Haiyuan Yu

This study shows KAT6 and KAT7 maintain mitotic chromatin acetylation and composition, supporting postmitotic transcriptional memory and developmental competence in stem cells.

This study compares divergent and unidirectional transcriptional regulatory elements, finding that unidirectional elements are younger and less constrained, with some transcription factors modulating their directionality.

This study identifies a conserved role of cystinosin/Ers1 in the early secretory pathway, revealing new insights into cystinosis pathology beyond lysosomal cystine transport.

This study identifies TRAPPC8 as a brake on UFMylation that suppresses tauopathy by interacting with UBA5, showing its therapeutic potential in Alzheimer's disease.

This study identifies UBE3C as a longevity-associated E3 ligase that maintains nuclear lamina integrity by modulating LMNB1 and LBR through selective autophagy, delaying senescence.

The paper investigates enhancer-promoter interactions in a native chromatin context, revealing transcription factor synergy and the role of facilitators and repressive promoters in regulating gene activation.

This paper introduces DeepDETAILS, a method to reconstruct cell-type-specific genomic signals from bulk data, outperforming traditional methods and providing insights into disease mechanisms.

This paper investigates the role of POGZ in maintaining chromatin architecture and transcription in the developing mouse cortex, finding that its loss disrupts H3K9me3 distribution and impairs neuronal gene expression.

The paper develops a neuron-specific interactome map in human excitatory neurons to identify novel risk genes and pathways in Alzheimer's disease.

The paper introduces CIERA-seq to study transcription regulatory element interactions and finds pdTREs and edTREs have distinct functional and chromatin features.

This study identifies a SWI/SNF-targeting therapy effective against a double-negative prostate cancer subtype, showing it reduces viability in both AR-dependent and WNT-dependent models.

This paper optimizes liquid chromatography conditions for single-cell proteomics across platforms, demonstrating its effectiveness even on non-specialized instruments and showing its ability to detect biological responses to oxidative stress.

This study evaluates multiple MPRA and STARR-seq datasets to identify enhancers, finding inconsistencies due to technical factors and proposing a unified pipeline to improve cross-assay agreement.

The paper identifies human RNA Polymerase II interactors at early transcription stages using biochemical purification and mass spectrometry, revealing new factors involved in PIC assembly and promoter-proximal pausing.

The paper identifies a metabolic connection between PI4P and PA at the plasma membrane, showing how PI4P depletion increases PA levels through PS synthesis and RhoB activation.

The paper presents a high-resolution atlas of transcriptional regulatory elements in human tissues and disease states using PRO-cap and deep learning models to prioritize noncoding variants.

This paper introduces a deep learning framework for predicting disease-specific variant pathogenicity, showing improved accuracy over existing methods.

This study identifies the A-type lamin rod domain as the key determinant of nuclear deformability, revealing its unique role in resisting large mechanical forces on the nucleus.

The paper introduces PIONEER2, a method that improves protein-protein interface prediction by integrating structural homology and deep learning, outperforming AlphaFold3 in certain scenarios.

The paper predicts 3D structures of viral-host protein interactions and identifies evolutionary patterns and functional mechanisms underlying viral pathogenesis.

The paper develops a dual reporter assay to measure promoter and enhancer potential, finding that they share regulatory logic and can function reciprocally.

This paper introduces NetFlow3D, a framework that integrates protein structure and interaction networks to identify significant mutation clusters and modules in cancer, revealing biological processes not found by traditional methods.

MAGMa is a tool for differential expression analysis in mass-spectrometry proteomic data, showing improved sensitivity and specificity in detecting true signals.

The paper trains a deep learning model to predict transcription initiation from DNA sequences, revealing shared cis-regulatory logic and epistatic interactions across promoters and enhancers.

The paper identifies stress-induced submitochondrial protein translocation as a mechanism regulating thermogenesis and obesity by mobilizing proteins from the mitochondrial inner membrane into the matrix.

The paper introduces PIONEER, a deep learning framework for predicting protein-protein interfaces, and shows disease mutations are enriched in these interfaces, identifying oncoPPIs linked to cancer survival and drug responses.

DeepERA uses deep learning to predict drug-target interactions by integrating heterogeneous data, identifying 45,603 novel interactions including those related to SARS-CoV-2.

This study uses PRO-cap to identify drug-responsive enhancers in prostate cancer, finding many not detected by ChIP-seq or ATAC-seq.

The paper compares RNA sequencing assays and introduces a tool to identify active enhancers, compiling a compendium of enhancer candidates across 120 cell types.

The paper develops a computational framework to analyze mutational hotspots in cancer genomes using a human 3D interactome network, identifying clinically relevant interactions and proteins.