Sara Mostafavi

The paper introduces immgenT-GP, a gene-program framework that identifies 200 reproducible gene programs in mouse T cells, capturing lineage identity, activation states, and tissue location.

This paper introduces SAGE-net, a framework for training sequence-to-function models using personal genomes, showing improved gene expression prediction but limited generalization, while DNA methylation shows better generalization.

The paper evaluates AlphaGenome's performance in predicting gene expression and chromatin accessibility from personal genomes, finding it excels in accessibility but struggles with gene expression.

The paper improves sequence-to-activity models by incorporating ATAC-seq profiles, enhancing predictions of chromatin accessibility across immune cell types.

GAME provides a standardized framework for benchmarking genomics models across tasks, enabling consistent evaluation and community contributions.

The paper describes a germline variant in IKZF2 causing immune dysregulation, including AIHA and atopic dermatitis, and reports single-cell RNA sequencing findings on T cell dysfunction.

This paper introduces DeepAllele, a deep learning model for allele-specific gene regulation analysis, showing improved detection of cis-regulatory elements in F1 hybrid mice immune cells.

The paper creates a curated MPRA dataset collection and builds sequence-to-function models to analyze how data size, diversity, and model complexity affect predictive performance.

This study benchmarks deep neural networks for predicting gene expression from DNA sequence across individuals and identifies limitations in predicting variant effects.

The paper introduces a method to approximate in silico saturation mutagenesis using first-order Taylor expansion, reducing computational cost while maintaining robustness.

ExplaiNN provides interpretable deep learning models for genomics tasks, achieving comparable performance to state-of-the-art methods while offering transparent predictions.

The paper inserts random 50-base sequences into yeast introns and finds that most reduce splicing efficiency, likely due to RNA secondary structures, but existing models poorly predict these effects.

Gemini integrates hundreds of gene networks using memory-efficient high-order pooling, improving protein function prediction performance.

The paper identifies small molecules that enhance stem cell differentiation into definitive endoderm, reducing costs by 90% through a drug repurposing approach.

The paper identifies cell-type and cell-subtype specific eQTLs in aging human cortex, revealing variant effects on gene expression and disease associations.

This study shows that region-specific gray matter myelin changes in rats and humans correlate with stress-induced behavioral differences, suggesting a common mechanism across species.

CEWAS improves gene detection from GWAS by modeling cascading epigenomic and transcriptomic effects, identifying disease-related genes and novel biological processes.

The paper quantifies mosaic loss of chromosome Y (LOY) in brain cells across multiple datasets and finds microglia are most affected by LOY, with associated gene expression changes linked to aging and neurodegeneration.

This study shows that transfer learning improves transcription factor binding prediction by pre-training with biologically relevant TFs and fine-tuning single-task models.

The paper introduces MD-AD, a deep learning framework that identifies non-linear gene expression-neuropathology relationships in Alzheimer's disease using multi-cohort data, revealing sex-specific immune response patterns.

This paper performs a large-scale eQTL meta-analysis across cerebral cortical and cerebellar brain regions, identifying distinct eQTL patterns and demonstrating improved power over existing studies.

The paper evaluates methods for estimating cell-type proportions in brain tissue using paired transcriptomic and immunohistochemistry data, showing some algorithms provide accurate estimates.

The paper identifies novel modifier genes for Huntington disease by combining population-scale gene expression data with patient genomic data.

This paper introduces DEcode, a deep learning model that improves differential gene expression prediction by integrating systems biology data, showing post-translational RNA-binding factors are key regulators.

metPropagate uses metabolomics data to prioritize candidate genes for neurometabolic diseases, achieving high accuracy in identifying causative genes for both known and novel IEMs.

This study uses multi-omic profiling of blood samples from ALD patients to identify molecular markers predictive of cerebral demyelination, but finds no statistically significant markers due to high inter-individual variation.

This study identifies traits associated with polygenic risk for depression using UK Biobank data and finds causal links between depression risk and brain structure, as well as environmental interactions.

This paper introduces AI-TAC, a convolutional neural network that infers cell-type-specific chromatin activity from DNA sequences using an ATAC-seq atlas of immune cells.

The paper develops PlaNET, a DNA methylation-based ethnicity classifier for placental samples, achieving high accuracy and outperforming existing methods in population stratification correction.

This study investigates whether genetic variation in the Major Histocompatibility Complex (MHC) is associated with depression risk, finding no significant associations at a conservative threshold.