SMILES Strings Research Articles

STOPLIGHT: A Hit Scoring Calculator.

We introduce STOPLIGHT, a web portal to assist medicinal chemists in prioritizing hits from screening campaigns and the selection of compounds for optimization. STOPLIGHT incorporates services to assess 6 physiochemical and structural properties, 6 assay liabilities, and 11 pharmacokinetic properties, for any small molecule represented by its SMILES string. We briefly describe each service and illustrate the utility of this portal with a case study. The STOPLIGHT portal provides a user-friendly tool to guide hit selection in early drug discovery campaigns, whereby compounds with unfavorable properties can be quickly recognized and eliminated.

A cautionary tale about properly vetting datasets used in supervised learning predicting metabolic pathway involvement.

The mapping of metabolite-specific data to pathways within cellular metabolism is a major data analysis step needed for biochemical interpretation. A variety of machine learning approaches, particularly deep learning approaches, have been used to predict these metabolite-to-pathway mappings, utilizing a training dataset of known metabolite-to-pathway mappings. A few such training datasets have been derived from the Kyoto Encyclopedia of Genes and Genomes (KEGG). However, several prior published machine learning approaches utilized an erroneous KEGG-derived training dataset that used SMILES molecular representations strings (KEGG-SMILES dataset) and contained a sizable proportion (~26%) duplicate entries. The presence of so many duplicates taint the training and testing sets generated from k-fold cross-validation of the KEGG-SMILES dataset. Therefore, the k-fold cross-validation performance of the resulting machine learning models was grossly inflated by the erroneous presence of these duplicate entries. Here we describe and evaluate the KEGG-SMILES dataset so that others may avoid using it. We also identify the prior publications that utilized this erroneous KEGG-SMILES dataset so their machine learning results can be properly and critically evaluated. In addition, we demonstrate the reduction of model k-fold cross-validation (CV) performance after de-duplicating the KEGG-SMILES dataset. This is a cautionary tale about properly vetting prior published benchmark datasets before using them in machine learning approaches. We hope others will avoid similar mistakes.

FraHMT: A Fragment-Oriented Heterogeneous Graph Molecular Generation Model for Target Proteins.

The molecular generation task stands as a pivotal step in the domains of computational chemistry and drug discovery, aiming to computationally generate molecular structures for specific properties. In contrast to previous models that focused primarily on SMILES strings or molecular graphs, our model placed a special emphasis on the substructure information on molecules, enabling the model to learn richer chemical rules and structure features from fragments and chemical reaction information on molecules. To accomplish this, we fragmented the molecules to construct heterogeneous graph representations based on atom and fragment information. Then our model mapped the heterogeneous graph data into a latent vector space by using an encoder and employed a self-regressive generative model as a decoder for molecular generation. Additionally, we performed transfer learning on the model using a small set of ligand molecules known to be active against the target protein to generate molecules that bind better to the target protein. Experimental results demonstrate that our model is highly competitive with state-of-the-art models. It can generate valid and diverse molecules with favorable physicochemical properties and drug-likeness. Importantly, they produce novel molecules with high docking scores against the target proteins.

AdmetSAR3.0: a comprehensive platform for exploration, prediction and optimization of chemical ADMET properties.

Absorption, distribution, metabolism, excretion and toxicity (ADMET) properties play a crucial role in drug discovery and chemical safety assessment. Built on the achievements of admetSAR and its successor, admetSAR2.0, this paper introduced the new version of the series, admetSAR3.0, as a comprehensive platform for chemical ADMET assessment, including search, prediction and optimization modules. In the search module, admetSAR3.0 hosted over 370000 high-quality experimental ADMET data for 104652 unique compounds, and supplemented chemical structure similarity search function to facilitate read-across. In the prediction module, we introduced comprehensive ADMET endpoints and two new sections for environmental and cosmetic risk assessments, empowering admetSAR3.0 to provide prediction for 119 endpoints, more than double numbers compared to the previous version. Furthermore, the advanced multi-task graph neural network framework offered robust and reliable support for ADMET prediction. In particular, a module named ADMETopt was added to automatically optimize the ADMET properties of query molecules through transformation rules or scaffold hopping. Finally, admetSAR3.0 provides user-friendly interfaces for multiple types of input data, such as SMILES string, chemical structure and batch molecule file, and supports various output types, including digital, chart displays and file downloads. In summary, admetSAR3.0 is anticipated to be a valuable and powerful tool in drug discovery and chemical safety assessment at http://lmmd.ecust.edu.cn/admetsar3/.

Cancer drug sensitivity estimation using modular deep Graph Neural Networks.

Computational drug sensitivity models have the potential to improve therapeutic outcomes by identifying targeted drugs components that are tailored to the transcriptomic profile of a given primary tumor. The SMILES representation of molecules that is used by state-of-the-art drug-sensitivity models is not conducive for neural networks to generalize to new drugs, in part because the distance between atoms does not generally correspond to the distance between their representation in the SMILES strings. Graph-attention networks, on the other hand, are high-capacity models that require large training-data volumes which are not available for drug-sensitivity estimation. We develop a modular drug-sensitivity graph-attentional neural network. The modular architecture allows us to separately pre-train the graph encoder and graph-attentional pooling layer on related tasks for which more data are available. We observe that this model outperforms reference models for the use cases of precision oncology and drug discovery; in particular, it is better able to predict the specific interaction between drug and cell line that is not explained by the general cytotoxicity of the drug and the overall survivability of the cell line. The complete source code is available at https://zenodo.org/doi/10.5281/zenodo.8020945. All experiments are based on the publicly available GDSC data.

Text-Guided Molecule Generation with Diffusion Language Model

Text-guided molecule generation is a task where molecules are generated to match specific textual descriptions. Recently, most existing SMILES-based molecule generation methods rely on an autoregressive architecture. In this work, we propose the Text-Guided Molecule Generation with Diffusion Language Model (TGM-DLM), a novel approach that leverages diffusion models to address the limitations of autoregressive methods. TGM-DLM updates token embeddings within the SMILES string collectively and iteratively, using a two-phase diffusion generation process. The first phase optimizes embeddings from random noise, guided by the text description, while the second phase corrects invalid SMILES strings to form valid molecular representations. We demonstrate that TGM-DLM outperforms MolT5-Base, an autoregressive model, without the need for additional data resources. Our findings underscore the remarkable effectiveness of TGM-DLM in generating coherent and precise molecules with specific properties, opening new avenues in drug discovery and related scientific domains. Code will be released at: https://github.com/Deno-V/tgm-dlm.

Predicting Lewis Acidity: Machine Learning the Fluoride Ion Affinity of p-Block-Atom-Based Molecules.

"How strong is this Lewis acid?" is a question researchers often approach by calculating its fluoride ion affinity (FIA) with quantum chemistry. Here, we present FIA49k, an extensive FIA dataset with 48,986 data points calculated at the RI-DSD-BLYP-D3(BJ)/def2-QZVPP//PBEh-3c level of theory, including 13 different p-block atoms as the fluoride accepting site. The FIA49k dataset was used to train FIA-GNN, two message-passing graph neural networks, which predict gas and solution phase FIA values of molecules excluded from training with a mean absolute error of 14 kJ mol-1 (r2=0.93) from the SMILES string of the Lewis acid as the only input. The level of accuracy is notable, given the wide energetic range of 750 kJ mol-1 spanned by FIA49k. The model's value was demonstrated with four case studies, including predictions for molecules extracted from the Cambridge Structural Database and by reproducing results from catalysis research available in the literature. Weaknesses of the model are evaluated and interpreted chemically. FIA-GNN and the FIA49k dataset can be reached via a free web app (www.grebgroup.de/fia-gnn).

Dual-View Learning Based on Images and Sequences for Molecular Property Prediction.

The prediction of molecular properties remains a challenging task in the field of drug design and development. Recently, there has been a growing interest in the analysis of biological images. Molecular images, as a novel representation, have proven to be competitive, yet they lack explicit information and detailed semantic richness. Conversely, semantic information in SMILES sequences is explicit but lacks spatial structural details. Therefore, in this study, we focus on and explore the relationship between these two types of representations, proposing a novel multimodal architecture named ISMol. ISMol relies on a cross-attention mechanism to extract information representations of molecules from both images and SMILES strings, thereby predicting molecular properties. Evaluation results on 14 small molecule ADMET datasets indicate that ISMol outperforms machine learning (ML) and deep learning (DL) models based on single-modal representations. In addition, we analyze our method through a large number of experiments to test the superiority, interpretability and generalizability of the method. In summary, ISMol offers a powerful deep learning toolbox for drug discovery in a variety of molecular properties.

Molecular Contrastive Pretraining with Collaborative Featurizations.

Molecular pretraining, which learns molecular representations over massive unlabeled data, has become a prominent paradigm to solve a variety of tasks in computational chemistry and drug discovery. Recently, prosperous progress has been made in molecular pretraining with different molecular featurizations, including 1D SMILES strings, 2D graphs, and 3D geometries. However, the role of molecular featurizations with their corresponding neural architectures in molecular pretraining remains largely unexamined. In this paper, through two case studies─chirality classification and aromatic ring counting─we first demonstrate that different featurization techniques convey chemical information differently. In light of this observation, we propose a simple and effective MOlecular pretraining framework with COllaborative featurizations (MOCO). MOCO comprehensively leverages multiple featurizations that complement each other and outperforms existing state-of-the-art models that solely rely on one or two featurizations on a wide range of molecular property prediction tasks.

A database of thermally activated delayed fluorescent molecules auto-generated from scientific literature with ChemDataExtractor

A database of thermally activated delayed fluorescent (TADF) molecules was automatically generated from the scientific literature. It consists of 25,482 data records with an overall precision of 82%. Among these, 5,349 records have chemical names in the form of SMILES strings which are represented with 91% accuracy; these are grouped in a subsidiary database. Each data record contains one of the following four properties: maximum emission wavelength (λEM), photoluminescence quantum yield (PLQY), singlet-triplet energy splitting (ΔEST), and delayed lifetime (τD). The databases were created through text mining using ChemDataExtractor, a chemistry-aware natural-language-processing toolkit, which has been adapted for TADF research. The text-mined corpus consisted of 2,733 papers from the Royal Society of Chemistry and Elsevier. To the best of our knowledge, these databases are the first databases that have been auto-generated for TADF molecules from existing publications. The databases have been publicly released for experimental and computational applications in the TADF research field.

PyComp: A Versatile Tool for Efficient Data Extraction, Conversion, and Management in High-throughput Virtual Drug Screening.

Virtual screening (VS) is essential for analyzing potential drug candidates in drug discovery. Often, this involves the conversion of large volumes of compound data into specific formats suitable for computational analysis. Managing and processing this wealth of information, especially when dealing with vast numbers of compounds in various forms, such as names, identifiers, or SMILES strings, can present significant logistical and technical challenges. To streamline this process, we developed PyComp, a software tool using Python's PyQt5 library, and compiled it into an executable with Pyinstaller. PyComp provides a systematic way for users to retrieve and convert a list of compound names, IDs (even in a range), or SMILES strings into the desired 3D format. PyComp greatly enhances the efficiency of data extraction, conversion, and storage processes involved in VS. It searches for similar compounds coupled with its ability to handle misidentified compounds and offers users an easy-to-use, customizable tool for managing largescale compound data. By streamlining these operations, PyComp allows researchers to save significant time and effort, thus accelerating the pace of drug discovery research. PyComp effectively addresses some of the most pressing challenges in highthroughput VS: efficient management and conversion of large volumes of compound data. As a user-friendly, customizable software tool, PyComp is pivotal in improving the efficiency and success of large-scale drug screening efforts, paving the way for faster discovery of potential therapeutic compounds.

BINDTI: A bi-directional Intention network for drug-target interaction identification based on attention mechanisms.

The identification of drug-target interactions (DTIs) is an essential step in drug discovery. In vitro experimental methods are expensive, laborious, and time-consuming. Deep learning has witnessed promising progress in DTI prediction. However, how to precisely represent drug and protein features is a major challenge for DTI prediction. Here, we developed an end-to-end DTI identification framework called BINDTI based on bi-directional Intention network. First, drug features are encoded with graph convolutional networks based on its 2D molecular graph obtained by its SMILES string. Next, protein features are encoded based on its amino acid sequence through a mixed model called ACmix, which integrates self-attention mechanism and convolution. Third, drug and target features are fused through bi-directional Intention network, which combines Intention and multi-head attention. Finally, unknown drug-target (DT) pairs are classified through multilayer perceptron based on the fused DT features. The results demonstrate that BINDTI greatly outperformed four baseline methods (i.e., CPI-GNN, TransfomerCPI, MolTrans, and IIFDTI) on the BindingDB, BioSNAP, DrugBank, and Human datasets. More importantly, it was more appropriate to predict new DTIs than the four baseline methods on imbalanced datasets. Ablation experimental results elucidated that both bi-directional Intention and ACmix could greatly advance DTI prediction. The fused feature visualization and case studies manifested that the predicted results by BINDTI were basically consistent with the true ones. We anticipate that the proposed BINDTI framework can find new low-cost drug candidates, improve drugs' virtual screening, and further facilitate drug repositioning as well as drug discovery. BINDTI is publicly available at https://github.com/plhhnu/BINDTI.

DEEPSIDE A DEEP LEARNING FRAMEWORK FOR DRUG SIDE EFFECT PREDICTION

Drug failures brought on by unanticipated side effects during clinical trials put participants' health at risk and result in significant financial losses. Algorithms for predicting side effects may direct the development of new drugs. The LINCS L1000 dataset establishes a knowledge foundation for context-specific characteristics and offers a wealth of cell line gene expression data that has been altered by various medications. The state-of-the-art method, which seeks to use context-specific information, discards a significant number of the trials and only uses the high-quality experiments in LINCS L1000. Our aim in this research is to maximize the prediction performance by fully using this data. Five deep learning architectures are used in our experiments. We discover that when drug chemical structure (CS) and the whole collection of drug altered gene expression profiles (GEX) are employed as modalities, a multi-modal architecture yields the greatest prediction performance among multi-layer perceptron-based designs. We find that overall, the CS provides more information than the GEX. The best results are obtained by a convolutional neural network-based model that just employs the SMILES string representation of the medications; this model outperforms the state-of-the-art by 3:1% micro-AUC and 13:0% macro-AUC. Additionally, we demonstrate that the model can forecast drug-side effect couples that are absent from the ground truth side effect dataset but are described in the literature.

DEEP SIDE-A DEEP LEARNING FRAMEWORK FOR DRUG SIDE EFFECT PREDICTION

Drug failures due to unforeseen adverse effects at clinical trials pose health risks for the participants and lead to substantial financial losses. Side effect prediction algorithms have the potential to guide the drug design process. LINCS L1000 dataset provides a vast resource of cell line gene expression data perturbed by different drugs and creates a knowledge base for context specific features. The state-of-the-art approach that aims at using context specific information relies on only the high quality experiments in LINCS L1000 and discards a large portion of the experiments. In this study, our goal is to boost the prediction performance by utilizing this data to its full extent. We experiment with 5 deep learning architectures. We find that a multimodal architecture produces the best predictive performance among multi-layer perceptron-based architectures when drug chemical structure (CS), and the full set of drug perturbed gene expression profiles (GEX) are used as modalities. Overall, we observe that the CS is more informative than the GEX. A convolutional neural network-based model that uses only SMILES string representation of the drugs achieves the best results and provides 13:0% macro-AUC and 3:1% micro-AUC improvements over the state-of-the-art. We also show that the model is able to predict side effect-drug pairs that are reported in the literature but was missing in the ground truth side effect dataset.

Transformer-Based Representation of Organic Molecules for Potential Modeling of Physicochemical Properties.

In this work, we study the use of three configurations of an autoencoder neural network to process organic substances with the aim of generating meaningful molecular descriptors that can be employed to develop property prediction models. A total of 18,322,500 compounds represented as SMILES strings were used to train the model, demonstrating that a latent space of 24 units is able to adequately reconstruct the data. After AE training, an analysis of the latent space properties in terms of compound similarity was carried out, indicating that this space possesses desired properties for the potential development of models for forecasting physical properties of organic compounds. As a final step, a QSPR model was developed to predict the boiling point of chemical substances based on the AE descriptors. 5276 substances were used for the regression task, and the predictive ability was compared with models available in the literature evaluated on the same database. The final AE model has an overall error of 1.40% (1.39% with augmented SMILES) in the prediction of the boiling temperature, while other models have errors between 2.0 and 3.2%. This shows that the SMILES representation is comparable and even outperforms the state-of-the-art representations widely used in the literature.

A pre-trained multi-representation fusion network for molecular property prediction

In the field of machine learning and cheminformatics, the prediction of molecular properties holds significant importance. Molecules can be represented in various formats, including 1D SMILES string, 2D graph, and 3D conformation. Numerous models have been proposed for different representations to accomplish molecular property prediction. However, most recent works have focused on one or two representations or combining embedding vectors from different perspectives in an unsophisticated manner. To address this issue, we present PremuNet, a novel pre-trained multi-representation fusion network for molecular property prediction. PremuNet can extract comprehensive molecular information from multiple views and combine them interactively through pre-training and fine-tuning. The framework of PremuNet consists of two branches: a Transformer-GNN branch that extracts SMILES and graph information, and a Fusion Net branch that extracts topology and geometry information, called PremuNet-L and PremuNet-H respectively. We employ masked self-supervised methods to enable the model to learn information fusion and achieve enhanced performance in downstream tasks. The proposed model has been evaluated on eight molecular property prediction tasks, including five classification and three regression tasks, and attained state-of-the-art performance in most cases. Additionally, we conduct the ablation studies to demonstrate the effect of each view and the branch combination approaches.

XGBoost odor prediction model: finding the structure-odor relationship of odorant molecules using the extreme gradient boosting algorithm.

Determining the structure-odor relationship has always been a very challenging task. The main challenge in investigating the correlation between the molecular structure and its associated odor is the ambiguous and obscure nature of verbally defined odor descriptors, particularly when the odorant molecules are from different sources. With the recent developments in machine learning (ML) technology, ML and data analytic techniques are significantly being used for quantitative structure-activity relationship (QSAR) in the chemistry domain toward knowledge discovery where the traditional Edisonian methods have not been useful. The smell perception of odorant molecules is one of the aforementioned tasks, as olfaction is one of the least understood senses as compared to other senses. In this study, the XGBoost odor prediction model was generated to classify smells of odorant molecules from their SMILES strings. We first collected the dataset of 1278 odorant molecules with seven basic odor descriptors, and then 1875 physicochemical properties of odorant molecules were calculated. To obtain relevant physicochemical features, a feature reduction algorithm called PCA was also employed. The ML model developed in this study was able to predict all seven basic smells with high precision (>99%) and high sensitivity (>99%) when tested on an independent test dataset. The results of the proposed study were also compared with three recently conducted studies. The results indicate that the XGBoost-PCA model performed better than the other models for predicting common odor descriptors. The methodology and ML model developed in this study may be helpful in understanding the structure-odor relationship.Communicated by Ramaswamy H. Sarma.

DataDTA: a multi-feature and dual-interaction aggregation framework for drug-target binding affinity prediction.

Accurate prediction of drug-target binding affinity (DTA) is crucial for drug discovery. The increase in the publication of large-scale DTA datasets enables the development of various computational methods for DTA prediction. Numerous deep learning-based methods have been proposed to predict affinities, some of which only utilize original sequence information or complex structures, but the effective combination of various information and protein-binding pockets have not been fully mined. Therefore, a new method that integrates available key information is urgently needed to predict DTA and accelerate the drug discovery process. In this study, we propose a novel deep learning-based predictor termed DataDTA to estimate the affinities of drug-target pairs. DataDTA utilizes descriptors of predicted pockets and sequences of proteins, as well as low-dimensional molecular features and SMILES strings of compounds as inputs. Specifically, the pockets were predicted from the three-dimensional structure of proteins and their descriptors were extracted as the partial input features for DTA prediction. The molecular representation of compounds based on algebraic graph features was collected to supplement the input information of targets. Furthermore, to ensure effective learning of multiscale interaction features, a dual-interaction aggregation neural network strategy was developed. DataDTA was compared with state-of-the-art methods on different datasets, and the results showed that DataDTA is a reliable prediction tool for affinities estimation. Specifically, the CI of DataDTA is 0.806 and the R value is 0.814 on the test dataset, which is higher than other methods. The codes of DataDTA are available at https://github.com/YanZhu06/DataDTA. Supplementary data are available at Bioinformatics online.

Reichardt’s Dye-Based Solvent Polarity and Abraham Solvent Parameters: Examining Correlations and Predictive Modeling

The concept of “solvent polarity” is widely used to explain the effects of using different solvents in various scientific applications. However, a consensus regarding its definition and quantitative measure is still lacking, hindering progress in solvent-based research. This study hopes to add to the conversation by presenting the development of two linear regression models for solvent polarity, based on Reichardt’s ET(30) solvent polarity scale, using Abraham solvent parameters and a transformer-based model for predicting solvent polarity directly from molecular structure. The first linear model incorporates the standard Abraham solvent descriptors s, a, b, and the extended model ionic descriptors j+ and j−, achieving impressive test-set statistics of R2 = 0.940 (coefficient of determination), MAE = 0.037 (mean absolute error), and RMSE = 0.050 (Root-Mean-Square Error). The second model, covering a more extensive chemical space but only using the descriptors s, a, and b, achieves test-set statistics of R2 = 0.842, MAE = 0.085, and RMSE = 0.104. The transformer-based model, applicable to any solvent with an associated SMILES string, achieves test-set statistics of R2 = 0.824, MAE = 0.066, and RMSE = 0.095. Our findings highlight the significance of Abraham solvent parameters, especially the dipolarity/polarizability, hydrogen-bond acidity/basicity, and ionic descriptors, in predicting solvent polarity. These models offer valuable insights for researchers interested in Reichardt’s ET(30) solvent polarity parameter and solvent polarity in general.

Macrocyclization of linear molecules by deep learning to facilitate macrocyclic drug candidates discovery

Interest in macrocycles as potential therapeutic agents has increased rapidly. Macrocyclization of bioactive acyclic molecules provides a potential avenue to yield novel chemical scaffolds, which can contribute to the improvement of the biological activity and physicochemical properties of these molecules. In this study, we propose a computational macrocyclization method based on Transformer architecture (which we name Macformer). Leveraging deep learning, Macformer explores the vast chemical space of macrocyclic analogues of a given acyclic molecule by adding diverse linkers compatible with the acyclic molecule. Macformer can efficiently learn the implicit relationships between acyclic and macrocyclic structures represented as SMILES strings and generate plenty of macrocycles with chemical diversity and structural novelty. In data augmentation scenarios using both internal ChEMBL and external ZINC test datasets, Macformer display excellent performance and generalisability. We showcase the utility of Macformer when combined with molecular docking simulations and wet lab based experimental validation, by applying it to the prospective design of macrocyclic JAK2 inhibitors.