variant effect prediction

MHCBind is a deep learning model that predicts MHC class I peptide binding affinity across alleles and improves upon existing methods with better accuracy and ranking performance.

This study evaluates sparse testing designs for genomic prediction in Miscanthus breeding to reduce phenotyping costs while maintaining predictive accuracy.

A ProtT5-embedding, ConvNext-plus-attention ensemble predicts human pathogenic capacity from whole bacterial genomes, outperforming prior predictors on species absent from training and flagging influential proteins.

This study evaluates uncertainty quantification methods to improve reliability of genomic language models under distribution shift in genomic and metagenomic tasks.

The paper introduces enhanced features in selscan v3.0 for analyzing selection signals using EHH statistics and demonstrates their application on simulated and real genomic data.

The paper identifies deleterious mutations in sugar beet by combining evolutionary models and deep learning, showing a decline in genetic load over time due to breeding.

The paper benchmarks five machine learning-based variant annotation methods for rare variant association testing across 350,377 UK Biobank participants, finding CADD provides best signal separation and GPN-MSA highest enrichment for loss-of-function intolerant genes.

This paper uses machine learning on MALDI-TOF mass spectrometry data to detect ampicillin susceptibility in Enterococcus faecium, achieving high accuracy across datasets.

The paper identifies proteins where variant effect predictors disagree, suggesting these are prioritized for experimental validation to improve variant interpretation.

G-VEP accelerates variant effect prediction in clinical whole-genome sequencing by using GPU acceleration, reducing runtime from 72 minutes to 4 minutes with no loss of accuracy.

The paper shows that pretraining deep neural networks on human genomic data improves prediction accuracy and reduces training time when applied to diverse species and tissues.

This study shows extending the spacer in SuperFi-Cas9 restores cleavage activity while maintaining specificity, offering a strategy to improve CRISPR efficiency and stability.

The paper quantifies selection efficiency across Escherichia coli lineages by linking mutation scores to population genetics, showing significant variation in selection intensity between synonymous and non-synonymous mutations and pathogenic populations.

This paper introduces ProteinEBM, a protein conformational energy model that outperforms existing methods in predicting mutation effects and protein structures.

This paper presents a structure-aware framework for interpreting genomic variants in genetic skeletal disorders by integrating structural data and variant annotations.

This study quantifies how mutations in GATA1 affect DNA binding affinity using a new assay, revealing that certain mutations reduce binding while others increase it.

VarDCL predicts pathogenic missense variants by integrating multimodal protein language model embeddings and self-distilled contrastive learning, achieving high performance on clinical variants.

CoNVict is an agentic AI system that prioritizes copy number variants in rare disease diagnosis by ranking candidates through structured reasoning, outperforming existing methods in identifying causal variants.

This study uses tensor decomposition to analyze CRISPR-Cas9 knockout screens, distinguishing essential from non-essential genes with performance comparable to JACKS without requiring control samples or logarithmic transformations.

The paper introduces ENGINE, a method to infer gene-environment interactions from large-scale biobank data, finding that combining multiple exposures increases detected GxE variance.

This paper introduces DiffEvol, a constrained diffusion framework to model evolution, and applies it to SARS-CoV-2 data to reconstruct fitness trends and vaccine-induced phase transitions.

This paper introduces an explainable boosting machine model to identify and filter variant artifacts caused by formalin-fixed paraffin embedding in cancer genomics.

The paper identifies stop-loss mutations in cancer genomes, showing enrichment in cancer-related genes and potential immunological consequences.

The paper introduces a method to compute Bayes factors for linear mixed models in genetic association studies using Bayesian regression and demonstrates its equivalence to BLUP predictions.

The paper establishes frequency criteria for secondary findings genes to improve variant classification consistency by setting disease allele frequency thresholds for 58 monogenic disease entities.

This paper introduces MR-UBRA, a Bayesian framework for analyzing genetic heterogeneity in Mendelian randomization, showing it can resolve multi-component structures in causal effect space.

ACENet predicts enzyme pHmin using a graph neural network integrating structural and evolutionary features, achieving high accuracy and stability across conditions.

This paper evaluates neural network-based methods for local ancestry inference in challenging admixture scenarios and finds they perform best with specialized preprocessing.

The paper applies a Sire MGS model to partition genetic variance into paternal and maternal components for lifetime productivity in tropical dairy cattle, finding significant maternal contributions.

Rosalind, a DNA foundation model, predicts variant-gene regulatory relationships without proximity bias, identifying distal genes as more impactful in osteoporosis.