Designing Target-specific Data Sets for Regioselectivity Predictions on Complex Substrates

Journal of the American Chemical Society, Journal Year: 2025, Volume and Issue: 147(9), P. 7476 - 7484

Published: Feb. 21, 2025

The development of machine learning models to predict the regioselectivity C(sp3)-H functionalization reactions is reported. A data set for dioxirane oxidations was curated from literature and used generate a model C-H oxidation. To assess whether smaller, intentionally designed sets could provide accuracy on complex targets, series acquisition functions were developed select most informative molecules specific target. Active learning-based that leverage predicted reactivity uncertainty found outperform those based molecular site similarity alone. use elaboration significantly reduced number points needed perform accurate prediction, it machine-designed can give predictions when larger, randomly selected fail. Finally, workflow experimentally validated five substrates shown be applicable predicting arene radical borylation. These studies quantitative alternative intuitive extrapolation "model substrates" frequently estimate molecules.

Language: Английский

---

Bayesian Optimization over Multiple Experimental Fidelities Accelerates Automated Discovery of Drug Molecules

ACS Central Science, Journal Year: 2025, Volume and Issue: 11(2), P. 346 - 356

Published: Feb. 5, 2025

Different experiments of differing fidelities are commonly used in the search for new drug molecules. In classic experimental funnels, libraries molecules undergo sequential rounds virtual, coarse, and refined screenings, with each level balanced between cost number screened. Bayesian optimization offers an alternative approach, using iterative to locate optimal fewer than large-scale screening, but without ability weigh costs benefits different types experiments. this work, we combine multifidelity approach funnel iteratively, taking full advantage experiments, their costs, quality data they produce. We first demonstrate utility (MF-BO) on a series targets reported ChEMBL, emphasizing what properties chemical space result substantial acceleration MF-BO. Then integrate MF-BO experiment selection algorithm into autonomous molecular discovery platform illustrate prospective histone deacetylase inhibitors docking scores, single-point percent inhibitions, dose–response IC50 values as low-, medium-, high-fidelity A appropriate diversity fidelity correlation use was constructed genetic generative algorithm. The integrated then docked more 3,500 molecules, automatically synthesized screened 120 inhibition, selected handful manual evaluation at highest fidelity. Many have never been any capacity. At end search, several were found submicromolar free problematic hydroxamate moieties that constrain current inhibitors.

Language: Английский

---

Direct (LC-)MS Identification of Regioisomers from C–H Functionalization by Partial Isotopic Labeling

ACS Central Science, Journal Year: 2025, Volume and Issue: 11(2), P. 272 - 278

Published: Feb. 14, 2025

C–H functionalization of complex substrates is highly enabling in total synthesis and the development late-stage drug candidates. Much work has been dedicated to developing new methods as well predictive modeling accelerate route scouting. However, workflows identify regioisomeric products are arduous, typically requiring chromatographic separation and/or nuclear magnetic resonance spectroscopy analysis. In addition, most reports focus on major or do not assign products, which biases models constructed from such data. Herein, we present a novel approach reaction analysis utilizing partial deuterium labels, enables direct product identification via liquid chromatography–mass spectrometry. When combined with spectral deconvolution, method generates ratios while circumventing chromatography altogether. Competitive kinetic isotope effects can also be determined. The resultant data expected useful construction across several dimensions including selectivity, impact structure mechanism, mass ionization patterns expedite metabolites.

Language: Английский

---

A review: Integration of NIRS and chemometric methods for tea quality control-principles, spectral preprocessing methods, machine learning algorithms, research progress, and future directions

Food Research International, Journal Year: 2025, Volume and Issue: 205, P. 115870 - 115870

Published: Feb. 15, 2025

Language: Английский

---

The Implementation and Impact of Chemical High-Throughput Experimentation at AstraZeneca

ACS Catalysis, Journal Year: 2025, Volume and Issue: unknown, P. 5229 - 5256

Published: March 13, 2025

Language: Английский

---

Machine Learning Approaches to Surpass the Limitations of the Beer–Lambert Law

ACS Omega, Journal Year: 2025, Volume and Issue: 10(16), P. 16597 - 16601

Published: April 18, 2025

Many scientific and industrial applications depend on the precise measurement of chemical concentrations. The current study demonstrates how an inventive method combining photographic images with a machine learning (ML) model successfully estimates concentration compound in solution. A using linear regression L2 regularization (ridge model) was developed as part predictive model. trained captured K2Cr2O7 solutions following standard setup. After completing training, evaluated data set test samples. prediction precision had been 210 high correlation between actual predicted concentrations obtained MAE, MSE, RMSE 1.4 × 10-5, 3.4 10-10, 1.0 respectively. ridge is also extended to predict potassium permanganate (KMnO4) highlights potential integrating techniques image analysis accurately quantify any species solution state. As this depends solely color intensity sample without molecular interactions, it exceeds limitations Beer-Lambert law. created minimizes requirement substantial expertise training hence bridges gap experienced novice analysts.

Language: Английский

---

Direct (LC-)MS Identification of Regioisomers in C-H Activation by Partial Isotopic Labeling

Published: July 31, 2024

C–H functionalization of complex substrates is highly enabling in total synthesis and the development late-stage drug candidates. Much work has been dedicated towards developing new methods as well predictive modeling to accelerate route scouting. However, workflows identify regioisomeric products are arduous, typically requiring chromatographic separation and/or nuclear magnetic resonance spectroscopy analysis. In addition, most reports focus on major or do not assign which biases models constructed from such data. Herein, we present a novel approach reaction analysis utilizing partial deuterium labels enables direct product identification via liquid chromatography–mass spectrometry. When combined with spectral deconvolution, method generates ratios while circumventing chromatography altogether. Competitive kinetic isotope effects can also be determined. The resultant data expected useful construction across several dimensions including selectivity, impact structure mechanism, mass ionization patterns.

Language: Английский

---

Application of an Electronic Eye to Address the Limitations of Beer-Lambert Law

Published: Aug. 16, 2024

Many scientific and industrial applications depend on the precise measurement of chemical concentrations. The current study demonstrates how an inventive method combining photographic images with a machine learning (ML) model successfully estimates concentration potassium dichromate (K2Cr2O7) in solution. A predictive is created by taking photographs K2Cr2O7 solutions evaluating color intensities those using ridge regression model. pre-diction accuracy was evaluated MAE, MSE, RMSE, high correlation between actual predicted concentrations obtained RMSE values 0.4%, 0.003%, 0.5%, respectively. also extended to predict KMnO4 highlights potential integrating techniques image analysis accurately quantify any species solution state. As this solely depends intensity sample without molecular interactions, it surpasses limitations Beer-Lambert law. minimizes requirement substantial expertise training, hence bridging gap experienced novice analysts.

Language: Английский

---

Overcoming Bottlenecks towards Complete Biocatalytic Conversions and Complete Product Recovery

Reaction Chemistry & Engineering, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 14, 2024

Sustainable processes by completing biocatalytic conversions and recovering products completely.

Language: Английский

---