Simplified Molecular Input Line Entry System Strings Research Articles

Diffusion-based generative drug-like molecular editing with chemical natural language

Recently, diffusion models have emerged as a promising paradigm for molecular design and optimization. However, most diffusion-based molecular generative models focus on modeling 2D graphs or 3D geometries, with limited research on molecular sequence diffusion models. The International Union of Pure and Applied Chemistry (IUPAC) names are more akin to chemical natural language than the Simplified Molecular Input Line Entry System (SMILES) for organic compounds. In this work, we apply an IUPAC-guided conditional diffusion model to facilitate molecular editing from chemical natural language to chemical language (SMILES) and explore whether the pre-trained generative performance of diffusion models can be transferred to chemical natural language. We propose DiffIUPAC, a controllable molecular editing diffusion model that converts IUPAC names to SMILES strings. Evaluation results demonstrate that our model outperforms existing methods and successfully captures the semantic rules of both chemical languages. Chemical space and scaffold analysis show that the model can generate similar compounds with diverse scaffolds within the specified constraints. Additionally, to illustrate the model's applicability in drug design, we conducted case studies in functional group editing, analogue design and linker design.

Can large language models understand molecules?

PurposeLarge Language Models (LLMs) like Generative Pre-trained Transformer (GPT) from OpenAI and LLaMA (Large Language Model Meta AI) from Meta AI are increasingly recognized for their potential in the field of cheminformatics, particularly in understanding Simplified Molecular Input Line Entry System (SMILES), a standard method for representing chemical structures. These LLMs also have the ability to decode SMILES strings into vector representations.MethodWe investigate the performance of GPT and LLaMA compared to pre-trained models on SMILES in embedding SMILES strings on downstream tasks, focusing on two key applications: molecular property prediction and drug-drug interaction prediction.ResultsWe find that SMILES embeddings generated using LLaMA outperform those from GPT in both molecular property and DDI prediction tasks. Notably, LLaMA-based SMILES embeddings show results comparable to pre-trained models on SMILES in molecular prediction tasks and outperform the pre-trained models for the DDI prediction tasks.ConclusionThe performance of LLMs in generating SMILES embeddings shows great potential for further investigation of these models for molecular embedding. We hope our study bridges the gap between LLMs and molecular embedding, motivating additional research into the potential of LLMs in the molecular representation field. GitHub: https://github.com/sshaghayeghs/LLaMA-VS-GPT.

Invalid SMILES are beneficial rather than detrimental to chemical language models

Generative machine learning models have attracted intense interest for their ability to sample novel molecules with desired chemical or biological properties. Among these, language models trained on SMILES (Simplified Molecular-Input Line-Entry System) representations have been subject to the most extensive experimental validation and have been widely adopted. However, these models have what is perceived to be a major limitation: some fraction of the SMILES strings that they generate are invalid, meaning that they cannot be decoded to a chemical structure. This perceived shortcoming has motivated a remarkably broad spectrum of work designed to mitigate the generation of invalid SMILES or correct them post hoc. Here I provide causal evidence that the ability to produce invalid outputs is not harmful but is instead beneficial to chemical language models. I show that the generation of invalid outputs provides a self-corrective mechanism that filters low-likelihood samples from the language model output. Conversely, enforcing valid outputs produces structural biases in the generated molecules, impairing distribution learning and limiting generalization to unseen chemical space. Together, these results refute the prevailing assumption that invalid SMILES are a shortcoming of chemical language models and reframe them as a feature, not a bug.

Enhanced Structure-Based Prediction of Chiral Stationary Phases for Chromatographic Enantioseparation from 3D Molecular Conformations.

The accurate prediction of suitable chiral stationary phases (CSPs) for resolving the enantiomers of a given compound poses a significant challenge in chiral chromatography. Previous attempts at developing machine learning models for structure-based CSP prediction have primarily relied on 1D SMILES strings [the simplified molecular-input line-entry system (SMILES) is a specification in the form of a line notation for describing the structure of chemical species using short ASCII strings] or 2D graphical representations of molecular structures and have met with only limited success. In this study, we apply the recently developed 3D molecular conformation representation learning algorithm, which uses rapid conformational analysis and point clouds of atom positions in the 3D space, enabling efficient chemical structure-based machine learning. By harnessing the power of the rapid 3D molecular representation learning and a data set comprising over 300,000 chromatographic enantioseparation records sourced from the literature, our models afford notable improvements for the chemical structure-based choice of appropriate CSP for enantioseparation, paving the way for more efficient and informed decision-making in the field of chiral chromatography.

A Molecular Generative Model of COVID-19 Main Protease Inhibitors Using Long Short-Term Memory-Based Recurrent Neural Network.

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a serious threat to public health and prompted researchers to find anti-coronavirus 2019 (COVID-19) compounds. In this study, the long short-term memory-based recurrent neural network was used to generate new inhibitors for the coronavirus. First, the model was trained to generate drug compounds in the form of valid simplified molecular-input line-entry system strings. Then, the structures of COVID-19 main protease inhibitors were applied to fine-tune the model. After fine-tuning, the network could generate new molecular structures as novel SARS-CoV-2 main protease inhibitors. Molecular docking exhibited that some generated compounds have the proper affinity to the active site of the protease. Molecular Dynamics simulations explored binding free energies of the compounds over simulation trajectories. In addition, in silico absorption, distribution, metabolism, and excretion studies showed that some novel compounds could be formulated as orally active agents. Based on molecular docking and molecular dynamics simulation studies, compound AADH possessed significant binding affinity and presumably inhibition against the SARS-CoV-2 main protease enzyme. Therefore, the proposed deep learning-based model was capable of generating promising anti-COVID-19 drugs.

EarlGAN: An enhanced actor–critic reinforcement learning agent-driven GAN for de novo drug design

Deep generative models, such as Generative Adversarial Networks (GANs), have attracted the attention of researchers in the denovo drug design field. However, traditional GANs are typically used for image processing. Therefore, they are unsuitable for Simplified Molecular-Input Line-Entry System (SMILES) strings due to their discrete nature. Previous studies addressed this problem by combining Reinforcement Learning (RL) approaches with Monte Carlo tree search. However, for large chemical datasets, the molecule generation process is time-consuming due to the lengthy atom-by-atom sampling process with cumulative reward, an essence of the Monte Carlo tree search-based RL approaches. To address this problem, we propose an enhanced actor–critic RL agent-driven GAN, called EarlGAN, for denovo drug design. Specifically, EarlGAN’s generator acts as an actor to generate SMILES strings, and the discriminator acts as a critic to perform discrimination. EarlGAN makes autoregressive predictions at the atomic level. While the generator is based on previously generated atoms, the discriminator discriminates using a bidirectional pass over the atoms, including the current atom that is being predicted. We integrate moment, global-level discrimination rewards, and information entropy maximization. The moment rewards reduce the computation time, and the global-level rewards ensure the consistency of the molecule, whereas the information entropy maximization leads to a more diverse sample generation. Experiments and ablation studies verify the effectiveness of EarlGAN for denovo drug design on the QM9 and ZINC datasets. In addition, the visualization analysis provides insight into EarlGAN and supports our conclusion.

Descriptor-Free Deep Learning QSAR Model for the Fraction Unbound in Human Plasma.

Chemical-specific parameters are either measured in vitro or estimated using quantitative structure-activity relationship (QSAR) models. The existing body of QSAR work relies on extracting a set of descriptors or fingerprints, subset selection, and training a machine learning model. In this work, we used a state-of-the-art natural language processing model, Bidirectional Encoder Representations from Transformers, which allowed us to circumvent the need for calculation of these chemical descriptors. In this approach, simplified molecular-input line-entry system (SMILES) strings were embedded in a high-dimensional space using a two-stage training approach. The model was first pre-trained on a masked SMILES token task and then fine-tuned on a QSAR prediction task. The pre-training task learned meaningful high-dimensional embeddings based upon the relationships between the chemical tokens in the SMILES strings derived from the "in-stock" portion of the ZINC 15 dataset─a large dataset of commercially available chemicals. The fine-tuning task then perturbed the pre-trained embeddings to facilitate prediction of a specific QSAR endpoint of interest. The power of this model stems from the ability to reuse the pre-trained model for multiple different fine-tuning tasks, reducing the computational burden of developing multiple models for different endpoints. We used our framework to develop a predictive model for fraction unbound in human plasma (fu,p). This approach is flexible, requires minimum domain expertise, and can be generalized for other parameters of interest for rapid and accurate estimation of absorption, distribution, metabolism, excretion, and toxicity.

NoiseMol: A noise-robusted data augmentation via perturbing noise for molecular property prediction

Simplified Molecular-Input Line-Entry System (SMILES) is one of a widely used molecular representation methods for molecular property prediction. We conjecture that all the characters in the SMILES string of a molecule are essential for making up the molecules, but most of them make little contribution to determining a particular property of the molecule. Therefore, we verified the conjecture in the pre-experiment. Motivated by the result, we propose to inject proper noisy information into the SMILES to augment the training data by increasing the diversity of the labeled molecules. To this end, we explore injecting perturbing noise into the original labeled SMILES strings to construct augmented data for alleviating the limitation of the labeled compound data and enhancing the model to extract more useful molecular representation for molecular property prediction. Specifically, we directly adopt mask, swap, deletion, and fusion operations on SMILES strings to randomly mask, swap, and delete atoms in SMILES strings. Then, the augmented data is used by two strategies: each epoch alternately feeds the original and perturbing noisy molecules, or each batch alternately feeds the original and perturbing noisy molecules. We conduct experiments on both Transformer and BiGRU models to validate the effectiveness by adopting widely used datasets from MoleculeNet and ZINC. Experimental results demonstrate that the proposed method outperforms strong baselines on all the datasets. NoiseMol obtains the best performance on BBBP and FDA when compared with state-of-the-art methods. Besides, NoiseMol achieves the best accuracy on LogP. Therefore, injecting perturbing noise into the labeled SMILES strings is an effective and efficient method, which improves the prediction performance, generalization, and robustness of the deep learning models.

De Novo Direct Inverse QSPR/QSAR: Chemical Variational Autoencoder and Gaussian Mixture Regression Models.

Herein, we propose a de novo direct inverse quantitative structure-property relationship/quantitative structure-activity relationship (QSPR/QSAR) analysis method, based on the chemical variational autoencoder (VAE) and Gaussian mixture regression (GMR) models, to generate molecules with the desired target variables of interest for properties and activities (y). A data set of molecules was analyzed, and an encoder was used to transform the simplified molecular input line entry system (SMILES) strings to latent variables (x), while a decoder was used to transform x to SMILES strings. A chemical VAE model was used for analysis and a GMR model (between x and y) was constructed for direct inverse analysis. The target y values were input into the GMR model to directly predict the x values. Following this, the predicted x values were input into the decoder associated with the chemical VAE model and the SMILES string representations (or chemical structures of molecules) were obtained as the output, indicating that the proposed method could be used to selectively obtain the molecules that were characterized by the target y values. We confirmed that the proposed method can be used to generate molecules within the target y ranges even when the conventional chemical VAE model failed to generate the target molecules.

The prediction of molecular toxicity based on BiGRU and GraphSAGE

The prediction of molecules toxicity properties plays an crucial role in the realm of the drug discovery, since it can swiftly screen out the expected drug moleculars. The conventional method for predicting toxicity is to use some in vivo or in vitro biological experiments in the laboratory, which can easily pose a threat significant time and financial waste and even ethical issues. Therefore, using computational approaches to predict molecular toxicity has become a common strategy in modern drug discovery. In this article, we propose a novel model named MTBG, which primarily makes use of both SMILES (Simplified molecular input line entry system) strings and graph structures of molecules to extract drug molecular feature in the field of drug molecular toxicity prediction. To verify the performance of the MTBG model, we opt the Tox21 dataset and several widely used baseline models. Experimental results demonstrate that our model can perform better than these baseline models.

Application of Machine Learning in Developing Quantitative Structure-Property Relationship for Electronic Properties of Polyaromatic Compounds.

The degree of π orbital overlap (DPO) model has been demonstrated to be an excellent quantitative structure–property relationship (QSPR) that can map two-dimensional structural information of polycyclic aromatic hydrocarbons (PAHs) and thienoacenes to their electronic properties, namely, band gaps, electron affinities, and ionization potentials. However, the model suffers from significant limitations that narrow its applications due to inefficient manual procedures in parameter optimization and descriptor formulation. In this work, we developed a machine learning (ML)-based method for efficiently optimizing DPO parameters and proposed a truncated DPO descriptor, which is simple enough that can be automatically extracted from simplified molecular-input line-entry system strings of PAHs and thienoacenes. Compared with the result from our previous studies, the ML-based methodology can optimize DPO parameters with four times fewer data, while it can achieve the same level of accuracy in predictions of the mentioned electronic properties to within 0.1 eV. The truncated DPO model also has similar accuracy to the full DPO model. Consequently, the ML-based DPO approach coupled with the truncated DPO model enables new possibilities for developing automatic pipelines for high-throughput screening and investigating new QSPR for new chemical classes.

Evaluation of Deep Learning Architectures for Aqueous Solubility Prediction.

Determining the aqueous solubility of molecules is a vital step in many pharmaceutical, environmental, and energy storage applications. Despite efforts made over decades, there are still challenges associated with developing a solubility prediction model with satisfactory accuracy for many of these applications. The goals of this study are to assess current deep learning methods for solubility prediction, develop a general model capable of predicting the solubility of a broad range of organic molecules, and to understand the impact of data properties, molecular representation, and modeling architecture on predictive performance. Using the largest currently available solubility data set, we implement deep learning-based models to predict solubility from the molecular structure and explore several different molecular representations including molecular descriptors, simplified molecular-input line-entry system strings, molecular graphs, and three-dimensional atomic coordinates using four different neural network architectures—fully connected neural networks, recurrent neural networks, graph neural networks (GNNs), and SchNet. We find that models using molecular descriptors achieve the best performance, with GNN models also achieving good performance. We perform extensive error analysis to understand the molecular properties that influence model performance, perform feature analysis to understand which information about the molecular structure is most valuable for prediction, and perform a transfer learning and data size study to understand the impact of data availability on model performance.

Generative Adversarial Networks for De Novo Molecular Design

In the chemical industry, the generation of novel molecular structures with beneficial pharmacological and physicochemical properties in de novo molecular design is a critical problem. The advent of deep learning and neural generative models has recently enabled significant achievements in constructing molecular design models in de novo design. Consequently, studies on new generative models continue to generate molecules that exhibit more useful chemical properties. In this study, we propose a method for de novo design that utilizes generative adversarial networks based on reinforcement learning for realistic molecule generation. This method learns to reproduce the training data distribution of simplified molecular-input line-entry system strings. The proposed method is demonstrated to effectively generate novel molecular structures from five benchmark results using a real-world public dataset, ChEMBL. The code is available at https://github.com/dudwojae/SMILES-MaskGAN.

Drug-Target Interaction Prediction: End-to-End Deep Learning Approach.

The discovery of potential Drug-Target Interactions (DTIs) is a determining step in the drug discovery and repositioning process, as the effectiveness of the currently available antibiotic treatment is declining. Although putting efforts on the traditional in vivo or in vitro methods, pharmaceutical financial investment has been reduced over the years. Therefore, establishing effective computational methods is decisive to find new leads in a reasonable amount of time. Successful approaches have been presented to solve this problem but seldom protein sequences and structured data are used together. In this paper, we present a deep learning architecture model, which exploits the particular ability of Convolutional Neural Networks (CNNs) to obtain 1D representations from protein sequences (amino acid sequence) and compounds SMILES (Simplified Molecular Input Line Entry System) strings. These representations can be interpreted as features that express local dependencies or patterns that can then be used in a Fully Connected Neural Network (FCNN), acting as a binary classifier. The results achieved demonstrate that using CNNs to obtain representations of the data, instead of the traditional descriptors, lead to improved performance. The proposed end-to-end deep learning method outperformed traditional machine learning approaches in the correct classification of both positive and negative interactions.

Storage Consumption Reduction using Improved Inverted Indexing for Similarity Search on LINGO Profiles

Millions of compounds which exist in huge datasets are represented using Simplified Molecular-Input Line- Entry System (SMILES) representation. Fragmenting SMILES strings into overlapping substrings of a defined size called LINGO Profiles avoids the otherwise time-consuming conversion process. One drawback of this process is the generation of numerous identical LINGO Profiles. Introduced by Kristensen et al, the inverted indexing approach represents a modification intended to deal with the large number of molecules residing in the database. Implementing this technique effectively reduced the storage space requirement of the dataset by half, while also achieving significant speedup and a favourable accuracy value when performing similarity searching. This report presents an in-depth analysis of results, with conclusions about the effectiveness of the working prototype for this study.

Improving Chemical Autoencoder Latent Space and Molecular De Novo Generation Diversity with Heteroencoders.

Chemical autoencoders are attractive models as they combine chemical space navigation with possibilities for de novo molecule generation in areas of interest. This enables them to produce focused chemical libraries around a single lead compound for employment early in a drug discovery project. Here, it is shown that the choice of chemical representation, such as strings from the simplified molecular-input line-entry system (SMILES), has a large influence on the properties of the latent space. It is further explored to what extent translating between different chemical representations influences the latent space similarity to the SMILES strings or circular fingerprints. By employing SMILES enumeration for either the encoder or decoder, it is found that the decoder has the largest influence on the properties of the latent space. Training a sequence to sequence heteroencoder based on recurrent neural networks (RNNs) with long short-term memory cells (LSTM) to predict different enumerated SMILES strings from the same canonical SMILES string gives the largest similarity between latent space distance and molecular similarity measured as circular fingerprints similarity. Using the output from the code layer in quantitative structure activity relationship (QSAR) of five molecular datasets shows that heteroencoder derived vectors markedly outperforms autoencoder derived vectors as well as models built using ECFP4 fingerprints, underlining the increased chemical relevance of the latent space. However, the use of enumeration during training of the decoder leads to a marked increase in the rate of decoding to different molecules than encoded, a tendency that can be counteracted with more complex network architectures.

A novel methodology on distributed representations of proteins using their interacting ligands.

MotivationThe effective representation of proteins is a crucial task that directly affects the performance of many bioinformatics problems. Related proteins usually bind to similar ligands. Chemical characteristics of ligands are known to capture the functional and mechanistic properties of proteins suggesting that a ligand-based approach can be utilized in protein representation. In this study, we propose SMILESVec, a Simplified molecular input line entry system (SMILES)-based method to represent ligands and a novel method to compute similarity of proteins by describing them based on their ligands. The proteins are defined utilizing the word-embeddings of the SMILES strings of their ligands. The performance of the proposed protein description method is evaluated in protein clustering task using TransClust and MCL algorithms. Two other protein representation methods that utilize protein sequence, Basic local alignment tool and ProtVec, and two compound fingerprint-based protein representation methods are compared.ResultsWe showed that ligand-based protein representation, which uses only SMILES strings of the ligands that proteins bind to, performs as well as protein sequence-based representation methods in protein clustering. The results suggest that ligand-based protein description can be an alternative to the traditional sequence or structure-based representation of proteins and this novel approach can be applied to different bioinformatics problems such as prediction of new protein–ligand interactions and protein function annotation.Availability and implementation https://github.com/hkmztrk/SMILESVecProteinRepresentation Supplementary information Supplementary data are available at Bioinformatics online.

Towards the Development of Global Nano-Quantitative Structure-Property Relationship Models: Zeta Potentials of Metal Oxide Nanoparticles.

Zeta potential indirectly reflects a charge of the surface of nanoparticles in solutions and could be used to represent the stability of the colloidal solution. As processes of synthesis, testing and evaluation of new nanomaterials are expensive and time-consuming, so it would be helpful to estimate an approximate range of properties for untested nanomaterials using computational modeling. We collected the largest dataset of zeta potential measurements of bare metal oxide nanoparticles in water (87 data points). The dataset was used to develop quantitative structure–property relationship (QSPR) models. Essential features of nanoparticles were represented using a modified simplified molecular input line entry system (SMILES). SMILES strings reflected the size-dependent behavior of zeta potentials, as the considered quasi-SMILES modification included information about both chemical composition and the size of the nanoparticles. Three mathematical models were generated using the Monte Carlo method, and their statistical quality was evaluated (R2 for the training set varied from 0.71 to 0.87; for the validation set, from 0.67 to 0.82; root mean square errors for both training and validation sets ranged from 11.3 to 17.2 mV). The developed models were analyzed and linked to aggregation effects in aqueous solutions.

UManSysProp v1.0: an online and open-source facility for molecular property prediction and atmospheric aerosol calculations

Abstract. In this paper we describe the development and application of a new web-based facility, UManSysProp (http://umansysprop.seaes.manchester.ac.uk), for automating predictions of molecular and atmospheric aerosol properties. Current facilities include pure component vapour pressures, critical properties, and sub-cooled densities of organic molecules; activity coefficient predictions for mixed inorganic–organic liquid systems; hygroscopic growth factors and CCN (cloud condensation nuclei) activation potential of mixed inorganic–organic aerosol particles; and absorptive partitioning calculations with/without a treatment of non-ideality. The aim of this new facility is to provide a single point of reference for all properties relevant to atmospheric aerosol that have been checked for applicability to atmospheric compounds where possible. The group contribution approach allows users to upload molecular information in the form of SMILES (Simplified Molecular Input Line Entry System) strings and UManSysProp will automatically extract the relevant information for calculations. Built using open-source chemical informatics, and hosted at the University of Manchester, the facilities are provided via a browser and device-friendly web interface, or can be accessed using the user's own code via a JSON API (application program interface). We also provide the source code for all predictive techniques provided on the site, covered by the GNU GPL (General Public License) license to encourage development of a user community. We have released this via a Github repository (doi:10.5281/zenodo.45143). In this paper we demonstrate its use with specific examples that can be simulated using the web-browser interface.