Accuracy, and Matthew’s correlation coefficient, all valued above 70% with a feature importance level of 60%, 70%, 80%, and 90%.ConclusionThe K-modes clustering algorithm can produce a representative subset of DPP-4 inhibitor molecules. Feature selection in the fingerprint dataset using CatBoost is best used before making QSAR Classification and QSAR Regression models. QSAR Classification using Machine Learning and QSAR Classification using Deep Learning, each of which has an accuracy of above 70%. The QSAR RFC-PCA and QSAR RFR-PCA models performed better than QSAR RFC-SPCA and QSAR RFR-SPCA models because QSAR RFC-PCA and QSAR RFR-PCA models have more effective time than the QSAR RFC-SPCA and QSAR RFR-SPCA models.

Assessing the ecotoxicity of pharmaceuticals is of urgent need due to the recognition of their possible adverse effects on nontarget organisms in the aquatic environment. The reality of ecotoxicity data scarcity promotes the development and application of quantitative structure activity relationship (QSAR) models. In the present study, we aimed to clarify whether a QSAR model of ecotoxicity specifically for pharmaceuticals is needed considering that pharmaceuticals are a class of chemicals with complex structures, multiple functional groups, and reactive properties. To this end, we conducted a performance comparison of two previously developed and validated QSAR models specifically for pharmaceuticals with the commonly used narcosis toxicity prediction model, i.e., Ecological Structure Activity Relationship (ECOSAR), using a subset of pharmaceuticals produced in China that had not been included in the training datasets of QSAR models under consideration. A variety of statistical measures demonstrated that the pharmaceutical specific model outperformed ECOSAR, indicating the necessity of developing a specific QSAR model of ecotoxicity for the active pharmaceutical contaminants. ECOSAR, which was generally used to predict the baseline or the minimum toxicity of a compound, generally underestimated the ecotoxicity of the analyzed pharmaceuticals. This could possibly be because some pharmaceuticals can react through specific modes of action. Nonetheless, it should be noted that 95% prediction intervals spread over approximately four orders of magnitude for both tested QSAR models specifically for pharmaceuticals.

ProteoDockNet: Novel GNN-based ligand binding affinities prediction architecture via SMILES to key liver, kidney and brain proteins using QSAR data

Development of quantitative structure activity relationship (QSAR) model for disinfection byproduct (DBP) research: A review of methods and resources

Deep learning has revolutionized the performance of classification and object detection, but meanwhile demands sufficient labeled data for training. Given insufficient data, while many techniques have been developed to help combat overfitting, the challenge remains if one tries to train deep networks, especially in the ill-posed <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">extremely low data regimes</i> : only a small set of labeled data are available, and nothing – including unlabeled data – else. Such regimes arise from practical situations where not only data labeling but also data collection itself is expensive. We propose a deep adversarial data augmentation (DADA) technique to address the problem, in which we elaborately formulate data augmentation as a problem of training a class-conditional and supervised generative adversarial network (GAN). Specifically, a new discriminator loss is proposed to fit the goal of data augmentation, through which both real and augmented samples are enforced to contribute to and be consistent in finding the decision boundaries. Tailored training techniques are developed accordingly. To quantitatively validate its effectiveness, we first perform extensive simulations to show that DADA substantially outperforms both traditional data augmentation and a few GAN-based options. We then extend experiments to three real-world small labeled classification datasets where existing data augmentation and/or transfer learning strategies are either less effective or infeasible. We also demonstrate that DADA to can be extended to the detection task. We improve the pedestrian synthesis work by substitute for our discriminator and training scheme. Validation experiment shows that DADA can improve the detection mean average precision (mAP) compared with some traditional data augmentation techniques in object detection. Source code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/SchafferZhang/DADA</uri> .

Automated segmentation of dental restorations using deep learning: exploring data augmentation techniques.

The extensive utilization of drones has led to numerous scenarios that encompass both advantageous and perilous outcomes. By using deep learning techniques, this study aimed to reduce the dangerous effects of drone use through early detection of drones. The purpose of this study is the evaluation of deep learning approaches such as pre-trained YOLOv8 drone detection for security issues. This study focuses on the YOLOv8 model to achieve optimal performance in object detection tasks using a publicly available dataset collected by Mehdi Özel for a UAV competition that is sourced from GitHub. These images are labeled using Roboflow, and the model is trained on Google Colab. YOLOv8, known for its advanced architecture, was selected due to its suitability for real-time detection applications and its ability to process complex visual data. Hyperparameter tuning and data augmentation techniques were applied to maximize the performance of the model. Basic hyperparameters such as learning rate, batch size, and optimization settings were optimized through iterative experiments to provide the best performance. In addition to hyperparameter tuning, various data augmentation strategies were used to increase the robustness and generalization ability of the model. Techniques such as rotation, scaling, flipping, and color adjustments were applied to the dataset to simulate different conditions and variations. Among the augmentation techniques applied to the specific dataset in this study, rotation was found to deliver the highest performance. Blurring and cropping methods were observed to follow closely behind. The combination of optimized hyperparameters and strategic data augmentation allowed YOLOv8 to achieve high detection accuracy and reliable performance on the publicly available dataset. This method demonstrates the effectiveness of YOLOv8 in real-world scenarios, while also highlighting the importance of hyperparameter tuning and data augmentation in increasing model capabilities. To enhance model performance, dataset augmentation techniques including rotation and blurring are implemented. Following these steps, a significant precision value of 0.946, a notable recall value of 0.9605, and a considerable precision–recall curve value of 0.978 are achieved, surpassing many popular models such as Mask CNN, CNN, and YOLOv5.

Advances in Artificial Intelligence (AI)-assisted approaches in drug screening

The three-dimensional quantitative structure-activity relationship (QSAR) technique of comparative molecular field analysis (CoMFA) has demonstrated the ability to provide accurate predictions for diverse chemical compounds when trained with molecules of diverse chemical type. Although predictive, the derivation and utilization of models of this type are quite computationally and person power intensive. It is this intensity that pragmatically limits the widespread implementation of these models as predictive tools. In this study, two newer QSAR techniques were evaluated as possible alternatives to CoMFA based QSAR models for the purpose of rapidly identifying estrogen receptor ligands from diverse collections of molecules. The first of these is Hologram QSAR, or HQSAR. HQSAR utilizes Tripos molecular fingerprints as descriptors in conjunction with partial least squares (PLS) regression and cross-validation routines. The HQSAR technique demonstrated the ability to rapidly develop QSAR models independent of the intense user input (i.e. geometry optimization, conformational analysis, and molecular superposition were not required). Second, a newly developed QSAR paradigm that utilizes Molecular Design Limited (MDL) substructure keys (SKEYS) as descriptors in combination with an evolutionary algorithm, Fast Random Elimination of Descriptors (FRED), was evaluated. By utilizing the FRED/SKEYS algorithm, a simple substructure-based QSAR model was derived that was comparable in statistical robustness and predictive ability to both CoMFA and HQSAR derived models. A comparison of the utility of these three approaches as computational tools for the rapid identification of estrogen receptor ligands as potential endocrine disruptors as assessed by model predictive ability will be described.

In this work a model for the prediction of daily solar radiation time series is proposed, this model is based on recurrent neural networks (RNN) and data augmentation. The RNN used is Long Short-Term Memory (LSTM), and the data augmentation technique is based on time-warping and jittering. It is experimented with data of 6 years,5 years for training and 1 year for testing. The results show that the proposal model with data augmentation (DA+LSTM) considerably outperforms an LSTM model without data augmentation, as well as other state-of-the-art models with which it was compared.

Quantitative Structure Activity Relationship (QSAR) models can inform on the correlation between activities and structure-based molecular descriptors. This information is important for the understanding of the factors that govern molecular properties and for designing new compounds with favorable properties. Due to the large number of calculate-able descriptors and consequently, the much larger number of descriptors combinations, the derivation of QSAR models could be treated as an optimization problem. For continuous responses, metrics which are typically being optimized in this process are related to model performances on the training set, for example, and . Similar metrics, calculated on an external set of data (e.g., ), are used to evaluate the performances of the final models. A common theme of these metrics is that they are context -” ignorant”. In this work we propose that QSAR models should be evaluated based on their intended usage. More specifically, we argue that QSAR models developed for Virtual Screening (VS) should be derived and evaluated using a virtual screening-aware metric, e.g., an enrichment-based metric. To demonstrate this point, we have developed 21 Multiple Linear Regression (MLR) models for seven targets (three models per target), evaluated them first on validation sets and subsequently tested their performances on two additional test sets constructed to mimic small-scale virtual screening campaigns. As expected, we found no correlation between model performances evaluated by “classical” metrics, e.g., and and the number of active compounds picked by the models from within a pool of random compounds. In particular, in some cases models with favorable and/or values were unable to pick a single active compound from within the pool whereas in other cases, models with poor and/or values performed well in the context of virtual screening. We also found no significant correlation between the number of active compounds correctly identified by the models in the training, validation and test sets. Next, we have developed a new algorithm for the derivation of MLR models by optimizing an enrichment-based metric and tested its performances on the same datasets. We found that the best models derived in this manner showed, in most cases, much more consistent results across the training, validation and test sets and outperformed the corresponding MLR models in most virtual screening tests. Finally, we demonstrated that when tested as binary classifiers, models derived for the same targets by the new algorithm outperformed Random Forest (RF) and Support Vector Machine (SVM)-based models across training/validation/test sets, in most cases. We attribute the better performances of the Enrichment Optimizer Algorithm (EOA) models in VS to better handling of inactive random compounds. Optimizing an enrichment-based metric is therefore a promising strategy for the derivation of QSAR models for classification and virtual screening.

The oral acute toxicity of per- and polyfluoroalkyl compounds (PFASs) to Rat and Mouse: A mechanistic interpretation and prioritization analysis of untested PFASs by QSAR, q-RASAR and interspecies modelling methods

Quantitative Structure-Activity Relationship (QSAR) models have been used in Pharmaceutical design and Medicinal Chemistry for the discovery of anti-parasite drugs. QSAR models predict biological activity using as input different types of structural parameters of molecules. Topological Indices (TIs) are a very interesting class of these parameters. We can derive TIs from graph representations based on only nodes (atoms) and edges (chemical bonds). TIs are not time-consuming in terms of computational resources because they depend only on atom-atom connectivity information. This information expressed in the molecular graphs can be tabulated in the form of adjacency matrices easy to manipulate with computers. Consequently, TIs allow the rapid collection, annotation, retrieval, comparison and mining of molecular structures within large databases. The interest in TIs has exploded because we can use them to describe also macromolecular and macroscopic systems represented by complex networks of interactions (links) between the different parts of a system (nodes) such as: drug-target, protein-protein, metabolic, host-parasite, brain cortex, parasite disease spreading, Internet, or social networks. In this work, we review and comment on the following topics related to the use of TIs in anti-parasite drugs and target discovery. The first topic reviewed was: Topological Indices and QSAR for antiparasitic drugs. This topic included: Theoretical Background, QSAR for anti-malaria drugs, QSAR for anti-Toxoplasma drugs. The second topic was: TOMO-COMD approach to QSAR of antiparasitic drugs. We included in this topic: TOMO-COMD theoretical background and TOMO-COMD models for antihelmintic activity, Trichomonas, anti-malarials, anti-trypanosome compounds. The third section was inserted to discuss Topological Indices in the context of Complex Networks. The last section is devoted to the MARCH-INSIDE approach to QSAR of antiparasitic drugs and targets. This begins with a theoretical background for drugs and parameters for proteins. Next, we reviewed MARCH-INSIDE models for Pharmaceutical Design of antiparasitic drugs including: flukicidal drugs and anti-coccidial drugs. We close MARCH-NSIDE topic with a review of multi-target QSAR of antiparasitic drugs, MARCH-INSIDE assembly of complex networks of antiparasitic drugs. We closed the MARCH-INSIDE section discussing the prediction of proteins in parasites and MARCH-INSIDE web-servers for Protein-Protein interactions in parasites: Plasmod-PPI and Trypano-PPI web-servers. We closed this revision with an important section devoted to review some legal issues related to QSAR models.

Deep learning (also known as deep structured learning, hierarchical learning, or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using a deep graph with multiple processing layers, composed of multiple linear and nonlinear transformations. Deep learning has been characterized as a class of machine learning algorithms with the following characteristics [257]: • They use a cascade of many layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. The algorithms may be supervised or unsupervised and applications include pattern analysis (unsupervised) and classification (supervised). • They are based on the (unsupervised) learning of multiple levels of features or representations of the data. Higher-level features are derived from lower-level features to form a hierarchical representation. • They are part of the broader machine learning field of learning representations of data. • They learn multiple levels of representations that correspond to different levels of abstraction; the levels form a hierarchy of concepts. These definitions have in common: multiple layers of nonlinear processing units and the supervised or unsupervised learning of feature representations in each layer, with the layers forming a hierarchy from low-level to high-level features. Various deep learning architectures such as deep neural networks, convolutional deep neural networks, deep belief networks (DBN), and recurrent neural networks have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition, and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. In this chapter we start in Section 7.1 with an introduction, giving a brief history of this field, the relevant literature, and its applications. Then we study some basic concepts of deep learning such as convolutional neural networks, recurrent neural networks, backpropagation algorithm, restricted Boltzmann machines, and deep learning networks in Section 7.2. Then we illustrate three examples for Apache Spark implementation for mobile big data (MBD), user moving pattern extraction, and combination with nonparametric Bayesian learning, respectively, in Sections 7.3 through 7.5. Finally, we have summary in Section 7.6.