Dina Schneidman‐Duhovny

The paper introduces SABLE, a benchmark for evaluating deep-learning methods in antibody structure prediction, with leakage-controlled test sets and versioned resources.

CLIPepPI predicts domain-peptide specificity using contrastive learning, achieving competitive performance across benchmarks and enabling proteome-wide analysis.

This paper shows that an orthosteric inhibitor of the COP9 signalosome achieves high potency through a molecular glue mechanism that enhances substrate-dependent binding.

This paper improves variant effect prediction by calibrating models at the residue level, enhancing both calibration and discrimination across variant subgroups.

The paper analyzes 1,100 SARS-CoV-2 antibody structures to show convergent paratope structures and near-universal immune escape by advanced variants.

The paper investigates the molecular mechanism of the Gαo E246K mutation in neurodevelopmental disorders, showing it disrupts the GTPase cycle by impairing Gβγ dissociation, leading to dominant-negative effects.

DockFormer predicts protein-small molecule binding affinity and structure using a flexible docking approach, outperforming existing methods on benchmark datasets.

SCOPER improves RNA structure validation using SAXS data by incorporating ion binding prediction and conformational sampling.

The paper identifies EspH's interactions with Rab GTPases and phosphoinositides, showing their roles in immune signaling and membrane trafficking during EPEC infection.

The paper introduces CombFold, a method that uses AlphaFold2 to predict large protein complexes with high accuracy, achieving 72% success in Top-10 predictions and improved structural coverage.

This study identifies new DUX4 protein interactions and activation pathways using AP-MS and CL-MS, revealing structural and functional insights into its role in FSHD.

The paper introduces MHCfold, a deep learning model that predicts peptide-MHC class I structures and binding specificity, achieving high accuracy and enabling high-throughput analysis.

ScanNet predicts protein binding sites using an interpretable geometric deep learning model, achieving accuracy on unseen protein folds and identifying epitopes in SARS-CoV-2 spike protein.

NanoNet is a deep learning model that predicts high-resolution 3D structures of nanobody variable domains from sequences, achieving sub-angstrom accuracy and enabling rapid high-throughput modeling.

The paper identifies a de novo STXBP1 mutation in a child with developmental delay and spasticity, showing how it disrupts the STXBP1-syntaxin 1A interaction through structural analysis.

The paper reports three patients with biallelic EXOSC5 variants causing developmental delays, short stature, and cerebellar hypoplasia, and shows functional defects in RNA exosome activity.

The paper uses molecular dynamics simulations to compare SARS-CoV-2 and other coronaviruses' interactions with the human ACE2 receptor, revealing distinct binding interfaces and dynamics.

This paper introduces ITCell, a method that predicts CD4 T-cell epitopes by integrating antigen cleavage, MHCII presentation, and TCR recognition, achieving higher accuracy than single-stage methods.