Journal of Cheminformatics最新文献

Computer-aided drug discovery (CADD) is nurtured by late advances in big data analytics and Artificial Intelligence (AI) towards enhanced drug discovery (DD) outcomes. In this context, reliable datasets are of utmost importance. We herein present CidalsDB a novel web server for AI-assisted DD against infectious pathogens, namely Leishmania parasites and Coronaviruses. We performed a literature search on molecules with validated anti-pathogen effects. Then, we consolidated these data with bioassays from PubChem. Finally, we constructed a database to store these datasets and make them accessible and ready-to-use for the scientific community through CidalsDB, a web-based interface. In a second step, we implemented and optimized four machine learning (ML) and three deep learning (DL) algorithms that optimally predicted the biological activity of molecules. Random Forests (RF), Multi-Layer Perceptron (MLP) and ChemBERTa were the best classifiers of anti-Leishmania molecules, while Gradient Boosting (GB), Graph-Convolutional Network (GCN) and ChemBERTa achieved the best performances on the Coronaviruses dataset. All six models were optimized and deployed through CidalsDB as anti-pathogen activity prediction models.

Scientific contribution

CidalsDB is an open access web-based tool that allows browsing and access to ready-to-use datasets of anti-pathogen molecules, alongside best performing AI models for biological activity prediction. It offers a democratized no-code platform for AI-based CADD, which shall foster innovation and collaboration within the DD community. CidalsDB is accessible through https://cidalsdb.streamlit.app/.

The exploration of chemical space holds promise for developing influential chemical entities. Molecular representations, which reflect features of molecular structure in silico, assist in navigating chemical space appropriately. Unlike atom-level molecular representations, such as SMILES and atom graph, which can sometimes lead to confusing interpretations about chemical substructures, substructure-level molecular representations encode important substructures into molecular features; they not only provide more information for predicting molecular properties and drug‒drug interactions but also help to interpret the correlations between molecular properties and substructures. However, it remains challenging to represent the entire molecular structure both intactly and simply with substructure-level molecular representations. In this study, we developed a novel substructure-level molecular representation and named it a group graph. The group graph offers three advantages: (a) the substructure of the group graph reflects the diversity and consistency of different molecular datasets; (b) the group graph retains molecular structural features with minimal information loss because the graph isomorphism network (GIN) of the group graph performs well in molecular properties and drug‒drug interactions prediction, showing higher accuracy and efficiency than the model of other molecular graphs, even without any pretraining; and (c) the molecular property may change when the substructure is substituted with another of differing importance in group graph, facilitating the detection of activity cliffs. In addition, we successfully predicted structural modifications to improve blood‒brain barrier permeability (BBBP) via the GIN of group graph. Therefore, the group graph takes advantages for simultaneously representing molecular local characteristics and global features.

Scientific contribution The group graph, as a substructure-level molecular representation, has the ability to retain molecular structural features with minimal information loss. As a result, it shows superior performance in predicting molecular properties and drug‒drug interactions with enhanced efficiency and interpretability.

With the advent of artificial intelligence (AI), it is now possible to design diverse and novel molecules from previously unexplored chemical space. However, a challenge for chemists is the synthesis of such molecules. Recently, there have been attempts to develop AI models for retrosynthesis prediction, which rely on the availability of a high-quality training dataset. In this work, we explore the suitability of large language models (LLMs) for extraction of high-quality chemical reaction data from patent documents. A comparative study on the same set of patents from an earlier study showed that the proposed automated approach can enhance the current datasets by addition of 26% new reactions. Several challenges were identified during reaction mining, and for some of them alternative solutions were proposed. A detailed analysis was also performed wherein several wrong entries were identified in the previously curated dataset. Reactions extracted using the proposed pipeline over a larger patent dataset can improve the accuracy and efficiency of synthesis prediction models in future.

Scientific contribution

In this work we evaluated the suitability of large language models for mining a high-quality chemical reaction dataset from patent literature. We showed that the proposed approach can significantly improve the quantity of the reaction database by identifying more chemical reactions and improve the quality of the reaction database by correcting previous errors/false positives.

In this work, inspired by the graph transformer, we presented an improved protocol, termed GT-NMR, which integrates 2D molecular graph representation with Transformer architecture, for accurate yet efficient prediction of NMR chemical shifts. The effectiveness of the GT-NMR was thoroughly examined with the standard nmrshiftdb2 dataset, 37 natural products and structural elucidation of 11 pairs of natural products. Systematical analysis affirms that GT-NMR outperforms traditional graph-based methods in all aspects, achieving state-of-the-art performance, with the mean absolute error of 0.158 and 1.189 ppm in predicting ¹H and ¹³C NMR chemical shifts, respectively, for the standard nmrshiftdb2 dataset. Further scrutiny of its practical applications indicates that GT-NMR's efficacy is closely tied to molecular complexity, as quantified by the size-normalized spatial score (nSPS). For relatively simple molecules (nSPS < = 27.71), GT-NMR performs comparably to the best density functional while its effectiveness significantly diminishes with complex molecules characterized by higher nSPS values (nSPS > = 38.42). This trend is consistent across other graph-based NMR chemical shift prediction methods as well. Therefore, while employing GT-NMR or other graph-based methods for the rapid and routine prediction of NMR chemical shifts, it is advisable to utilize nSPS to assess their suitability. The source codes and trained model of GT-NMR are publicly available at GitHub.

Scientific contribution

GT-NMR, which combines the 2D molecular graph representation with the Transformer architecture, was implemented for the first time to predict atom-level NMR chemical shifts, achieving state-of-the-art performance. More importantly, the reliability of the GT-NMR and graph-based methods was assessed for the first time in terms of molecular complexity, as quantified by the size-normalized spacial score (nSPS). Systematical scrutiny demonstrated that GT-NMR offer a valuable way for routine application in structural screening and elucidation of relatively simple molecules.

Non–Contact Atomic Force Microscopy with CO–functionalized metal tips (referred to as HR-AFM) provides access to the internal structure of individual molecules adsorbed on a surface with totally unprecedented resolution. Previous works have shown that deep learning (DL) models can retrieve the chemical and structural information encoded in a 3D stack of constant-height HR–AFM images, leading to molecular identification. In this work, we overcome their limitations by using a well-established description of the molecular structure in terms of topological fingerprints, the 1024–bit Extended Connectivity Chemical Fingerprints of radius 2 (ECFP4), that were developed for substructure and similarity searching. ECFPs provide local structural information of the molecule, each bit correlating with a particular substructure within the molecule. Our DL model is able to extract this optimized structural descriptor from the 3D HR–AFM stacks and use it, through virtual screening, to identify molecules from their predicted ECFP4 with a retrieval accuracy on theoretical images of 95.4%. Furthermore, this approach, unlike previous DL models, assigns a confidence score, the Tanimoto similarity, to each of the candidate molecules, thus providing information on the reliability of the identification. By construction, the number of times a certain substructure is present in the molecule is lost during the hashing process, necessary to make them useful for machine learning applications. We show that it is possible to complement the fingerprint-based virtual screening with global information provided by another DL model that predicts from the same HR–AFM stacks the chemical formula, boosting the identification accuracy up to a 97.6%. Finally, we perform a limited test with experimental images, obtaining promising results towards the application of this pipeline under real conditions.

Scientific contribution

Previous works on molecular identification from AFM images used chemical descriptors that were intuitive for humans but sub–optimal for neural networks. We propose a novel method to extract the ECFP4 from AFM images and identify the molecule via a virtual screening, beating previous state-of-the-art models.

Discovering new chemical compounds with specific properties can provide advantages for fields that rely on materials for their development, although this task comes at a high cost in terms of complexity and resources. Since the beginning of the data age, deep learning techniques have revolutionized the process of designing molecules by analyzing and learning from representations of molecular data, greatly reducing the resources and time involved. Various deep learning approaches have been developed to date, using a variety of architectures and strategies, in order to explore the extensive and discontinuous chemical space, providing benefits for generating compounds with specific properties. In this study, we present a systematic review that offers a statistical description and comparison of the strategies utilized to generate molecules through deep learning techniques, utilizing the metrics proposed in Molecular Sets (MOSES) or Guacamol. The study included 48 articles retrieved from a query-based search of Scopus and Web of Science and 25 articles retrieved from citation search, yielding a total of 72 retrieved articles, of which 62 correspond to chemical language models approaches to molecule generation and other 10 retrieved articles correspond to molecular graph representations. Transformers, recurrent neural networks (RNNs), generative adversarial networks (GANs), Structured Space State Sequence (S4) models, and variational autoencoders (VAEs) are considered the main deep learning architectures used for molecule generation in the set of retrieved articles. In addition, transfer learning, reinforcement learning, and conditional learning are the most employed techniques for biased model generation and exploration of specific chemical space regions. Finally, this analysis focuses on the central themes of molecular representation, databases, training dataset size, validity-novelty trade-off, and performance of unbiased and biased chemical language models. These themes were selected to conduct a statistical analysis utilizing graphical representation and statistical tests. The resulting analysis reveals the main challenges, advantages, and opportunities in the field of chemical language models over the past four years.

Building reliable and robust quantitative structure–property relationship (QSPR) models is a challenging task. First, the experimental data needs to be obtained, analyzed and curated. Second, the number of available methods is continuously growing and evaluating different algorithms and methodologies can be arduous. Finally, the last hurdle that researchers face is to ensure the reproducibility of their models and facilitate their transferability into practice. In this work, we introduce QSPRpred, a toolkit for analysis of bioactivity data sets and QSPR modelling, which attempts to address the aforementioned challenges. QSPRpred’s modular Python API enables users to intuitively describe different parts of a modelling workflow using a plethora of pre-implemented components, but also integrates customized implementations in a “plug-and-play” manner. QSPRpred data sets and models are directly serializable, which means they can be readily reproduced and put into operation after training as the models are saved with all required data pre-processing steps to make predictions on new compounds directly from SMILES strings. The general-purpose character of QSPRpred is also demonstrated by inclusion of support for multi-task and proteochemometric modelling. The package is extensively documented and comes with a large collection of tutorials to help new users. In this paper, we describe all of QSPRpred’s functionalities and also conduct a small benchmarking case study to illustrate how different components can be leveraged to compare a diverse set of models. QSPRpred is fully open-source and available at https://github.com/CDDLeiden/QSPRpred.

Scientific Contribution

QSPRpred aims to provide a complex, but comprehensive Python API to conduct all tasks encountered in QSPR modelling from data preparation and analysis to model creation and model deployment. In contrast to similar packages, QSPRpred offers a wider and more exhaustive range of capabilities and integrations with many popular packages that also go beyond QSPR modelling. A significant contribution of QSPRpred is also in its automated and highly standardized serialization scheme, which significantly improves reproducibility and transferability of models.

Background

Target identification and hit identification can be transformed through the application of biomedical knowledge analysis, AI-driven virtual screening and robotic cloud lab systems. However there are few prospective studies that evaluate the efficacy of such integrated approaches.

Results

We synergistically integrate our in-house-developed target evaluation (SpectraView) and deep-learning-driven virtual screening (HydraScreen) tools with an automated robotic cloud lab designed explicitly for ultra-high-throughput screening, enabling us to validate these platforms experimentally. By employing our target evaluation tool to select IRAK1 as the focal point of our investigation, we prospectively validate our structure-based deep learning model. We can identify 23.8% of all IRAK1 hits within the top 1% of ranked compounds. The model outperforms traditional virtual screening techniques and offers advanced features such as ligand pose confidence scoring. Simultaneously, we identify three potent (nanomolar) scaffolds from our compound library, 2 of which represent novel candidates for IRAK1 and hold promise for future development.

Conclusion

This study provides compelling evidence for SpectraView and HydraScreen to provide a significant acceleration in the processes of target identification and hit discovery. By leveraging Ro5’s HydraScreen and Strateos’ automated labs in hit identification for IRAK1, we show how AI-driven virtual screening with HydraScreen could offer high hit discovery rates and reduce experimental costs.

Scientific contribution

We present an innovative platform that leverages Knowledge graph-based biomedical data analytics and AI-driven virtual screening integrated with robotic cloud labs. Through an unbiased, prospective evaluation we show the reliability and robustness of HydraScreen in virtual and high-throughput screening for hit identification in IRAK1. Our platforms and innovative tools can expedite the early stages of drug discovery.

The accurate identification of protein–ligand binding sites is of critical importance in understanding and modulating protein function. Accordingly, ligand binding site prediction has remained a research focus for over three decades with over 50 methods developed and a change of paradigm from geometry-based to machine learning. In this work, we collate 13 ligand binding site predictors, spanning 30 years, focusing on the latest machine learning-based methods such as VN-EGNN, IF-SitePred, GrASP, PUResNet, and DeepPocket and compare them to the established P2Rank, PRANK and fpocket and earlier methods like PocketFinder, Ligsite and Surfnet. We benchmark the methods against the human subset of our new curated reference dataset, LIGYSIS. LIGYSIS is a comprehensive protein–ligand complex dataset comprising 30,000 proteins with bound ligands which aggregates biologically relevant unique protein–ligand interfaces across biological units of multiple structures from the same protein. LIGYSIS is an improvement for testing methods over earlier datasets like sc-PDB, PDBbind, binding MOAD, COACH420 and HOLO4K which either include 1:1 protein–ligand complexes or consider asymmetric units. Re-scoring of fpocket predictions by PRANK and DeepPocket display the highest recall (60%) whilst IF-SitePred presents the lowest recall (39%). We demonstrate the detrimental effect that redundant prediction of binding sites has on performance as well as the beneficial impact of stronger pocket scoring schemes, with improvements up to 14% in recall (IF-SitePred) and 30% in precision (Surfnet). Finally, we propose top-N+2 recall as the universal benchmark metric for ligand binding site prediction and urge authors to share not only the source code of their methods, but also of their benchmark.

Scientific contributions

This study conducts the largest benchmark of ligand binding site prediction methods to date, comparing 13 original methods and 15 variants using 10 informative metrics. The LIGYSIS dataset is introduced, which aggregates biologically relevant protein–ligand interfaces across multiple structures of the same protein. The study highlights the detrimental effect of redundant binding site prediction and demonstrates significant improvement in recall and precision through stronger scoring schemes. Finally, top-N+2 recall is proposed as a universal benchmark metric for ligand binding site prediction, with a recommendation for open-source sharing of both methods and benchmarks.