Results 151 to 160 of about 130,387 (281)

Automatically Defining Protein Words for Diverse Functional Predictions Based on Attention Analysis of a Protein Language Model

Advanced Science, EarlyView.
Understanding protein sequence–function relationships remains challenging due to poorly defined motifs and limited residue‐level annotations. An annotation‐agnostic framework is introduced that segments protein sequences into “protein words” using attention patterns from protein language models.
Hedi Chen, Jingrui Zhong, Xiaochun Zhang, Jingke Chen, Lin Guo, Xiaoliang Xiong, Xiaonan Zhang, Xiangyu Liu, Bailong Xiao, Boxue Tian   +9 more
wiley   +1 more source

Lattice Genome Framework for Regionally Tailored Component‐Level Multi‐Objective Design in Additive Manufacturing

Advanced Science, EarlyView.
A Lattice Genome framework links geometric and process “genes” to lattice “phenotypes” via correction‐calibrated high‐throughput simulations and a growing performance database. Genome‐driven retrieval and recombination of unit cells enables component‐level, regionally tailored multi‐objective design: stress fields are programmed under constant relative
Haoyuan Deng, Yufan Zhao, Mingyang Cao, Haiou Yang, Pan Wang, Hongze Wang, Hao Wang, Akihiko Chiba, Xin Lin   +8 more
wiley   +1 more source

Mechanistic Analysis of Large Atomic Models of Molten Salt

Advanced Science, EarlyView.
This work uncovers the physical mechanism of large atomic models for molten salts by linking atomic contribution to electronic structure features. We demonstrate that energy predictions are physically determined by the local occupancy of frontier orbitals.
Yuliang Guo, Xiaobo Sun, Xiaoli Xi, Zuoren Nie   +3 more
wiley   +1 more source

Understanding Fabrication Variability in Core‐Shell Soft Biomaterials Using Stochastic Artificial Intelligence

Advanced Science, EarlyView.
Fabrication‐induced variability remains a fundamental limitation in the scalable design of soft biomaterials. In this work, a stochastic machine learning approach based on Gaussian processes modeling is employed to establish quantitative links between biofabrication parameters, material properties, and their intrinsic variability.
Maria Alexaki, Lília M. S. Dias, Raquel C. Gonçalves, Dinis O. Abranches, Albano N. Carneiro Neto, Rute A. S. Ferreira, Paulo S. B. André, João F. Mano, Mariana B. Oliveira   +8 more
wiley   +1 more source

Physics‐Embedded Neural Network: A Novel Approach to Design Polymeric Materials

Advanced Science, EarlyView.
Traditional black‐box models for polymer mechanics rely solely on data and lack physical interpretability. This work presents a physics‐embedded neural network (PENN) that integrates constitutive equations into machine learning. The approach ensures reliable stress predictions, provides interpretable parameters, and enables performance‐driven, inverse ...
Siqi Zhan, Hengheng Zhao, Haotian Wang, Zhanjie Liu, Qian Li, Weifeng Zhang, Qingsong Zhao, Liqun Zhang, Jun Liu   +8 more
wiley   +1 more source

Forecasting Root Rot Disease through Predictive Microbial Functional Profiling

Advanced Science, EarlyView.
Predicting soil‐borne disease moves beyond observation with a framework that elevates microbial functional genes into reliable forecasting biomarkers. By coupling targeted qPCR assays for core stress‐response genes with machine learning, this method detects root rot risks in pre‐symptomatic soils with over 80% accuracy.
Chuan You, Peng Ren, Ying Guan, Kai Gong, Zhen Hua, Wuxian Zhou, Xinlan Mei, Yikui Wang, Xiaofang Wang, Yangchun Xu, Qirong Shen, Zhong Wei   +11 more
wiley   +1 more source

Be aware of overfitting by hyperparameter optimization!

Journal of Cheminformatics
Hyperparameter optimization is very frequently employed in machine learning. However, an optimization of a large space of parameters could result in overfitting of models.
Igor V. Tetko, Ruud van Deursen, Guillaume Godin   +2 more
doaj   +1 more source

Design and Optimization of Full‐Stokes Hyperspectro‐Polarimetric Encoding Metasurfaces Based on Conditional Multi‐Task Deep Learning

Advanced Science, EarlyView.
A conditional multi‐task deep learning framework is developed for designing and optimizing Full‐Stokes Hyperspectro‐Polarimetric Encoding Metasurfaces (FHPEMs). This framework achieves joint spectro‐polarimetric learning and unified forward–inverse design.
Chenjie Gong, Haodong Shi, Qi Wang, Yingchao Li, Hongyu Sun, Jiayu Wang, Guanlin Li, Jian Zhang, Qiang Fu, Huilin Jiang   +9 more
wiley   +1 more source

ML Workflows for Screening Degradation‐Relevant Properties of Forever Chemicals

Advanced Science, EarlyView.
The environmental persistence of per‐ and polyfluoroalkyl substances (PFAS) necessitates efficient remediation strategies. This study presents physics‐informed machine learning workflows that accurately predict critical degradation properties, including bond dissociation energies and polarizability.
Pranoy Ray, Andrew R. Castillo, Manoj Kolel‐Veetil, Surya R. Kalidindi   +3 more
wiley   +1 more source

Machine‐Learning‐Guided Design of Incommensurate Antiferroelectrics via Field‐Driven Phase Engineering

Advanced Science, EarlyView.
The key to enhancing the energy storage performance of antiferroelectrics lies in regulating the phase transition and reverse phase transition. A phase‐field‐machine learning framework is employed to predict the energy storage performance of Pb‐based incommensurate antiferroelectrics with multi‐scale regulation strategy, thereby revealing the dynamic ...
Ke Xu, Xiaoming Shi, Zhaochen Xi, Shouzhe Dong, Changqing Guo, Rongzhen Gao, Letao Yang, Jing Wang, Di Zhou, Houbing Huang   +9 more
wiley   +1 more source