Recent progress in modeling protein structures and interactions suggests that high-resolution prediction and design will significantly contribute to biology and medicine.

Progress in Modeling of Protein Structures and Interactions

Our graph convolutional model consistently outperforms existing models in molecular property prediction, offering significant improvements over existing industrial workflows.

Analyzing Learned Molecular Representations for Property Prediction

Advanced computational methodologies enable accurate molecular structure determinations for small and medium-sized molecules, crucial for understanding their chemical and physical properties.

Accurate molecular structures of small- and medium-sized molecules

This theory demonstrates that atoms and bonds can be defined and stabilized using topological properties of the observable distribution of charge, linking molecular structure to quantum mechanics.

A topological theory of molecular structure

Convolutional neural networks can effectively predict physical properties of molecules by identifying important features of atom clusters.

Convolutional Embedding of Attributed Molecular Graphs for Physical Property Prediction

A joint theoretical-experimental approach can accurately determine molecular structures, addressing challenges in understanding thermochemistry, kinetics, and spectroscopic signatures of molecular systems.

Diving for Accurate Structures in the Ocean of Molecular Systems with the Help of Spectroscopy and Quantum Chemistry.

Graph-To-Structure (G2S) effectively predicts atomistic structures of molecules, transition states, and solids without energy optimization, improving accuracy and efficiency in physics, chemistry, materials, and biology.

Machine learning based energy-free structure predictions of molecules, transition states, and solids

Current methods for predicting molecular organic crystal structures show progress and limitations, aiding in materials design and avoiding undesirable changes in form.

Current approaches to predicting molecular organic crystal structures

Electronic structure methods are crucial for modeling molecular crystals and predicting their properties, including polymorphism, for various applications in pharmaceuticals, organic semiconductor materials, and foods.

Modeling Polymorphic Molecular Crystals with Electronic Structure Theory.

Deep learning can rapidly predict the structural properties of liquids from a single molecular configuration, providing rapid insight into liquid structure under various pressure and temperature conditions.

Rapid Prediction of a Liquid Structure from a Single Molecular Configuration Using Deep Learning

UV resonance Raman spectroscopy has developed into a powerful tool for studying molecular structure and dynamics in both small and large molecules, with applications in physical and biophysical chemistry.

UV resonance Raman studies of molecular structure and dynamics: applications in physical and biophysical chemistry.

TransFoxMol improves molecular property prediction by focusing attention on key features, enhancing AI's understanding of molecular structure and property relationships.

TransFoxMol: predicting molecular property with focused attention

Rosetta is a protein structure prediction software that uses statistical information from known structures to predict the structure of proteins with unknown folds.

Protein Structure Prediction

Molecular dynamics simulations can accurately predict peptide and protein structures with similar accuracy to experiments, and can reproduce experimental 3D structures of small proteins.

Applications of Molecular Dynamics Simulation in Structure Prediction of Peptides and Proteins

Valence bond theory provides a direct interpretation of molecular mechanics formulations and parameters, enabling quantitative calculations and understanding of electronic effects in physical organic chemistry.

Understanding molecular structure from molecular mechanics

Protein structure refinement to experimental accuracy via molecular dynamics simulations is possible, but significant challenges need to be overcome in practical refinement protocols.

Experimental accuracy in protein structure refinement via molecular dynamics simulations

Students' appreciation of three-dimensional molecular structures improves with concrete and pseudo-concrete representations, while abstract types are less effective.

Students' understanding of molecular structure representations

MaSIF, a geometric deep learning method, can predict protein-ligand and protein-protein interactions by analyzing common chemical and geometric features on protein molecular surfaces.

Deciphering interaction fingerprints from protein molecular surfaces using geometric deep learning

Our method using long structural prototypes (LSPs) effectively predicts protein flexibility with a prediction rate of 49.6%, compared to cutting-edge methods based on true flexibility data learning with sophisticated algorithms.

Predicting protein flexibility through the prediction of local structures

How do we understand and predict molecular structures?

These studies suggest that understanding and predicting molecular structures can be achieved through advanced computational methodologies, machine learning models, and joint theoretical-experimental approaches, which are crucial for applications in biology, medicine, chemistry, and materials science.