David Benjamin Ascher

This study determines the crystal structure of AbTir, revealing how its coiled-coil domain inhibits NAD+ nucleosidase activity by disrupting symmetry and preventing NAD+ binding site formation.

This paper uses prime editing to classify LDLR variants by measuring their impact on LDL-C uptake, reclassifying many previously unclassified variants.

This study uses Methylscape to track breast cancer progression by detecting epigenetic changes during EMT and developing a low-cost platform for monitoring.

This study identifies Copine-6 as a protein that links calcium, phosphoinositide, and Rab11 signals to regulate glutamate receptor recycling in neurons.

This paper develops UshEffect-3D, a variant effect predictor for USH2A missense variants using sequence and predicted structure features, reclassifying 37.2% of ClinVar VUS as likely pathogenic.

CharacTERT classifies hTERT missense variants using machine learning models that integrate sequence and structure-based features, achieving high accuracy and sensitivity in predicting pathogenicity.

The paper introduces a metric to quantify conformational noise in structural alphabets and shows its impact on phylogenetic inference using NMR ensembles and molecular dynamics simulations.

PhenotypeToGeneDownloaderR retrieves and validates phenotype-associated genes across multiple databases, achieving high recall and providing a reproducible framework for downstream analysis.

AnnotateMissense integrates and benchmarks missense variant pathogenicity predictors using genome-wide data, achieving high performance with XGBoost.

This study benchmarks 86 heritability estimation configurations and evaluates their impact on polygenic risk score performance across 10 UK Biobank phenotypes.

The paper introduces the Duplicate Monophyly Criterion to calibrate distance-matrix perturbations for phylogenetic support estimation using synthetic duplicates.

This paper benchmarks 46 polygenic risk score tools across binary and continuous phenotypes, showing no single method is universally optimal due to varying factors like phenotype architecture and computational demands.

This study benchmarks machine learning, deep learning, and polygenic risk score tools on 80 binary phenotypes from openSNP data to evaluate their predictive performance.

The paper uses machine and deep learning to identify phenotype-associated genes by classifying cases/controls and calculating feature importance, achieving an 84% gene identification ratio.

GWASPoker is a tool that identifies and processes GWAS summary statistics for polygenic risk score calculation by scanning files for column availability without full download.

Structome-TM assembles datasets for structural phylogenetics by prioritizing local structural similarity to address size-based biases in homologous structure identification.

Structome-AlignViewer evaluates structure-aware alignments by mapping alignment columns to protein structures and providing confidence scores to identify reliable regions.

This paper presents a 9-step pipeline for fine-tuning LLMs on bioinformatics data, demonstrating PRSGPT and BioStarsGPT with performance metrics and QA datasets.

The paper introduces AMRscope, a protein language model-based tool for predicting antimicrobial resistance likelihood from single mutations, showing strong performance across bacterial species and identifying new resistance candidates.

The paper introduces a quantum surrogate model framework for biophysical landscapes, demonstrating its ability to simulate transitions between localized and delocalized states using variational quantum eigensolver simulations.

The paper demonstrates a quantum-inspired kernel achieves high accuracy and Matthews correlation coefficient in classifying protein torsion angles, matching classical models.

This paper introduces DDMuffin, a deep learning framework for predicting protein-ligand binding affinities and mutation-induced changes, showing strong performance on benchmark datasets and offering insights for drug design.

The paper presents a dashboard for comparing and visualizing pathogenicity predictions of missense mutations in the human proteome within a 3D structural context.

PRIMITI predicts miRNA-target mRNA interactions using machine learning with CLIP-seq and expression data, achieving high AUC scores and outperforming existing methods.

AI-m6ARS predicts human m6A methylation sites using sequence, conservation, and geographical features, achieving high accuracy and offering a user-friendly web server.

This study evaluates pathogenicity predictors for GPCR mutations using ClinVar and MutHTP datasets, finding reliable performance on ClinVar but poor performance on MutHTP, and identifies key mutation characteristics linked to disease.

This study identifies chemically constrained peptides that inhibit PfUCHL3, a critical deubiquitinating enzyme in Plasmodium falciparum, by blocking substrate binding rather than active site interaction.

This paper improves variant effect quantification in base editing screens by integrating genotypic and phenotypic data using a Bayesian network called BEAN.

The paper uses structure-based hidden Markov models to predict protein folds and infer function, particularly for proteins with weak sequence homology.

This paper uses tertiary-interaction characters to improve structural phylogenetics beyond the twilight zone, achieving accurate results with fast computation.