ARTIFICIAL INTELLIGENCE PROVES ITS PROTEIN-FOLDING POWER

Article Multidisciplinary Sciences

Computed structures of core eukaryotic protein complexes

Ian R. Humphreys et al.

Summary: By utilizing proteome-wide amino acid coevolution analysis and deep-learning-based structure modeling, we systematically identified and built accurate models of core eukaryotic protein complexes within the Saccharomyces cerevisiae proteome. These models cover nearly 800 previously uncharacterized assemblies, providing comprehensive insights into key processes in eukaryotic cells.

SCIENCE (2021)

Add to Collection

Article Multidisciplinary Sciences

Highly accurate protein structure prediction with AlphaFold

John Jumper et al.

Summary: Proteins are essential for life, and accurate prediction of their structures is a crucial research problem. Current experimental methods are time-consuming, highlighting the need for accurate computational approaches to address the gap in structural coverage. Despite recent progress, existing methods fall short of atomic accuracy in protein structure prediction.

NATURE (2021)

Add to Collection

Article Multidisciplinary Sciences

Highly accurate protein structure prediction for the human proteome

Kathryn Tunyasuvunakool et al.

Summary: Using the AlphaFold method, the structural coverage of the proteome has been significantly expanded, covering 98.5% of human proteins with 58% of residues having confident predictions and 36% having very high confidence. Introducing new metrics to interpret the dataset and identify disordered regions, this study aims to provide high-quality predictions for generating biological hypotheses.

NATURE (2021)

Add to Collection

Article Biochemistry & Molecular Biology

Protein structure prediction using deep learning distance and hydrogen-bonding restraints in CASP14

Wei Zheng et al.

Summary: In this article, 3D structure prediction results by two top server groups (Zhang-Server and QUARK) in CASP14 were reported, with significant improvements achieved by introducing new components, particularly the newly added spatial restraints from DeepPotential and the well-tuned force field. However, challenges remain in modeling multi-domain proteins and protein domains from oligomer complexes, suggesting the need for further adjustments in deep learning-based predictors to address these issues.

PROTEINS-STRUCTURE FUNCTION AND BIOINFORMATICS (2021)

Add to Collection

Article Multidisciplinary Sciences

Accurate prediction of protein structures and interactions using a three-track neural network

Minkyung Baek et al.

Summary: Through the three-track network, we achieved accuracies approaching those of DeepMind in CASP14, enabling rapid solution of challenging x-ray crystallography and cryo-electron microscopy structure modeling problems, and providing insights into the functions of proteins with currently unknown structure.

SCIENCE (2021)

Add to Collection

Article Multidisciplinary Sciences