A Nested Model for AI Design and Validation DOI Creative Commons
Akshat Dubey, Zewen Yang, Georges Hattab

et al.

iScience, Journal Year: 2024, Volume and Issue: 27(9), P. 110603 - 110603

Published: July 30, 2024

The growing AI field faces trust, transparency, fairness, and discrimination challenges. Despite the need for new regulations, there is a mismatch between regulatory science AI, preventing consistent framework. A five-layer nested model design validation aims to address these issues streamline application validation, improving adoption. This aligns with addresses practitioners' daily challenges, offers prescriptive guidance determining appropriate evaluation approaches by identifying unique validity threats. We have three recommendations motivated this model: (1) Authors should distinguish layers when claiming contributions clarify specific areas in which contribution made avoid confusion; (2) authors explicitly state upstream assumptions ensure that context limitations of their system are clearly understood, (3) venues promote thorough testing systems compliance requirements.

Language: Английский

Artificial intelligence in drug development DOI
Kang Zhang, Xin Yang, Yifei Wang

et al.

Nature Medicine, Journal Year: 2025, Volume and Issue: 31(1), P. 45 - 59

Published: Jan. 1, 2025

Language: Английский

Citations

11
From Data to Discovery: Recent Trends of Machine Learning in Metal–Organic Frameworks DOI Creative Commons
Junkil Park, Honghui Kim, Yeonghun Kang

et al.

JACS Au, Journal Year: 2024, Volume and Issue: 4(10), P. 3727 - 3743

Published: Sept. 12, 2024

Renowned for their high porosity and structural diversity, metal-organic frameworks (MOFs) are a promising class of materials wide range applications. In recent decades, with the development large-scale databases, MOF community has witnessed innovations brought by data-driven machine learning methods, which have enabled deeper understanding chemical nature MOFs led to novel structures. Notably, is continuously rapidly advancing as new methodologies, architectures, data representations actively being investigated, implementation in discovery vigorously pursued. Under these circumstances, it important closely monitor research trends identify technologies that introduced. this Perspective, we focus on emerging within field MOFs, challenges they face, future directions development.

Language: Английский

Citations

8
Scaling Graph Neural Networks to Large Proteins DOI
Justin Airas, Bin Zhang

Journal of Chemical Theory and Computation, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 6, 2025

Graph neural network (GNN) architectures have emerged as promising force field models, exhibiting high accuracy in predicting complex energies and forces based on atomic identities Cartesian coordinates. To expand the applicability of GNNs, machine learning fields more broadly, optimizing their computational efficiency is critical, especially for large biomolecular systems classical molecular dynamics simulations. In this study, we address key challenges existing GNN benchmarks by introducing a dataset, DISPEF, which comprises large, biologically relevant proteins. DISPEF includes 207,454 proteins with sizes up to 12,499 atoms features diverse chemical environments, spanning folded disordered regions. The implicit solvation free energies, used training targets, represent particularly challenging case due many-body nature, providing stringent test evaluating expressiveness models. We benchmark performance seven GNNs emphasizing importance directly accounting long-range interactions enhance model transferability. Additionally, present novel multiscale architecture, termed Schake, delivers transferable computationally efficient energy predictions Our findings offer valuable insights tools advancing protein modeling applications.

Language: Английский

Citations

1
Transformers for Molecular Property Prediction: Lessons Learned from the Past Five Years DOI
A.R. Sultan, Jochen Sieg, Miriam Mathea

et al.

Journal of Chemical Information and Modeling, Journal Year: 2024, Volume and Issue: 64(16), P. 6259 - 6280

Published: Aug. 13, 2024

Molecular Property Prediction (MPP) is vital for drug discovery, crop protection, and environmental science. Over the last decades, diverse computational techniques have been developed, from using simple physical chemical properties molecular fingerprints in statistical models classical machine learning to advanced deep approaches. In this review, we aim distill insights current research on employing transformer MPP. We analyze currently available explore key questions that arise when training fine-tuning a model These encompass choice scale of pretraining data, optimal architecture selections, promising objectives. Our analysis highlights areas not yet covered research, inviting further exploration enhance field's understanding. Additionally, address challenges comparing different models, emphasizing need standardized data splitting robust analysis.

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
Deep learning-assisted methods for accelerating the intelligent screening of novel 2D materials: New perspectives focusing on data collection and description DOI

Yuandong Lin,

Ji Ma, Yong‐Guang Jia

et al.

Coordination Chemistry Reviews, Journal Year: 2025, Volume and Issue: 529, P. 216436 - 216436

Published: Jan. 16, 2025

Language: Английский

Citations

0
One-Class Learning for Data Stream Through Graph Neural Networks DOI
Marcos Paulo Silva Gôlo, João Gama, Ricardo Marcondes Marcacini

et al.

Lecture notes in computer science, Journal Year: 2025, Volume and Issue: unknown, P. 61 - 75

Published: Jan. 1, 2025

Language: Английский

Citations

0
GNINA 1.3: the next increment in molecular docking with deep learning DOI Creative Commons
Andrew T. McNutt, Yanjing Li, Rocco Meli

et al.

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

Published: March 2, 2025

Abstract Computer-aided drug design has the potential to significantly reduce astronomical costs of development, and molecular docking plays a prominent role in this process. Molecular is an silico technique that predicts bound 3D conformations two molecules, necessary step for other structure-based methods. Here, we describe version 1.3 open-source software Gnina . This release updates underlying deep learning framework PyTorch, resulting more computationally efficient paving way seamless integration methods into pipeline. We retrained our CNN scoring functions on updated CrossDocked2020 v1.3 dataset introduce knowledge-distilled facilitate high-throughput virtual screening with Furthermore, add functionality covalent docking, where atom ligand covalently receptor. update expands scope further positions as user-friendly, framework. available at https://github.com/gnina/gnina Scientific contributions : GNINA open source tool enhanced support models effective screening.

Language: Английский

Citations

0
Learning Spatial Density Functions of Random Waypoint Mobility over Irregular Triangles and Convex Quadrilaterals DOI Creative Commons
Yiming Feng, Wanxin Gao, Lefeng Zhang

et al.

Mathematics, Journal Year: 2025, Volume and Issue: 13(6), P. 927 - 927

Published: March 11, 2025

For the optimization and performance evaluation of mobile ad hoc networks, a beneficial but challenging act is to derive from nodal movement behavior steady-state spatial density function locations over given finite area. Such derivation, however, often intractable when any assumption mobility model not basic, e.g., area irregular in shape. As first endeavor, we address this derivation problem for classic random waypoint convex polygons including triangles (i.e., 3-gons) quadrilaterals 4-gons). By mixing multiple Dirichlet distributions, devise mixture neural network tailored approximation then extend accommodate quadrilaterals. Experimental results show that our (DMM) can accurately capture irregularity ground-truth distributions at low training cost, markedly outperforming Gaussian (GMM).

Language: Английский

Citations

0
A Heterogeneous Image Registration Model for an Apple Orchard DOI Creative Commons
Debao Huang, Liqun Liu

Agronomy, Journal Year: 2025, Volume and Issue: 15(4), P. 889 - 889

Published: April 2, 2025

The current image registration models have problems such as low feature point matching accuracy, high memory consumption, and significant computational complexity in heterogeneous registration, especially complex environments. In this context, differences lighting leaf occlusion orchards can result inaccurate extraction during registration. To address these issues, study proposes an AD-ResSug model for First, a VGG16 network was included the encoder system, positional encoding embedded into network. This enabled us to better understand spatial relationships between points. addition of residual structures aimed solve gradient diffusion problem enhance flexibility scalability architecture. Then, we used Sinkhorn AutoDiff algorithm iteratively optimize optimal transmission problem, achieving Finally, carried out pruning compression operations minimize parameters computation cost while maintaining model’s performance. new uses evaluation indicators peak signal-to-noise ratio root mean square error well efficiency. proposed method achieved robust efficient performance, verified through experimental results quantitative comparisons processing color with ToF images captured using cameras natural apple orchards.

Language: Английский

Citations

0