Review
Chemistry, Multidisciplinary
Su-Qing Yang, Qing Ye, Jun-Jie Ding, Ming-Zhu Yin, Ai-Ping Lu, Xiang Chen, Ting-Jun Hou, Dong-Sheng Cao
Summary: Target identification for bioactive molecules is crucial in modern drug discovery, with computational methods being proposed and widely developed to accelerate the validation process. Ligand-based target prediction methods have made significant progress in the past decade, offering flexibility, low computational cost, and remarkable predictive performance.
WILEY INTERDISCIPLINARY REVIEWS-COMPUTATIONAL MOLECULAR SCIENCE
(2021)
Article
Biochemical Research Methods
Zhenxing Wu, Minfeng Zhu, Yu Kang, Elaine Lai-Han Leung, Tailong Lei, Chao Shen, Dejun Jiang, Zhe Wang, Dongsheng Cao, Tingjun Hou
Summary: A study on learning QSAR models using various ML algorithms for 14 public datasets showed that rbf-SVM, rbf-GPR, XGBoost, and DNN generally perform better than other algorithms. SVM and XGBoost are recommended for regression learning on small datasets, while XGBoost is an excellent choice for large datasets. Ensemble models integrating multiple algorithms can improve prediction accuracy.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Biochemistry & Molecular Biology
Zahra Bastami, Razieh Sheikhpour, Parvin Razzaghi, Ali Ramazani, Sajjad Gharaghani
Summary: Caspases are important enzymes involved in inflammation and cell death processes. This study used Proteochemometrics Modeling to summarize and predict the interactions between caspases and ligands. The ensemble model showed superior performance compared to other models.
MOLECULAR DIVERSITY
(2023)
Review
Pharmacology & Pharmacy
Jennifer M. Cantrell, Carolina H. Chung, Sriram Chandrasekaran
Summary: The text discusses the use of machine learning algorithms in designing combination therapies to combat antimicrobial resistance. It also compares different ML-based approaches based on the type of input information used and provides a compilation of relevant drug interaction datasets. Limitations of current methods are discussed, along with proposed strategies for enhancing efficacy in designing combination therapies.
DRUG DISCOVERY TODAY
(2022)
Review
Biochemical Research Methods
Axel Kowald, Israel Barrantes, Steffen Moeller, Daniel Palmer, Hugo Murua Escobar, Anne Schwerk, Georg Fuellen
Summary: Accurate transfer learning of clinical outcomes is highly useful for transferring prediction tasks from one cellular context to another. Transductive transfer learning focuses on learning the predictor in the source domain and transferring its label calculations to the target domain, while inductive transfer learning considers cases where the target predictor performs a different yet related task compared with the source predictor. Mapping variables may also be required. Various transfer learning approaches have been discussed and compared in the context of predicting clinical outcomes based on preclinical molecular data.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemistry & Molecular Biology
Ruifeng Liu, Srinivas Laxminarayan, Jaques Reifman, Anders Wallqvist
Summary: The main limitation in developing DNN models to predict bioactivity properties of chemicals is the lack of sufficient assay data. This study explores transfer learning by using the dense layers of a pre-trained DNN to develop models for other related or unrelated properties. The results show that transfer learning can significantly reduce prediction errors and training sample size depending on dataset correlation.
JOURNAL OF COMPUTER-AIDED MOLECULAR DESIGN
(2022)
Review
Engineering, Biomedical
Jacob Kerner, Alan Dogan, Horst von Recum
Summary: Machine learning has been widely utilized in various fields, including biomaterials, optimizing data collection and analysis. Recent advances in biomaterials have focused on quantitative structure properties relationships, introducing four basic models for rapid development and addressing the lack of machine learning implementation in the field. This article aims to spark greater interest and awareness in utilizing computational methods for biomaterials research.
ACTA BIOMATERIALIA
(2021)
Article
Engineering, Environmental
Jose Andres Cordero, Kai He, Kanjira Janya, Shinya Echigo, Sadahiko Itoh
Summary: This study used machine learning algorithms and chemical descriptors to predict the formation of haloacetic acids in drinking water, identified important predictors, and demonstrated a potential method for predicting precursors of other disinfection byproducts.
JOURNAL OF HAZARDOUS MATERIALS
(2021)
Article
Chemistry, Medicinal
Jannis Born, Tien Huynh, Astrid Stroobants, Wendy D. Cornell, Matteo Manica
Summary: The use of reduced active site sequence representation in kinase-ligand binding affinity prediction has shown significantly higher performance compared to using the full primary structure. This trend persists across different models, data sets, and performance metrics, and holds true when predicting pIC(50) for both unseen ligands and kinases.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2022)
Article
Chemistry, Applied
Katrina W. Lexa, Kevin M. Belyk, Jeremy Henle, Bangping Xiang, Robert P. Sheridan, Scott E. Denmark, Rebecca T. Ruck, Edward C. Sherer
Summary: Molecular design benefits from the partnership between chemical intuition and machine learning. In this study, different methods for modeling a catalyst system were investigated, and it was found that the 2D random forest approach outperformed other methods. With the passage of time, the model performance improved, and a high-throughput approach was crucial for large-scale optimization.
ORGANIC PROCESS RESEARCH & DEVELOPMENT
(2022)
Article
Biochemistry & Molecular Biology
Amelie Tjaden, Apirat Chaikuad, Eric Kowarz, Rolf Marschalek, Stefan Knapp, Martin Schroeder, Susanne Mueller
Summary: Phenotypical screening is a common method in drug discovery, but functional annotation of identified hits is often challenging. This study presents an optimized live-cell multiplexed assay for comprehensive time-dependent characterization of the effect of small molecules on cellular health.
Article
Biochemistry & Molecular Biology
Magdi E. A. Zaki, Sami A. Al-Hussain, Vijay H. Masand, Manoj K. Sabnani, Abdul Samad
Summary: A QSAR model was developed to explore safer anti-thrombotic drugs, revealing correlations between anti-thrombotic activity and concealed structural traits. The model captured reported and novel pharmacophoric features, validated by crystal structures of compounds with factor Xa, which could be beneficial for future lead compound optimization.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2021)
Article
Chemistry, Medicinal
Daiguo Deng, Xiaowei Chen, Ruochi Zhang, Zengrong Lei, Xiaojian Wang, Fengfeng Zhou
Summary: The study explores how to automatically extract chemical molecule features using graph neural networks (GNN) and build an accurate prediction model of molecular properties by combining with the XGBoost classifier. Experimental results suggest that the proposed XGraphBoost framework can facilitate efficient and accurate predictions of various molecular properties.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2021)
Article
Chemistry, Medicinal
Magdi E. A. Zaki, Sami A. Al-Hussain, Syed Nasir Abbas Bukhari, Vijay H. Masand, Mithilesh M. Rathore, Sumer D. Thakur, Vaishali M. Patil
Summary: Cancer is a major life-threatening disease with a high mortality rate in many countries. Chemotherapy is the preferred treatment option for patients, but serious side effects of anti-cancer drugs have led to a search for safer alternatives. This study developed a multi-linear QSAR model to identify crucial pharmacophoric features for developing an Hsp90 inhibitor. The results revealed that Hsp90 inhibitory activity is correlated with various nitrogen atoms and other structural features.
Article
Computer Science, Artificial Intelligence
Zhen Fang, Jie Lu, Feng Liu, Guangquan Zhang
Summary: Semi-supervised heterogeneous domain adaptation (SsHeDA) aims to train a classifier for the target domain with limited labeled and unlabeled data, by leveraging knowledge from a heterogeneous source domain. Although several methods have been proposed, there is still a lack of theoretical foundation to explain and guide better solutions for the SsHeDA problem.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Biochemical Research Methods
Karina Pikalyova, Alexey Orlov, Arkadii Lin, Olga Tarasova, Marcou Gilles, Dragos Horvath, Vladimir Poroikov, Alexandre Varnek
Summary: A new methodology based on generative topographic mapping (GTM) was introduced for predicting the drug resistance of HIV strains. The approach combines high accuracy and interpretability, allowing for visualization and analysis of sequence space and treatment optimization. Several case studies demonstrate the practicality of this method.
Article
Chemistry, Medicinal
Yuliana Zabolotna, Dmitriy M. Volochnyuk, Sergey Ryabukhin, Kostiantyn Gavrylenko, Dragos Horvath, Olga Klimchuk, Oleksandr Oksiuta, Gilles Marcou, Alexandre Varnek
Summary: Most existing computational tools for de novo library design focus on generating, selecting, and combining structural motifs to form new library members. However, these approaches appear to be more theoretical and disconnected from reality due to the lack of a direct link between the chemical space of the retrosynthesized fragments and the pool of available reagents. This paper presents a new open-source toolkit called Synthons Interpreter (SynthI), which merges these two chemical spaces into a single synthons space.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2022)
Article
Chemistry, Medicinal
Yuliana Zabolotna, Dmitriy M. Volochnyuk, Sergey V. Ryabukhin, Dragos Horvath, Konstantin S. Gavrilenko, Gilles Marcou, Yurii S. Moroz, Oleksandr Oksiuta, Alexandre Varnek
Summary: Efficient synthesis of desired compounds is crucial for chemical space exploration in drug discovery, which is influenced by both established synthetic protocols and the availability of corresponding building blocks (BBs). This study analyzes the chemical space of 400,000 purchasable BBs, examining their physicochemical properties and diversity to assess their coverage of medicinal chemistry needs. The analysis is based on a universal topographic map that visualizes libraries and their differences in coverage.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2022)
Article
Chemistry, Medicinal
Regina Pikalyova, Yuliana Zabolotna, Dmitriy M. Volochnyuk, Dragos Horvath, Gilles Marcou, Alexandre Varnek
Summary: DNA-Encoded Library (DEL) technology is a method for discovering bioactive molecules in medicinal chemistry. This project aimed to generate and analyze an ultra-large chemical space of DEL using commercially available building blocks. The study compared the DEL compounds to biologically relevant compounds from ChEMBL and identified optimal DELs covering the chemical space of ChEMBL. Different combinations of DELs were analyzed to achieve even higher coverage of ChEMBL than with a single DEL.
MOLECULAR INFORMATICS
(2022)
Article
Chemistry, Multidisciplinary
Zlata Boiarska, Helena Perez-Pena, Anne-Catherine Abel, Paola Marzullo, Beatriz Alvarez-Bernad, Francesca Bonato, Benedetta Santini, Dragos Horvath, Daniel Lucena-Agell, Francesca Vasile, Maurizio Sironi, J. Fernando Diaz, Andrea E. Prota, Stefano Pieraccini, Daniele Passarella
Summary: Maytansinoids, a potent class of tubulin binders with cytotoxic activity, have been limited in their application as cytotoxins and chemical probes due to the complexity of natural product chemistry. In this study, the synthesis of long-chain derivatives and maytansinoid conjugates was reported, confirming that bulky substituents do not affect their activity or binding mode. These results provide new opportunities for the design of maytansine-based probes.
CHEMISTRY-A EUROPEAN JOURNAL
(2023)
Article
Chemistry, Medicinal
Polina Oleneva, Yuliana Zabolotna, Dragos Horvath, Gilles Marcou, Fanny Bonachera, Alexandre Varnek
Summary: The Chimiotheque Nationale (CN) was compared with ZINC and ChEMBL to analyze its screening and biologically relevant compounds, including chemical space coverage, physicochemical properties, and Bemis-Murcko scaffold populations. Over 5 K CN-unique scaffolds were identified. Generative Topographic Maps (GTMs) were generated to compare compound populations. Hierarchical GTM (<< zooming >>) was used to create an ensemble of maps at different resolutions, from global overview to individual structure mapping. These maps were added to the ChemSpace Atlas website. The analysis of synthetic accessibility showed that only 29.7% of CN compounds can be fully synthesized using commercially available building blocks.
MOLECULAR INFORMATICS
(2023)
Review
Biochemistry & Molecular Biology
Helena Perez-Pena, Anne-Catherine Abel, Maxim Shevelev, Andrea E. E. Prota, Stefano Pieraccini, Dragos Horvath
Summary: Microtubules are essential in cellular processes and have potential as targets for cancer and neurodegeneration research. However, current tubulin binders have limitations, making the discovery of safer and more efficient agents necessary. Computer-aided design techniques and accessible tubulin-ligand structures can aid in the selection and design of new tubulin-targeting agents.
Article
Chemistry, Medicinal
Giuseppe Lamanna, Pietro Delre, Gilles Marcou, Michele Saviano, Alexandre Varnek, Dragos Horvath, Giuseppe Felice Mangiatordi
Summary: This study introduces a new de novo design algorithm called GENERA that combines the capabilities of a deep-learning algorithm for automated drug-like analogue design, called DeLA-Drug, with a genetic algorithm for generating molecules with desired target-oriented properties. GENERA was applied to the angiotensin-converting enzyme 2 (ACE2) target, and its ability to de novo design promising candidates was assessed using docking programs PLANTS and GLIDE. The study demonstrates that GENERA can effectively perform multiobjective optimization and generate focused libraries with better scores compared to a starting set of known ACE-2 binders.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Chemistry, Medicinal
Regina Pikalyova, Yuliana Zabolotna, Dragos Horvath, Gilles Marcou, Alexandre Varnek
Summary: The development of DNA-encoded library (DEL) technology has brought new challenges to the analysis of chemical libraries. This study introduces the concept of chemical library space (CLS) and compares four representations obtained using generative topographic mapping. These encodings allow for effective comparison of libraries and fine-tuning of matching criteria. The proposed CLS can be used for efficient analysis and selection of chemical libraries.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Chemistry, Medicinal
Regina Pikalyova, Yuliana Zabolotna, Dragos Horvath, Gilles Marcou, Alexandre Varnek
Summary: In chemical library analysis, it can be beneficial to describe libraries as individual items rather than collections of compounds. This is especially true for large non-selectable compound mixtures like DNA-encoded libraries (DELs). The chemical library space (CLS) is useful for managing a portfolio of libraries, similar to how chemical space (CS) helps manage portfolios of molecules. Mapping the CLS on meta-GTMs allows for analysis beyond pairwise library comparison, facilitating the selection of the most suitable libraries for specific projects.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Biochemistry & Molecular Biology
Julia Revillo Imbernon, Celien Jacquemard, Guillaume Bret, Gilles Marcou, Esther Kellenberger
Summary: The screening of fragment libraries is crucial in drug discovery, with the success depending on the quality and design of the library meeting specific research requirements. This study conducted an inventory of commercial fragment libraries and developed a methodology to classify any library based on its similarity, coverage, and structural features, leading to the creation of a model that considers fragment diversity and ease of interpretation.
RSC MEDICINAL CHEMISTRY
(2022)