Li Shen

EG-BINN improves epigenetic age estimation by combining parameter-efficient ensembling of biologically informed neural networks with sample-specific gating, achieving lower error rates than existing methods.

The paper develops a platform integrating multisource target evidence with machine learning for Alzheimer's drug discovery, improving potency prediction and model performance.

This paper develops Scratch-AID+, an automated mouse scratching detection framework, showing that fine-tuned video foundation models outperform classic models in generalizability and accuracy.

The paper quantifies the divergence between polygenic and proteomic risk scores in neurodegenerative diseases by analyzing absorption and co-expression modules across multiple cohorts.

The paper identifies candidate regulators and signaling axes in Alzheimer's disease lipid-processing microglia through multi-omics analysis across multiple cohorts.

The paper introduces DG-UMAP, a supervised dimensionality reduction method that improves class structure preservation in biomedical data while maintaining local geometry.

JASMINE improves analysis of incomplete multi-omics data by learning representations that preserve modality-specific and shared information, achieving better performance across tasks.

The paper uses machine learning to predict autopsy-confirmed neuropathology from clinical, neuroimaging, and CSF biomarkers, finding that different modalities best predict specific pathologies.

MethylCurate automates DNA methylation dataset retrieval and metadata harmonization from Gene Expression Omnibus, enabling scalable evaluation of epigenetic aging clocks.

This study improves modeling of amyloid-β spread in Alzheimer's Disease using a gene-modulated network diffusion framework, showing better performance in correlation and replicability across cohorts.

This study compares deep learning methods to predict cognitive decline using brain MRI and clinical data over two years, showing the value of integrating image and tabular data.

GROTIA aligns multi-omics single-cell datasets using optimal transport and graph regularization, identifying key biological markers and enabling novel subpopulation discovery.

The paper identifies nine neuroimaging endophenotypes linked to genetic loci and shows they improve prediction of systemic diseases and mortality.

This paper introduces UNITO, a deep learning framework for automated cytometric gating that matches human performance and provides interpretable results without requiring manual input.

The paper uses tiled whole genome sequence data to identify genetic signatures associated with Alzheimer's disease using machine learning methods.

This study identifies genetic subtypes of Alzheimer's Disease using cell-based gene networks and polygenic risk scores, then repositions existing drugs for AD treatment.

The paper uses GWAS summary statistics to analyze the genetic architecture of 2024 multi-organ AI endophenotypes, revealing genetic correlations and causal relationships across diseases.

This study introduces H-VIP to quantify vertex contributions to topological cycles in brain networks and finds a positive correlation with cognitive measures in both structural and functional connectomes.

This study identifies sixteen genome-wide significant loci associated with brain age gaps derived from MRI modalities and highlights potential causal relationships between brain aging and diseases.

This paper introduces QOT, a method for efficiently computing sample-level distance matrices from single-cell omics data, showing improved accuracy and robustness over existing OT-based approaches.

The paper introduces a homological formalism to analyze higher-order properties of human brain functional sub-circuits and identifies task-specific homological differences in brain connectivity.

The paper uses genetic data to predict imaging-derived phenotypes from the UK Biobank, creating polygenic scores that identify disease risk in external cohorts.

This study identifies and validates two molecular subtypes within early and late mild cognitive impairment stages of Alzheimer's disease using integrated metabolomics and transcriptomics data.

The paper identifies 617 novel genomic loci and pathways associated with structural covariance patterns in the human brain using a stochastic multivariate factorization method.

This study identifies genetic influences on brain connectivity across age groups using GWAS and finds a consistent effect of rs7937515 on network segregation in the left middle temporal gyrus.

The paper performs genome-wide association analysis on Alzheimer's disease endophenotypes to identify main effects and stage-specific genetic interactions.

This study uses an epidemic spreading model to show that tau spreads through neuronal connections in Alzheimer's disease, with beta-amyloid accelerating this process in some brain regions.

This study compares whole hippocampus and subfield biomarkers for Alzheimer’s disease detection and prediction using multimodal MRI data, finding subiculum biomarkers improve prediction accuracy.

The paper develops deep learning models to assess the quality of subcortical brain shape models from MRI, reducing human rater time by 46-70%.

The paper examines the heritability of regional brain volumes using genetic and neuroimaging data from multiple cohorts, finding high heritability and consistent patterns across populations.