ChemPrint: An AI-Driven Framework for Enhanced Drug Discovery DOI Creative Commons

Tyler Umansky,

Virgil A. Woods,

Sean M. Russell

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: March 27, 2024

ABSTRACT Traditional High-Throughput Screening (HTS) drug discovery is inefficient. Hit rates for compounds with clinical therapeutic potential are typically 0.5% and only up to 2% maximally. Deep learning models have enriched screening 28%; however, these results include hits non-therapeutic relevant concentrations, insufficient novelty their training set, traverse limited chemical space. This study introduces a novel artificial intelligence (AI)-driven platform, GALILEO, the Molecular-Geometric Learning (Mol-GDL) model, ChemPrint. model deploys both t-distributed Stochastic Neighbor Embedding (t-SNE) data splitting maximize dissimilarity during adaptive molecular embeddings enhance predictive capabilities navigate uncharted territories. When tested retrospectively, ChemPrint outperformed panel of five difficult-to-drug oncology targets, AXL BRD4, achieving an average AUROC score 0.897 0.876 BRD4 using t-SNE split, compared benchmark scores ranging from 0.826 0.885 0.801 0.852 BRD4. In zero-shot prospective study, in vitro testing demonstrated that 19 41 nominated by against inhibitory activity at concentrations ≤ 20 µM, 46% hit rate. The reported average-maximum Tanimoto similarity 0.36 relative set 0.13 (AXL) 0.10 (BRD4) stage targets. Our findings demonstrate increasing test difficulty through on datasets maximal enhances model. compound libraries high low novelty. Taken together, proposed platform sets new performance standard.

Language: Английский

Machine learning for synthetic gene circuit engineering DOI
Sebastian Palacios,

James J. Collins,

Domitilla Del Vecchio

et al.

Current Opinion in Biotechnology, Journal Year: 2025, Volume and Issue: 92, P. 103263 - 103263

Published: Jan. 27, 2025

Language: Английский

Citations

2
A Review of Large Language Models and Autonomous Agents in Chemistry DOI Creative Commons
Mayk Caldas Ramos, Christopher J. Collison, Andrew Dickson White

et al.

Chemical Science, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 9, 2024

Large language models (LLMs) have emerged as powerful tools in chemistry, significantly impacting molecule design, property prediction, and synthesis optimization. This review highlights LLM capabilities these domains their potential to accelerate scientific discovery through automation. We also LLM-based autonomous agents: LLMs with a broader set of interact surrounding environment. These agents perform diverse tasks such paper scraping, interfacing automated laboratories, planning. As are an emerging topic, we extend the scope our beyond chemistry discuss across any domains. covers recent history, current capabilities, design agents, addressing specific challenges, opportunities, future directions chemistry. Key challenges include data quality integration, model interpretability, need for standard benchmarks, while point towards more sophisticated multi-modal enhanced collaboration between experimental methods. Due quick pace this field, repository has been built keep track latest studies: https://github.com/ur-whitelab/LLMs-in-science.

Language: Английский

Citations

13
Image-based generation for molecule design with SketchMol DOI
Zixu Wang, Yangyang Chen,

Pengsen Ma

et al.

Nature Machine Intelligence, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 13, 2025

Language: Английский

Citations

1
AI‐Driven Electrolyte Additive Selection to Boost Aqueous Zn‐Ion Batteries Stability DOI Open Access
Haobo Li, Junnan Hao, Shi Zhang Qiao

et al.

Advanced Materials, Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 23, 2024

Abstract In tackling the stability challenge of aqueous Zn‐ion batteries (AZIBs) for large‐scale energy storage, adoption electrolyte additive emerges as a practical solution. Unlike current trial‐and‐error methods selecting additives, data‐driven strategy is proposed using theoretically computed surface free descriptor, benchmarked against experimental results. Numerous additives are calculated from existing literature, forming database machine learning (ML) training. Importantly, this ML model relies solely on values, effectively addressing large solvent molecule models that difficult to handle with quantum chemistry computation. The interpretable linear regression algorithm identifies number heavy atoms in and liquid tension key factors. Artificial intelligence (AI) clustering categorizes molecules, identifying regions most significant impact enhancing battery stability. Experimental verification successfully confirms exceptional performance 1,2,3‐butanetriol acetone optimal region. This integrated methodology, combining theoretical models, ML, validation, provides insights into rational design additives.

Language: Английский

Citations

7
Overcoming Challenges in Small-Molecule Drug Bioavailability: A Review of Key Factors and Approaches DOI Open Access
Ke Wu,

Soon Hwan Kwon,

Xuhan Zhou

et al.

International Journal of Molecular Sciences, Journal Year: 2024, Volume and Issue: 25(23), P. 13121 - 13121

Published: Dec. 6, 2024

The bioavailability of small-molecule drugs remains a critical challenge in pharmaceutical development, significantly impacting therapeutic efficacy and commercial viability. This review synthesizes recent advances understanding overcoming limitations, focusing on key physicochemical biological factors influencing drug absorption distribution. We examine cutting-edge strategies for enhancing bioavailability, including innovative formulation approaches, rational structural modifications, the application artificial intelligence design. integration nanotechnology, 3D printing, stimuli-responsive delivery systems are highlighted as promising avenues improving delivery. discuss importance holistic, multidisciplinary approach to optimization, emphasizing early-stage consideration ADME properties need patient-centric also explores emerging technologies such CRISPR-Cas9-mediated personalization microbiome modulation tailored enhancement. Finally, we outline future research directions, advanced predictive modeling, barriers, addressing challenges modalities. By elucidating complex interplay affecting this aims guide efforts developing more effective accessible therapeutics.

Language: Английский

Citations

7
Data-centric challenges with the application and adoption of artificial intelligence for drug discovery DOI
Ghita Ghislat,

Saiveth Hernández-Hernández,

Chayanit Piyawajanusorn

et al.

Expert Opinion on Drug Discovery, Journal Year: 2024, Volume and Issue: 19(11), P. 1297 - 1307

Published: Sept. 24, 2024

Artificial intelligence (AI) is exhibiting tremendous potential to reduce the massive costs and long timescales of drug discovery. There are however important challenges currently limiting impact scope AI models.

Language: Английский

Citations

6
Strategies for Redesigning Withdrawn Drugs to Enhance Therapeutic Efficacy and Safety: A Review DOI
Chirag Patel, Adeeba Shakeel, Raghvendra Mall

et al.

Wiley Interdisciplinary Reviews Computational Molecular Science, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 1, 2025

ABSTRACT Drug toxicity and market withdrawals are two issues that often obstruct the lengthy intricate drug discovery process. In order to enhance effectiveness safety, this review examines withdrawn drugs presents a novel paradigm for their redesign. addition addressing methodological with datasets, study highlights important shortcomings in silico prediction models suggests solutions. High‐throughput screening (HTS) has greatly progressed advent of 3D organoid organ‐on‐chip (OoC) technologies, which provide physiologically appropriate systems replicate structure function human tissue. These accurate, human‐relevant data development, evaluation, disease modeling, overcoming limitations traditional 2D cell cultures animal models. Their integration into HTS pipelines shown have major influence, promoting redesign efforts enabling improved accuracy preclinical research. The potential fragment‐based pharmacokinetics (PK) pharmacodynamics (PD) when combined conventional techniques is highlighted study. limits discussed, focus on need bioengineered humanized such OoC technologies organoids. To improve candidate simulate real illnesses, advanced crucial. This leads target affinity fewer adverse effects.

Language: Английский

Citations

0
Artificial Intelligence in Retrosynthesis Prediction and its Applications in Medicinal Chemistry DOI

Lanxin Long,

Rui Li, Jian Zhang

et al.

Journal of Medicinal Chemistry, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 30, 2025

Retrosynthesis is a strategy to analyze the synthetic routes for target molecules in medicinal chemistry. However, traditional retrosynthesis predictions performed by chemists and rule-based expert systems struggle adapt vast chemical space of real-world scenarios. Artificial intelligence (AI) has revolutionized prediction recent decades, significantly increasing accuracy diversity compounds. Single-step AI-driven models can be generalized into three types based on their dependence predefined reaction templates (template-based, semitemplate-based methods, template-free models), with respective advantages limitations, common challenges that limit chemistry applications. Moreover, there are relatively inadequate multi-step which lack strong links single-step methods. Herein, we review advancements AI applications summarizing related techniques landscape current representative propose feasible solutions tackle existing problems outline future directions this field.

Language: Английский

Citations

0
DrugDiff: small molecule diffusion model with flexible guidance towards molecular properties DOI Creative Commons
Marie Oestreich, Erinç Merdivan, Michael Lee

et al.

Journal of Cheminformatics, Journal Year: 2025, Volume and Issue: 17(1)

Published: Feb. 25, 2025

Abstract With the cost/yield-ratio of drug development becoming increasingly unfavourable, recent work has explored machine learning to accelerate early stages process. Given current success deep generative models across domains, we here investigated their application property-based proposal new small molecules for development. Specifically, trained a latent diffusion model— DrugDiff —paired with predictor guidance generate novel compounds variety desired molecular properties. The architecture was designed be highly flexible and easily adaptable future scenarios. Our experiments showed successful generation unique, diverse targeted code is available at https://github.com/MarieOestreich/DrugDiff . Scientific Contribution This expands use modelling in field from previously introduced proteins RNA presented molecules. making up majority drugs, but simultaneously being difficult model due elaborate chemical rules, this tackles level difficulty comparison sequence-based molecule as case RNA. Additionally, demonstrated framework flexible, allowing easy addition or removal considered properties without need retrain model, it research settings shows compelling performance wide

Language: Английский

Citations

0
Knowledge-guided diffusion model for 3D ligand-pharmacophore mapping DOI Creative Commons

Junlin Yu,

Zhou Cong, Xiang-Li Ning

et al.

Nature Communications, Journal Year: 2025, Volume and Issue: 16(1)

Published: March 6, 2025

Pharmacophores are abstractions of essential chemical interaction patterns, holding an irreplaceable position in drug discovery. Despite the availability many pharmacophore tools, adoption deep learning for pharmacophore-guided discovery remains relatively rare. We herein propose a knowledge-guided diffusion framework 'on-the-fly' 3D ligand-pharmacophore mapping, named DiffPhore. It leverages matching knowledge to guide ligand conformation generation, meanwhile utilizing calibrated sampling mitigate exposure bias iterative search process. By training on two self-established datasets pairs, DiffPhore achieves state-of-the-art performance predicting binding conformations, surpassing traditional tools and several advanced docking methods. also manifests superior virtual screening power lead target fishing. Using DiffPhore, we successfully identify structurally distinct inhibitors human glutaminyl cyclases, their modes further validated through co-crystallographic analysis. believe this work will advance AI-enabled techniques.

Language: Английский

Citations

0