Roland L. Dunbrack

This paper introduces ipSAE, a directional scoring framework for assessing confidence in predicted biomolecular complexes, showing improved performance over existing methods in protein-ligand and protein-DNA interactions.

The paper identifies a germline KDM3C polymorphism linked to impaired DNA repair and increased sensitivity to chemoradiotherapy in cancers like rectal and head and neck squamous cell carcinomas.

The paper reveals how phosphorylation of SMO in the Hedgehog pathway stabilizes a conformation that inhibits PKA-C, providing a structural mechanism for noncanonical GPCR signaling.

The paper defines structural criteria for the active form of human protein kinases using substrate-bound structures and uses AlphaFold2 to generate active-state models for all 437 human catalytic kinases.

The paper predicts functional E2-E3 ligase pairings using structural analysis and machine learning, identifying new interactions and providing a web resource for ubiquitination network mapping.

The interaction prediction Score from Aligned Errors (ipSAE) reformulates AlphaFold's interface predicted template modeling (ipTM) metric using filtered interchain predicted aligned errors, improving separation of true and false protein complexes.

This study shows that Id3 deficiency affects the development of gamma delta T cell subsets by regulating the generation and function of the Vgamma1Vdelta6.3 TCR.

This paper identifies NGL-1 as a stromal immunomodulator in pancreatic cancer that promotes tumor growth by modulating the tumor microenvironment through TGF-β and transcription factor pathways.

With active-state templates and shallow ortholog alignments, AlphaFold2 models all 437 human catalytic kinase domains as active; 80% of 130 benchmarked kinases are under 1 Å activation-loop RMSD.

This paper identifies FAM122A as a SLiM-dependent inhibitor of B55/PP2A that competes with substrates to regulate cell cycle progression and checkpoint control.

The paper introduces ProtCAD, a database that clusters protein assembly structures from independent experiments to improve accuracy in biological assembly annotations.

The paper identifies DNA repair and replication defects in early-onset renal cancer patients without known germline variants, suggesting these defects contribute to cancer development and may inform treatment strategies.

The paper classifies over 700 RAS structures into conformations based on SW1 and SW2 loops, identifies mutation impacts, and creates a database for analysis.

Kincore provides a web resource for classifying kinase conformations and inhibitor types using structural data from the PDB.

PDBrenum renumbers PDB files to align with UniProt sequences, resolving inconsistent amino acid numbering across PDB entries.

The paper introduces a new test set, a convolutional neural network model, and evaluates factors affecting accuracy in protein secondary structure prediction.

The paper builds a structure-supported human interactome by integrating structural information with PPI data, identifying domain-domain and peptide-binding interactions to improve the resolution and biological relevance of the network.

ProtCID is a database that provides structural information on protein interactions, including domain-level analysis and annotations for clusters of protein interfaces.

The paper analyzes the structure and sequence of the antibody DE loop, revealing its role in antigen binding and its variability in HIV-1 antibodies.

The paper presents a structure-based multiple sequence alignment of 497 human protein kinase domains and identifies ten kinases that can be confidently placed into the CAMK group.

The paper analyzes biological assemblies in protein structures, highlighting challenges in determining correct assemblies and recent computational methods that use interface similarity.

The paper clusters kinase activation loop conformations using structural features and introduces a nomenclature to classify active and inactive states based on DFG motif positioning and backbone angles.

The paper clusters beta turns from high-resolution protein structures into 18 types using DBSCAN and k-medoids, proposing a new nomenclature and providing tools for analysis.