A5046201147- Computational Drug Discovery Methods
- Protein Structure and Dynamics
- Machine Learning in Materials Science
- Chemical Synthesis and Analysis
- thermodynamics and calorimetric analyses
- Fibroblast Growth Factor Research
- Genetics, Bioinformatics, and Biomedical Research
- Heat shock proteins research
- Mass Spectrometry Techniques and Applications
- Chemical Reaction Mechanisms
- Monoclonal and Polyclonal Antibodies Research
- Biotin and Related Studies
- Cancer therapeutics and mechanisms
- Chemistry and Chemical Engineering
- Machine Learning and Algorithms
- Biochemical and Structural Characterization
- Proteoglycans and glycosaminoglycans research
- Kruppel-like factors research
- Molecular Sensors and Ion Detection
- Innovative Microfluidic and Catalytic Techniques Innovation
- Enzyme Structure and Function
- Click Chemistry and Applications
- Mast cells and histamine
- Angiogenesis and VEGF in Cancer
- Biotechnology and Related Fields
Sanofi (France)
2010-2025
Sanofi (China)
2023-2024
Sanofi (United States)
2023-2024
Design Science (United Kingdom)
2023
Spanish National Cancer Research Centre
2010
Italian Institute of Technology
2010
Molécules d'Intérêt Biologique
2002
Institut de Recherche sur les Systèmes Atomiques et Moléculaires Complexes
2002
Université Toulouse III - Paul Sabatier
2002
Drug-target residence time (τ), one of the main determinants drug efficacy, remains highly challenging to predict computationally and, therefore, is usually not considered in early stages design. Here, we present an efficient computational method, τ-random acceleration molecular dynamics (τRAMD), for ranking candidates by their and obtaining insights into ligand-target dissociation mechanisms. We assessed τRAMD on a data set 70 diverse drug-like ligands N-terminal domain HSP90α,...
Molecular generative artificial intelligence is drawing significant attention in the drug design community, with several experimentally validated proof of concepts already published. Nevertheless, models are known for sometimes generating unrealistic, unstable, unsynthesizable, or uninteresting structures. This calls methods to constrain those algorithms generate structures drug-like portions chemical space. While concept applicability domains predictive well studied, its counterpart not yet...
Abstract Despite growing interest and success in automated in-silico molecular design, questions remain regarding the ability of goal-directed generation algorithms to perform unbiased exploration novel chemical spaces. A specific phenomenon has recently been highlighted: guided with machine learning models produce molecules high scores according optimization model, but low control models, even when trained on same data distribution target. In this work, we show that worrisome behavior is...
Free-energy pathway methods show great promise in computing the mode of action and free energy profile associated with binding small molecules proteins, but are generally very computationally demanding. Here we apply a novel approach based on metadynamics path collective variables. We that this combination is able to find an optimal reaction coordinate explicit solvent full flexibility, while minimizing human intervention computational costs. it predict affinity congeneric series 5 CDK2...
The accurate yet efficient evaluation of the free energy profiles ligand-target association is a long sought goal in rational drug design. Methods that calculate along realistic pathways, such as metadynamics, have been shown to provide reliable profiles, while accounting properly for solvation and target flexibility. However, these approaches usually require prohibitive computational resources expert human intervention. Here, we show how multiple walkers when performed with optimal path...
A key challenge in drug discovery is to optimize, silico, various absorption and affinity properties of small molecules. One strategy that was proposed for such optimization process active learning. In learning molecules are selected testing based on their likelihood improving model performance. To enable the use with advanced neural network models we developed two novel batch selection methods. These methods were tested several public datasets different goals sizes. We have also curated new...
Over the past decade, collective intelligence, i.e., intelligence that emerges from efforts, has transformed complex problem-solving and decision-making. In drug discovery, decision-making often relies on medicinal chemistry intuition. The present study explores application of in focusing lead optimization. Ninety-two Sanofi researchers with diverse expertise participated anonymously an exercise centered ADMET-related questions. Their feedback was used to build a agent, which compared...
Fibroblast growth factor (FGF) signaling regulates mammalian development and metabolism, its dysregulation is implicated in many inherited acquired diseases, including cancer. Heparan sulfate glycosaminoglycans (HSGAGs) are essential for FGF as they promote FGF.FGF receptor (FGFR) binding dimerization. Using novel organic synthesis protocols to prepare homogeneously sulfated heparin mimetics (HM), hexasaccharide (HM(6)), octasaccharide (HM(8)), decasaccharide (HM(10)), we tested the ability...
The dissociative hydrolysis reaction of the methyl phosphate monoanion has been studied for reactant species CH3OPO3H- (1) and CH3OPO3H-·H2O (1a) in gas aqueous phases by density functional theory (B3LYP) calculations. Nonspecific solvation effects were taken into account with polarizable continuum model PCM either solvating gas-phase paths or performing geometry searches directly presence correction. In agreement previous theoretical studies, our calculations indicate that proton transfer...
One of the major applications generative models for drug Discovery targets lead-optimization phase. During optimization a lead series, it is common to have scaffold constraints imposed on structure molecules designed. Without enforcing such constraints, probability generating with required extremely low and hinders practicality de-novo design. To tackle this issue, we introduce new algorithm perform scaffold-constrained in-silico molecular We build well-known SMILES-based Recurrent Neural...
The heat shock protein 90 (Hsp90) is a molecular chaperone that controls the folding and activation of client proteins using free energy ATP hydrolysis. Hsp90 active site in its N-terminal domain (NTD). Our goal to characterize dynamics NTD an autoencoder-learned collective variable (CV) conjunction with adaptive biasing force Langevin dynamics. Using dihedral analysis, we cluster all available experimental structures into distinct native states. We then perform unbiased (MD) simulations...
The binding kinetic properties of potential drugs may significantly influence their subsequent clinical efficacy. Predictions these based on computer simulations provide a useful alternative to expensive and time-consuming experimental counterparts, even at an early drug discovery stage. Herein, we perform scaled molecular dynamics (ScaledMD) set 27 ligands HSP90 belonging more than seven chemical series estimate relative residence times. We introduce two new techniques for the analysis...
In a previous theoretical study [J. Am. Chem. Soc., in press], using combination of DFT and continuum solvation (PCM) methods, the anionic zwitterion CH3O+(H)PO32- (2) has been identified as key intermediate mechanism for dissociative hydrolysis methyl phosphate anion CH3OPO3H- (1). To confirm this finding, DFT/B3LYP calculations which few solvent molecules are explicitly considered, reported. Hydrogen-bonded complexes 2·(H2O)n (n = 2−4) have fully optimized characterized on their respective...
Over the last decade, combination of collective intelligence with computational methods has transformed complex problem-solving. Here, we investigate if and how can be applied to drug discovery, focusing on lead optimization stage discovery process. For this study, 92 Sanofi researchers diverse scientific expertise participated anonymously in a exercise. Their feedback was used build agent that compared an artificial model developed parallel. This work led three major conclusions. First,...
<title>Abstract</title> Over the last decade, combination of collective intelligence with computational methods has transformed complex problem-solving. Here, we investigate if and how can be applied to drug discovery, focusing on lead optimization stage discovery process. For this study, 92 Sanofi researchers diverse scientific expertise participated anonymously in a exercise. Their feedback was used build agent that compared an artificial model developed parallel. This work led three major...
A key challenge in drug discovery is to optimize, silico, various absorption and affinity properties of small molecules. One strategy that was proposed for such optimization process active learning. In learning molecules are selected testing based on their likelihood improving model performance. To enable the use with advanced neural network models we developed two novel batch selection methods. These methods were tested several public datasets different goals sizes. We have also curated new...
Molecular generative artificial intelligence is drawing significant attention in the drug design community, with several experimentally validated proofs of concepts already published. Nevertheless, models are known for sometimes generating unrealistic, unsynthesizable or unstable structures. This calls methods to constrain those algorithms generate structures reasonable portions chemical space. While concept applicability domains (AD) predictive well studied, its counterpart not yet defined....
Flexible regions in proteins, such as loops, cannot be represented by a single conformation. Instead, conformational ensembles are needed to provide more global picture. In this context, identifying statistically meaningful conformations within an ensemble generated loop sampling techniques remains open problem. The difficulty is primarily related the lack of structural data about these flexible regions. With majority coming from x-ray crystallography and ignoring plasticity, conception...
<div>The binding kinetic properties of potential drugs may significantly influence their subsequent clinical efficacy. Predictions these based on computer simulations provide a useful alternative to expensive and time-demanding experimental counterparts, even at an early drug discovery stage.</div><div>Herein, we perform Scaled Molecular Dynamics (ScaledMD) set 27 ligands HSP90 belonging more than 7 chemical series in order estimate relative residence time. We introduce two...
One of the common strategies to identify novel chemical matter in drug discovery consists performing a High Throughput Screening (HTS). However, large amount data generated at dose-response (DR) step an HTS campaign requires careful analysis detect artifacts and correct erroneous datapoints before validating experiments. This which review each DR experiment can be time consuming prone human errors or inconsistencies. AI4DR is system that has been developed for classification curves based on...