Genomics and Rare Diseases

This paper describes the Rare Genomes Project's efforts to address inequities in rare disease genomics by diversifying enrollment and understanding barriers to access.

The paper analyzes VUS rates in genetic testing and finds lower rates in exome/genome sequencing compared to multi-gene panels, suggesting improvements in variant reporting practices.

The paper introduces cost models to compare pathogenicity predictors in clinical contexts, showing no optimal predictor exists for all scenarios and highlighting the importance of rejection classifiers.

The paper re-analyzes 152 exomes from developmental disorder families five years later, finding 20% clinically relevant changes and reclassifying variants.

The paper identifies 267 genes reclassified as protein-coding in GRCh38 that were previously labeled as non-protein-coding in GRCh37, highlighting the impact of outdated annotations on variant prioritization and clinical diagnostics.

The paper assesses improvements in genomic variant classification accuracy over time in ClinVar and HGMD by analyzing misclassification rates in inborn errors of metabolism across ancestry groups.

The paper combines expert guideline-based and machine learning approaches to improve automated prediction of the clinical effect of copy number variations.

This study evaluates clinical genome sequencing (GS) over three years at a children's hospital, finding a 35% diagnostic yield and frequent misordering, with higher yields in patients without prior exome sequencing.

This study benchmarks 55 variant effect predictors using deep mutational scanning data from 26 human proteins, finding that unsupervised methods like EVE and ESM-1v perform best, while supervised methods like VARITY show improved handling of data circularity.

SPRI predicts the pathogenicity of missense variants and identifies spatially organized mutational clusters linked to cancer and Mendelian diseases.

The paper uses primate genetic data to predict human variant pathogenicity with a deep learning classifier trained on 4.3 million common missense variants.

The paper identifies discordant genotype calls across sequencing methods in gnomAD, provides a blacklist of problematic variants, and introduces a classifier to detect such variants in other datasets.

The paper shows that cotranslational protein complex assembly is less common for subunits with dominant-negative mutations and develops a model to predict such mutations.

The paper compares protein stability predictors with DMS-derived functional scores, finding FoldX and Rosetta show strong correlations, especially when considering intermolecular interactions.

This paper introduces a tissue-aware machine learning framework to improve understanding and diagnosis of Mendelian and rare diseases by leveraging tissue-specific gene mechanisms.

The paper introduces REGatta, a method to estimate clinical risk in gene regions using population sequencing data, identifying significant risk differences in breast cancer genes for rare missense variants.

The paper investigates why studies fail to find associations between de novo noncoding variants and complex disorders like ASD, identifying annotation quality and variant reliability as key issues.

This paper introduces Sequence UNET, a deep learning model for variant effect prediction that is highly scalable and computationally efficient, achieving comparable performance to existing methods on ClinVar data.

The paper identifies clinically relevant variants in poison exons (PEs) that improve diagnostic yield for neurodevelopmental disorders by uncovering variants previously overlooked in standard analyses.

This study compares the diagnostic yield of genome sequencing and targeted gene panel testing in diverse pediatric patients, finding higher yields with genome sequencing in most groups but not all.

The paper analyzes genetic disorders and variants in Emiratis using the CTGA database, highlighting high prevalence of rare, Mendelian conditions and discrepancies in variant classification.

The paper introduces AlphScore, a pathogenicity prediction score for missense variants using AlphaFold2-derived features, showing improved performance when combined with existing scores.

satmut_utils is a tool for simulating and analyzing saturation mutagenesis data, showing that single-codon variants can alter mRNA abundance and demonstrating improved variant calling performance.

DeMAG improves variant classification by integrating structural and evolutionary epistatic features, reducing false positives and improving accuracy for clinically actionable genes.

This study validates variants of uncertain significance in ECHS1 using functional assays and multi-omics analysis, identifying new pathogenic variants and cases of mitochondrial disease.

POSTRE predicts the pathogenicity of structural variants in human congenital disorders by identifying disease-causative genes and mechanisms.

Missense3D-PPI predicts the impact of missense variants on protein-protein interfaces using structural data, showing improved sensitivity over previous methods.

This study uses a computational framework to analyze genetic data from 77,539 genomes, identifying 21 novel genetic associations with 269 rare diseases.

KidneyNetwork predicts kidney-specific gene functions and disease phenotypes using co-expression networks, identifying ALG6 as a candidate gene for kidney and liver cysts in undiagnosed patients.