Huge Chemical Space Research Articles

Exploring the unexplored chemical space: Rational identification of new Tafenoquine analogs with antimalarial properties

Patents tend to define a huge chemical space described by the combinatorial nature of Markush structures. However, the optimization of new principal active ingredient is frequently driven by a simple Free Wilson approach. This procedure leads to a highly focused study on the chemical space near a hit compound leaving many unexplored regions that may present highly biological active reservoirs. This study aims to demonstrate that this unveiled chemical space can hide compounds with interesting potential biological activity that would be worth pursuing. This underlines the value and necessity of broadening an approach beyond conventional strategies. Hence, we advocate for an alternative methodology that may be more efficient in the early drug discovery stages. We have selected the case of Tafenoquine, a single-dose treatment for the radical cure of P. vivax malaria approved by the FDA in 2018, as an example to illustrate the process. Through the deep exploration of the Tafenoquine chemical space, seven compounds with potential antimalarial activity have been rationally identified and synthesized. This small set is representative of the chemical diversity unexplored by the 58 analogs reported to date. After biological assessment, results evidence that our approach for rational design has proven to be a very efficient exploratory methodology suitable for the early drug discovery stages.

Machine learning for prediction of CO2/N2/H2O selective adsorption and separation in metal-zeolites

Carbon dioxide (CO2) capture, utilization, and storage technologies are crucial in reducing global warming and producing various high-value chemicals and fuels. It is challenging to effectively separate the molecules of CO2, N2, and H2O, whose kinetic diameters are close to each other. Although zeolites have garnered considerable attention in gas separation, the huge chemical space of metal-doped zeolites (metal-zeolites) coming from the combination of different metal active sites, topology, and Si/Al ratios poses a difficulty in finding an optimal material for selectively trapping CO2. In this study, we build machine learning (ML) models to predict the selective adsorption of CO2/N2/H2O on metal-zeolites through the regulation of electrostatic polarization interaction. The stability of 208 metal-zeolites encompassing five distinct topological structures is estimated through the formation energy (E f), indicating the potential accessibility in the experiment for most of the studied systems, especially for Sc-, Y-, and Zr-zeolites. Adsorption of CO2 on metal sites has two possible configurations: linear vs. bent CO2, depending on different embedded metals. The concerted binding of CO2 with both carbon and oxygen atoms on the metal center leads to the bent geometry and larger binding energies on metal-zeolites (Zr-, Nb-, Mo-zeolites). Accessible descriptors associated with the zeolites, adsorbates, and metals are selected to train the adsorption strength index (I ), showing good performance [mean absolute error (MAE) = 0.04, R2 = 0.88]. The predicted adsorption selectivity is in agreement with the experimental systems (Co-, Zn-, Cu-, Fe-SSZ-13). It is found that medium-pore-sized zeolites [pore limiting diameter (PLD) = ~7 Å] anchored with Zr, Nb, or Mo are promising materials for the CO2 adsorption and separation. The proposed ML scheme may also be applicable to give a fast prediction of CO2 adsorption and separation ability in other porous metal-organic frameworks or amorphous materials.

Glycoproteomics of a Single Protein: Revealing Tens of Thousands of Myozyme Glycoforms by Hybrid HPLC-MS Approaches

Characterization of highly glycosylated biopharma-ceuticals by mass spectrometry is challenging because of the huge chemical space of coexistent glycoforms present. Here, we report the use of an array of HPLC-mass spectrometry-based approaches at different structural levels of released glycan, glycopeptide, and hitherto unexplored intact glycoforms to scrutinize the biopharmaceutical Myozyme, containing the highly complex lysosomal enzyme recombinant acid α-glucosidase. The intrinsic heterogeneity of recombinant acid α-glucosidase glycoforms was unraveled using a novel strong anion exchange HPLC-mass spectrometry approach involving a pH-gradient of volatile buffers to facilitate chromatographic separation of glycoforms based on their degree of sialylation, followed by the acquisition of native mass spectra in an Orbitrap mass spectrometer. Upon considering the structures of 60 different glycans attached to seven glycosylation sites in the intact protein, the large set of interdependent data acquired at different structural levels was integrated using a set of bioinformatic tools and allowed the annotation of intact glycoforms unraveling more than 1,000,000 putative intact glycoforms. Detectable isoforms also included several mannose-6-phosphate variants, which are essential for directing the drug toward its target, the lysosomes. Finally, for the first time, we sought to validate the intact glycoform annotations by integrating experimental data on the enzymatically dissected proteoforms, which reduced the number of glycoforms supported by experimental evidence to 42,104. The latter verification clearly revealed the strengths but also intrinsic limitations of this approach for fully characterizing such highly complex glycoproteins by mass spectrometry.

Distilling Accurate Descriptors from Multi-Source Experimental Data for Discovering Highly Active Perovskite OER Catalysts.

Perovskite oxides are promising catalysts for the oxygen evolution reaction, yet the huge chemical space remains largely unexplored due to the lack of effective approaches. Here, we report the distilling of accurate descriptors from multi-source experimental data for accelerated catalyst discovery by using the newly designed method of sign-constrained multi-task learning within the framework of sure independence screening and sparsifying operator that overcomes the challenge of data inconsistency between different sources. While many previous descriptors for the catalytic activity were proposed based on respective small data sets, we obtained a new 2D descriptor (dB, nB) based on 13 experimental data sets collected from different publications. Great universality and predictive accuracy, and the bulk-surface correspondence, of this descriptor have been demonstrated. With this descriptor, hundreds of unreported candidate perovskites with activity greater than the benchmark catalyst Ba0.5Sr0.5Co0.8Fe0.2O3 were identified from a large chemical space. Our experimental validations on five candidates confirmed the three highly active perovskite catalysts SrCo0.6Ni0.4O3, Rb0.1Sr0.9Co0.7Fe0.3O3, and Cs0.1Sr0.9Co0.4Fe0.6O3. This work provides an important new approach in dealing with inconsistent multi-source data for applications in the field of data-driven catalysis and beyond.

The importance of a charge transfer descriptor for screening potential CO2 reduction electrocatalysts

It has been over twenty years since the linear scaling of reaction intermediate adsorption energies started to coin the fields of heterogeneous and electrocatalysis as a blessing and a curse at the same time. It has established the possibility to construct activity volcano plots as a function of a single or two readily accessible adsorption energies as descriptors, but also limited the maximal catalytic conversion rate. In this work, it is found that these established adsorption energy-based descriptor spaces are not applicable to electrochemistry, because they are lacking an important additional dimension, the potential of zero charge. This extra dimension arises from the interaction of the electric double layer with reaction intermediates which does not scale with adsorption energies. At the example of the electrochemical reduction of CO2 it is shown that the addition of this descriptor breaks the scaling relations, opening up a huge chemical space that is readily accessible via potential of zero charge-based material design. The potential of zero charge also explains product selectivity trends of electrochemical CO2 reduction in close agreement with reported experimental data highlighting its importance for electrocatalyst design.

Deep multi-agent fusion Q-Network for graph generation

Deep Generative models for graph generation have made important advances in medicine, and particularly in drug research. Using graph representations of the molecules, these models proved efficient for predicting new molecular drug candidates with chemically valid structures. The main objective of the molecular graph generation task is to provide new molecular structures with the required chemical properties. Using a Reinforcement Learning (RL) agent to conduct the generation process while defining the reward function as the target chemical property, enables to generate new molecular drug candidates. However, sampling molecular candidates from the huge chemical space can be challenging, if only one single RL agent is deployed. Thus, we propose the Deep Fusion Q-Network (DFQN), a multi-agent RL-based Deep Neural Network for graph generation. We disclose that DFQN allows a better exploration of the environment, since multiple RL agents are exploited to generate candidate graph structures. Using a heterogeneous graph representation and an attention based fusion network, DFQN promotes coordination and communication between agents to jointly generate new graph structures with the required target properties. We investigate the optimal number of RL-agents using different degrees of communication in order to generate new graph structures under different property constraints. Moreover, we propose to conduct an adversarial training to incorporate prior knowledge of realistic drug molecules in the newly generated molecular graphs. We propose a new approach to check the chemical validity of the designed molecules in each iteration of the generation process. We evaluate the performance of our model for generating new molecules with different chemical properties and achieve remarkable results compared to state-of-the-art models for molecular graph generation.

Gex2SGen: Designing Drug-like Molecules from Desired Gene Expression Signatures

Drug-induced gene expression profiling provides a lot of useful information covering various aspects of drug discovery and development. Most importantly, this knowledge can be used to discover drugs' mechanisms of action. Recently, deep learning-based drug design methods are in the spotlight due to their ability to explore huge chemical space and design property-optimized target-specific drug molecules. Recent advances in accessibility of open-source drug-induced transcriptomic data along with the ability of deep learning algorithms to understand hidden patterns have opened opportunities for designing drug molecules based on desired gene expression signatures. In this study, we propose a deep learning model, Gex2SGen (Gene Expression 2 SMILES Generation), to generate novel drug-like molecules based on desired gene expression profiles. The model accepts desired gene expression profiles in a cell-specific manner as input and designs drug-like molecules which can elicit the required transcriptomic profile. The model was first tested against individual gene-knocked-out transcriptomic profiles, where the newly designed molecules showed high similarity with known inhibitors of the knocked-out target genes. The model was next applied on a triple negative breast cancer signature profile, where it could generate novel molecules, highly similar to known anti-breast cancer drugs. Overall, this work provides a generalized method, where the method first learned the molecular signature of a given cell due to a specific condition, and designs new small molecules with drug-like properties.

Adding open spectral data to MassBank and PubChem using open source tools to support non-targeted exposomics of mixtures.

The term "exposome" is defined as a comprehensive study of life-course environmental exposures and the associated biological responses. Humans are exposed to many different chemicals, which can pose a major threat to the well-being of humanity. Targeted or non-targeted mass spectrometry techniques are widely used to identify and characterize various environmental stressors when linking exposures to human health. However, identification remains challenging due to the huge chemical space applicable to exposomics, combined with the lack of sufficient relevant entries in spectral libraries. Addressing these challenges requires cheminformatics tools and database resources to share curated open spectral data on chemicals to improve the identification of chemicals in exposomics studies. This article describes efforts to contribute spectra relevant for exposomics to the open mass spectral library MassBank (https://www.massbank.eu) using various open source software efforts, including the R packages RMassBank and Shinyscreen. The experimental spectra were obtained from ten mixtures containing toxicologically relevant chemicals from the US Environmental Protection Agency (EPA) Non-Targeted Analysis Collaborative Trial (ENTACT). Following processing and curation, 5582 spectra from 783 of the 1268 ENTACT compounds were added to MassBank, and through this to other open spectral libraries (e.g., MoNA, GNPS) for community benefit. Additionally, an automated deposition and annotation workflow was developed with PubChem to enable the display of all MassBank mass spectra in PubChem, which is rerun with each MassBank release. The new spectral records have already been used in several studies to increase the confidence in identification in non-target small molecule identification workflows applied to environmental and exposomics research.

New Insights into the Structural Requirements of Isatin-Derived Pro-Apoptotic Agents against Acute Myeloid Leukemia.

Searching for bioactive compounds within the huge chemical space is like trying to find a needle in a haystack. Isatin is a unique natural compound which is endowed with different bio-pertinent activities, especially in cancer therapy. Herein, we envisaged that adopting a hybrid strategy of isatin and α,β-unsaturated ketone would afford new chemical entities with strong chemotherapeutic potential. Of interest, compounds 5b and 5g demonstrated significant antiproliferative activities against different cancer genotypes according to NCI-60 screening. Concomitantly, their IC50 against HL-60 cells were 0.38 ± 0.08 and 0.57 ± 0.05 µM, respectively, demonstrating remarkable apoptosis and moderate cell cycle arrest at G1 phase. Intriguingly, an impressive safety profile for 5b was reflected by a 37.2 times selectivity against HL-60 over PBMC from a healthy donor. This provoked us to further explore their mechanism of action by in vitro and in silico tools. Conclusively, 5b and 5g stand out as strong chemotherapeutic agents that hold clinical promise against acute myeloid leukemia.

Improved graph‐based multitask learning model with sparse sharing for quantitative structure–property relationship prediction of drug molecules

AbstractThe quantitative structure–property relationship (QSPR) is a fundamental technique for evaluating and screening potentially valuable molecules in the field of drug discovery. There is an urgent need to speed up pharmaceutical research and development and a huge chemical space to explore, which necessitate effective and precise computer‐aided QSPR modeling methods. Previous studies with various deep learning models are limited because they are trained on separate small datasets, known as the small‐sample problem. Using transfer learning, this article describes a sparse sharing method that uses advanced graph‐based models to construct an efficient and reasonable multitask learning workflow for QSPR prediction. The proposed workflow is systematically and comprehensively tested with four benchmark datasets containing different targets, and several precisely predicted molecular examples are illustrated. The results demonstrate that an obvious improvement in the prediction of molecular properties is achieved, along with the ability to predict multiple properties simultaneously.

Graph Neural Tree: A novel and interpretable deep learning-based framework for accurate molecular property predictions

Determining various properties of molecules is a critical step in drug discovery. Recently, with the improvement of large heterogeneous datasets and the development of deep learning approaches, more and more scientists have turned their attention to neural network-based virtual preliminary screening to reduce the time and monetary cost of drug discovery. However, the poor interpretability of deep learning masks causality, so models' conclusions are often beyond the comprehension of human users, which reduces the credibility of the model and makes it difficult for chemists to further narrow the huge chemical space based on models’ results. Thus, this study develops a novel framework consisting of Graph Neural Networks for feature extraction, Curriculum-Based Learning Strategies for optimization, and a Learning Binary Neural Tree (LBNT) for prediction, to improve the performance of neural networks and reveal their decision-making process to chemists. The framework encodes molecular graph data with graph neural networks (GNNs), then retrains the encoder with curriculum-based learning strategies to reduce uncertainty and improve accuracy, and finally uses LBNT as the predictor, which joint retrains with the encoder after independently training, for prediction and visualization. The framework is validated on the public datasets and compared to single GNNs with normal training strategies as well as GNN encoders with common machine learning predictors instead of the LBNT predictor. The result reveals that the proposed framework enhances the point prediction accuracy of the completely trained GNN and reduces its uncertainty through curriculum-based learning, and further improves the accuracy by combining LBNT. Besides, compared with common machine learning tools, the LBNT predictor generally has the best performance because of joint retraining with the GNN encoder. The decision-making process of LBNT is also better and easier to explain than that of other models.

Addressing Noise and Estimating Uncertainty in Biomedical Data through the Exploration of Chemical Space.

Noise is a basic ingredient in data, since observed data are always contaminated by unwanted deviations, i.e., noise, which, in the case of overdetermined systems (with more data than model parameters), cause the corresponding linear system of equations to have an imperfect solution. In addition, in the case of highly underdetermined parameterization, noise can be absorbed by the model, generating spurious solutions. This is a very undesirable situation that might lead to incorrect conclusions. We presented mathematical formalism based on the inverse problem theory combined with artificial intelligence methodologies to perform an enhanced sampling of noisy biomedical data to improve the finding of meaningful solutions. Random sampling methods fail for high-dimensional biomedical problems. Sampling methods such as smart model parameterizations, forward surrogates, and parallel computing are better suited for such problems. We applied these methods to several important biomedical problems, such as phenotype prediction and a problem related to predicting the effects of protein mutations, i.e., if a given single residue mutation is neutral or deleterious, causing a disease. We also applied these methods to de novo drug discovery and drug repositioning (repurposing) through the enhanced exploration of huge chemical space. The purpose of these novel methods that address the problem of noise and uncertainty in biomedical data is to find new therapeutic solutions, perform drug repurposing, and accelerate and optimize drug discovery, thus reestablishing homeostasis. Finding the right target, the right compound, and the right patient are the three bottlenecks to running successful clinical trials from the correct analysis of preclinical models. Artificial intelligence can provide a solution to these problems, considering that the character of the data restricts the quality of the prediction, as in any modeling procedure in data analysis. The use of simple and plain methodologies is crucial to tackling these important and challenging problems, particularly drug repositioning/repurposing in rare diseases.

Machine-learning-assisted discovery of highly efficient high-entropy alloy catalysts for the oxygen reduction reaction

SummaryHigh-entropy alloys (HEAs) have recently been applied in the field of heterogeneous catalysis benefiting from vast chemical space. However, huge chemical space also brings extreme challenges for the comprehensive study of HEAs by traditional trial-and-error experiments. Therefore, the machine learning (ML) method is presented to investigate the oxygen reduction reaction (ORR) catalytic activity of millions of reactive sites on HEA surfaces. The well-performed ML model is constructed based on the gradient boosting regression (GBR) algorithm with high accuracy, generalizability, and simplicity. In-depth analysis of the results demonstrates that adsorption energy is a mixture of the individual contributions of coordinated metal atoms near the reactive site. An efficient strategy is proposed to further boost the ORR catalytic activity of promising HEA catalysts by optimizing the HEA surface structure, which recommends a highly efficient HEA catalyst of Ir48Pt74Ru30Rh30Ag74. Our work offers a guide to the rational design and nanostructure synthesis of HEA catalysts.

3D-Scaffold: A Deep Learning Framework to Generate 3D Coordinates of Drug-like Molecules with Desired Scaffolds

The prerequisite of therapeutic drug design and discovery is to identify novel molecules and developing lead candidates with desired biophysical and biochemical properties. Deep generative models have demonstrated their ability to find such molecules by exploring a huge chemical space efficiently. An effective way to generate new molecules with desired target properties is by constraining the critical fucntional groups or the core scaffolds in the generation process. To this end, we developed a domain aware generative framework called 3D-Scaffold that takes 3D coordinates of the desired scaffold as an input and generates 3D coordinates of novel therapeutic candidates as an output while always preserving the desired scaffolds in generated structures. We demonstrated that our framework generates predominantly valid, unique, novel, and experimentally synthesizable molecules that have drug-like properties similar to the molecules in the training set. Using domain specific data sets, we generate covalent and noncovalent antiviral inhibitors targeting viral proteins. To measure the success of our framework in generating therapeutic candidates, generated structures were subjected to high throughput virtual screening via docking simulations, which shows favorable interaction against SARS-CoV-2 main protease (Mpro) and nonstructural protein endoribonuclease (NSP15) targets. Most importantly, our deep learning model performs well with relatively small 3D structural training data and quickly learns to generalize to new scaffolds, highlighting its potential application to other domains for generating target specific candidates.

Individual and collective human intelligence in drug design: evaluating the search strategy

In recent years, individual and collective human intelligence, defined as the knowledge, skills, reasoning and intuition of individuals and groups, have been used in combination with computer algorithms to solve complex scientific problems. Such approach was successfully used in different research fields such as: structural biology, comparative genomics, macromolecular crystallography and RNA design. Herein we describe an attempt to use a similar approach in small-molecule drug discovery, specifically to drive search strategies of de novo drug design. This is assessed with a case study that consists of a series of public experiments in which participants had to explore the huge chemical space in silico to find predefined compounds by designing molecules and analyzing the score associate with them. Such a process may be seen as an instantaneous surrogate of the classical design-make-test cycles carried out by medicinal chemists during the drug discovery hit to lead phase but not hindered by long synthesis and testing times. We present first findings on (1) assessing human intelligence in chemical space exploration, (2) comparing individual and collective human intelligence performance in this task and (3) contrasting some human and artificial intelligence achievements in de novo drug design.

Organic Spacers in 2D Perovskites: General Trends and Structure‐Property Relationships from Computational Studies

AbstractAddition of large organic molecules to halide perovskites has been shown to provoke dimensionality reduction and formation of two‐dimensional phases that demonstrate improved long‐term stabilities. Optoelectronic properties of the resulting 2D layered perovskites are strongly influenced by the chemical nature of the additive molecules, which opens immense possibilities for preparation of materials with tailored properties. However, given the huge chemical space of possible organic spacers, a systematic and exhaustive search for optimal compounds is impossible and general structure–property relationships that could guide a rational design are still largely absent. Here, we provide an overview of a series of recent computational studies from our group on different types of spacers. We first develop a simplified universal monovalent cation model to map out approximate structural stability maps as a function of the van der Waals radius and the magnitude of dispersion interactions to monitor the possible emergence of 2D phases. We further provide structural and photophysical insights from classical and first‐principles molecular dynamics simulations and density functional theory calculations on 2D hybrid perovskites based on a wide range of spacers with different chemical nature and varying conformational properties. Our computational predictions are validated through comparison with powder diffraction, conductivity and optical measurements. Such comparative study allows for providing some general structure–property correlations, which can serve as design guidelines in the search for optimal 2D and mixed 2D/3D perovskite photovoltaic materials.

High-Throughput Tertiary Amine Deoxygenated Photopolymerizations for Synthesizing Polymer Libraries.

The huge chemical space potential of synthetic polymers means that in many studies only a small parameter range can be realistically synthesized in a short time and hence the “best” formulations may not be optimum. Throughput is traditionally limited by the need for deoxygenation in radical polymerizations, but advances in photopolymerization now provide opportunities for “in-air” polymerizations. Here, we have developed a protocol using liquid handling robots (or multichannel pipettes) with blue light photolysis of reversible addition fragmentation chain transfer agents and tertiary amine deoxygenation to enable the synthesis of polymer libraries in industry-standard 96-well plates with no specialized infrastructure and no degassing step. The roles of solvents and amine deoxygenators are explored to optimize the polymerization, particularly to look at alternatives to dimethyl sulfoxide (DMSO) for hydrophobic monomer (co)polymerization. DMSO is shown to aid the degassing process but is not easy to remove and hence prevents isolation of pure polymers. In contrast, using dioxane in-plate evaporation or precipitation of the tertiary amine allowed isolation of polymers in-plate. This removed all reaction components yielding pure polymers, which is not easily achieved with systems using metal catalysts and secondary reductants, such as ascorbic acid. As an example of the throughput, in just under 40 h, 392 polymers were synthesized and subsequently analyzed direct from plates by a 96-well plate sampling size exclusion chromatography system to demonstrate reproducibility. Due to less efficient degassing in dioxane (compared to DMSO), the molecular weight and dispersity control were limited in some cases (with acrylates giving the lowest dispersities), but the key aim of this system is to access hundreds of polymers quickly and in a format needed to enable testing. This method enables easy exploration of chemical space and development of screening libraries to identify hits for further study using precision polymerization methods.

Multifidelity Information Fusion with Machine Learning: A Case Study of Dopant Formation Energies in Hafnia.

Cost versus accuracy trade-offs are frequently encountered in materials science and engineering, where a particular property of interest can be measured/computed at different levels of accuracy or fidelity. Naturally, the most accurate measurement is also the most resource and time intensive, while the inexpensive quicker alternatives tend to be noisy. In such situations, a number of machine learning (ML) based multifidelity information fusion (MFIF) strategies can be employed to fuse information accessible from varying sources of fidelity and make predictions at the highest level of accuracy. In this work, we perform a comparative study on traditionally employed single-fidelity and three MFIF strategies, namely, (1) Δ-learning, (2) low-fidelity as a feature, and (3) multifidelity cokriging (CK) to compare their relative prediction accuracies and efficiencies for accelerated property predictions and high throughput chemical space explorations. We perform our analysis using a dopant formation energy data set for hafnia, which is a well-known high-k material and is being extensively studied for its promising ferroelectric, piezoelectric, and pyroelectric properties. We use a dopant formation energy data set of 42 dopants in hafnia-each studied in six different hafnia phases-computed at two levels of fidelities to find merits and limitations of these ML strategies. The findings of this work indicate that the MFIF based learning schemes outperform the traditional SF machine learning methods, such as Gaussian process regression and CK provides an accurate, inexpensive and flexible alternative to other MFIF strategies. While the results presented here are for the case study of hafnia, they are expected to be general. Therefore, materials discovery problems that involve huge chemical space explorations can be studied efficiently (or even made feasible in some situations) through a combination of a large number of low-fidelity and a few high-fidelity measurements/computations, in conjunction with the CK approach.

Hit discovery of 4-amino-N-(4-(3-(trifluoromethyl)phenoxy)pyrimidin-5-yl)benzamide: A novel EGFR inhibitor from a designed small library

Searching for hit compounds within the huge chemical space resembles the attempt to find a needle in a haystack. Cheminformatics-guided selection of few representative molecules of a rationally designed virtual combinatorial library is a powerful tool to confront this challenge, speed up hit identification and cut off costs. Herein, this approach has been applied to identify hit compounds with novel scaffolds able to inhibit EGFR kinase. From a generated virtual library, six 4-aryloxy-5-aminopyrimidine scaffold-derived compounds were selected, synthesized and evaluated as hit EGFR inhibitors. 4-Aryloxy-5-benzamidopyrimidines inhibited EGFR with IC50 1.05–5.37 μM. Cell-based assay of the most potent EGFR inhibitor hit (10ac) confirmed its cytotoxicity against different cancerous cells. In spite of no EGFR, HER2 or VEGFR1 inhibition was elicited by 4-aryloxy-5-(thio)ureidopyrimidine derivatives, cell-based evaluation suggested them as antiproliferative hits acting by other mechanism(s). Molecular docking study provided a plausible explanation of incapability of 4-aryloxy-5-(thio)ureidopyrimidines to inhibit EGFR and suggested a reasonable binding mode of 4-aryloxy-5-benzamidopyrimidines which provides a basis to develop more optimized ligands.

De Novo Design of Drug-Like Molecules by a Fragment-Based Molecular Evolutionary Approach

This paper describes a similarity-driven simple evolutionary approach to producing candidate molecules of new drugs. The aim of the method is to explore the candidates that are structurally similar to the reference molecule and yet somewhat different in not only peripheral chains but also their scaffolds. The method employs a known active molecule of our interest as a reference molecule which is used to navigate a huge chemical space. The reference molecule is also used to obtain seed fragments. An initial set of individual structures is prepared with the seed fragments and additional fragments using several connection rules. The fragment library is preferably prepared from a collection of known molecules related to the target of the reference molecule. Every fragment of the library can be used for fragment-based mutation. All the fragments are categorized into three classes; rings, linkers, and side chains. New individuals are produced by the crossover and the fragment-based mutation with the fragment library. Computer experiments with our own fragment library prepared from GPCR SARfari verified the feasibility of our approach to drug discovery.