variant effect prediction

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

This paper evaluates generalist large language models (LLMs) in tumor genomic interpretation tasks, finding they can complement or outperform task-specific models in some cases.

The paper identifies autoimmune disease-associated variants that disrupt transcription factor-cofactor complex assembly and enhancer activity using the CASCADE method in Jurkat T cells.

This paper applies FINEMAP and SuSiE to fine-map schizophrenia risk loci, identifying new genes with high concordance between methods and validating them using LoFI and rare variant enrichment.

MARRVEL-MCP enables natural language variant interpretation by allowing LLMs to autonomously construct workflows using curated genomic tools, improving performance over models without tool access.

ParaDISM improves read placement and variant calling accuracy in highly homologous genomic regions by reducing misalignment artifacts.

The paper recalibrates BAP1 MAVE data for different disease phenotypes to enable accurate variant classification across related disorders.

This study identifies TaMYB83-7B as a wheat transcription factor regulating seed dormancy by modulating the ABA-GA balance, with specific promoter mutations linked to dormancy strength.

The paper predicts gene-rare disease associations using a network-based algorithm to improve variant prioritization and guide functional validation.

This paper introduces PfPan, a P. falciparum pangenome that improves structural variant detection and aids malaria control by identifying drug resistance and diagnostic evasion variants.

The paper derives an exact formula for the contribution of sampling error to r2, showing it equals 1/(S - 1) when sampling error is the only factor affecting LD.

This paper improves mRNA stability prediction by incorporating fine-scale structural metrics into a regression model, achieving better accuracy than existing methods.

The paper develops a proteomic polygenic score to predict IL-18 levels and IBD risk, identifying ~30% of IBD patients as IL-18 High.

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 Seq2Kd, a deep learning model predicting TF-DNA binding affinity using high-throughput Kd measurements, and provides a web portal for functional genomics and disease research.

This paper introduces MPRabc, an MPRA-informed model that improves enhancer-gene interaction prediction across cell types.

This study identifies a homozygous MGME1 variant in an Indian cohort causing mitochondrial dysfunction through impaired mtDNA replication and repair.

This paper compares two methods for creating calibrated prediction intervals for polygenic scores, showing their performance and the importance of normalization and preprocessing.

The paper compares deep learning models with pairwise-additive models for predicting TF-DNA binding affinity using high-throughput experimental data, showing improved accuracy with non-additive models.

This paper introduces ClinVar-BERT, a model that prioritizes variants for clinical review by analyzing variant summaries and distinguishing functional evidence, achieving high accuracy in predicting loss of function.

The paper compares automated and manual reanalysis of rare disease genome sequencing data, finding automated methods achieve similar results with less effort.

The paper introduces CANCAN, a computational framework for analyzing copy number and mutation heterogeneity in DNA sequencing data, showing high concordance with existing methods and improved performance on targeted sequencing data.

This paper develops a calibrated electrophysiological framework to classify CACNA1G missense variants, identifying two variants as likely pathogenic and others as benign.

EpiReasoner integrates AI for phenotype-to-genotype reasoning in plant epidermal development, identifying a major QTL for stomatal density in tomatoes.

The paper identifies genetic drivers of clinically relevant traits in Candida parapsilosis using genome-wide association studies and machine learning models.

HAUDI improves polygenic score construction for admixed populations by efficiently extending the GAUDI method to multi-way admixture with faster computation and better performance.

The paper combines enriched case data with population-based case-control analyses using a likelihood-ratio methodology to classify variants in breast cancer susceptibility genes.

This paper evaluates Evo 2's ability to predict pathogenicity and mechanisms of cardiomyopathy variants, showing high accuracy and structural feature identification.

HiFiMAP provides high-resolution IBD mapping at SNP level, outperforming window-based methods and detecting novel associations in UK Biobank and brain imaging data.

This paper benchmarks synthetic data generation methods for clinical transcriptomic cancer data, finding no single method outperforms all metrics, with Gaussian Copula showing the most balanced performance.