Chengxin Zhang

The paper predicts the structure and function of all proteins in uropathogenic E. coli UTI89 using computational methods and identifies novel host-microbe and microbe-phage interaction pathways.

StarFunc combines deep learning with template information from multiple sources to improve protein function prediction, outperforming existing methods in benchmark tests.

StoPred predicts protein complex stoichiometry using protein language models and graph attention networks, achieving higher accuracy for both homomeric and heteromeric complexes than existing methods.

The paper predicts the structural impact of human alternative splicing using AlphaFold2 on over 11,000 isoforms and identifies structural changes across multiple metrics.

EZpred improves enzyme function prediction by using unlabeled sequence homologs, achieving higher F1-scores than existing methods.

The paper quantifies UFMylation sites in mouse tissues and ALS patient biopsies, finding increased myosin UFMylation in ALS.

InterLabelGO+ improves protein function prediction by combining deep learning and alignment-based methods with a novel loss function and dynamic weighting, achieving strong performance in the CAFA5 challenge.

FURNA provides a database of functional annotations for experimental RNA structures, including Gene Ontology terms and ligand binding sites.

This study evaluates sequence database search tools for homology-based protein function prediction, finding that BLASTp and MMseqs2 outperform DIAMOND under default parameters, but DIAMOND can match their performance with proper settings.

BeEM converts mmCIF structure files to PDB format accurately and efficiently, handling complex chain IDs and outperforming existing tools.

This paper introduces DRfold, a method that improves RNA tertiary structure prediction by integrating deep learning with geometric potentials, achieving significant performance gains over existing methods.

US-align provides a universal platform for aligning diverse macromolecular structures, improving accuracy and speed in structural comparisons.

The paper introduces ATGO, a deep-learning method that improves Gene Ontology prediction for proteins using self-attention transformers and triplet networks, achieving higher accuracy than existing methods.

TripletGO improves gene function annotations by integrating transcript expression, sequence alignment, and probability models across multiple species and proteins.

This paper introduces DEMO-EM, a method for assembling multi-domain protein structures from cryo-EM maps, achieving high accuracy and outperforming existing methods.

This study identifies RTL8 as a novel UBQLN2 interactor that promotes its nuclear translocation and subnuclear localization in protein quality control.

TripletRes predicts high-accuracy protein contact-maps from primary sequences using deep residual convolutional networks and coevolutionary matrices, outperforming existing methods on benchmark datasets.

The paper develops a pipeline integrating multi-chain threading with high-throughput experiments to improve PPI detection and structural modeling in E. coli, achieving higher confidence PPIs and accurate complex structures.

The paper models myoglobin structures in 302 vertebrates, finds positive selection for aggregation resistance in penguin myoglobins and purifying selection in galliforms, and proposes a new equation for net charge and concentration relationship.

This study identifies potential intermediate hosts of SARS-CoV-2 by modeling Spike-RBD binding affinity to ACE2 proteins across multiple species.

The paper re-analyzes computational findings from two studies, concluding that snakes are not intermediate hosts of 2019-nCoV and that spike protein insertions are naturally evolved from bat coronaviruses.

The paper reports results from the CAFA3 challenge, showing improved computational protein function prediction and new annotations for over 1000 genes through experimental screens in C. albicans, P. aureginosa, and Drosophila.