Ziheng Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ziheng Chen has a total of 84 co-authors. Here are the top co-authors by number of shared publications:
- Andreas R. Pfenning
Carnegie Mellon University
4 shared publications - BaDoi Nguyen Phan
Department of Computational Biology, School of Computer Science, Carnegie Mellon University
4 shared publications - Easwaran Ramamurthy
Carnegie Mellon University
3 shared publications - Heather H. Sestili
Carnegie Mellon University
3 shared publications Aditi Trehan
Cardiometabolic Genomics Program, Division of Cardiology, Department of Medicine, Columbia University Irving Medical Center, New York, USA
2 shared publications- Alexis O. Aparicio
Department of Medicine, Department of Systems Biology, Columbia University Medical Center, New York, USA
2 shared publications
Latest Publications
- Open Access
Context-dependent regulatory variants in Alzheimer’s disease
Noncoding genetic variants underlie many complex diseases, yet identifying and interpreting their functional impacts remains challenging. Late-onset Alzheimer’s disease (LOAD), a polygenic neurodegenerative disorder, exemplifies this challenge. The disease is strongly associated with noncoding variation, including common variants enriched in microglial enhancers and rare variants that are hypothesized to influence neurodevelopment and synaptic plasticity. These variants often perturb regulatory sequences by disrupting transcription factor (TF) motifs or altering local TF interactions, thereby reshaping gene expression and chromatin accessibility. However, assessing their impact is complicated by the context-dependent functions of regulatory sequences, underscoring the need to systematically examine variant effects across diverse tissues, cell types, and cellular states. Here, we combined in vitro and in vivo massively parallel reporter assays (MPRAs) with interpretable machine-learning models to systematically characterize common and rare variants across myeloid and neural contexts. Parallel profiling of variants in four immune states in vitro and three mouse brain regions in vivo revealed that individual variants can differentially and even oppositely modulate regulatory function depending on cell-type and cell-state contexts. Common variants associated with LOAD tended to exert stronger effects in immune contexts, whereas rare variants showed more pronounced impacts in brain contexts. Interpretable sequence-to-function deep-learning models elucidated how genetic variation leads to cell-type-specific differences in regulatory activity, pinpointing both direct transcription-factor motif disruptions and subtler tuning of motif context. To probe the broader functional consequences of a locus prioritized by our reporter assays and models, we used CRISPR interference to silence an enhancer within the SEC63-OSTM1 locus that harbors four functional rare variants, revealing its gatekeeper role in inflammation and amyloidogenesis. These findings underscore the context-dependent nature of noncoding variant effects in LOAD and provide a generalizable framework for the mechanistic interpretation of risk alleles in complex diseases.
Open AccessContext-dependent regulatory variants in Alzheimer’s disease
Noncoding genetic variants underlie many complex diseases, yet identifying and interpreting their functional impacts remains challenging. Late-onset Alzheimer’s disease (LOAD), a polygenic neurodegenerative disorder, exemplifies this challenge. The disease is strongly associated with noncoding variation, including common variants enriched in microglial enhancers and rare variants that are hypothesized to influence neurodevelopment and synaptic plasticity. These variants often perturb regulatory sequences by disrupting transcription factor (TF) motifs or altering local TF interactions, thereby reshaping gene expression and chromatin accessibility. However, assessing their impact is complicated by the context-dependent functions of regulatory sequences, underscoring the need to systematically examine variant effects across diverse tissues, cell types, and cellular states. Here, we combined in vitro and in vivo massively parallel reporter assays (MPRAs) with interpretable machine-learning models to systematically characterize common and rare variants across myeloid and neural contexts. Parallel profiling of variants in four immune states in vitro and three mouse brain regions in vivo revealed that individual variants can differentially and even oppositely modulate regulatory function depending on cell-type and cell-state contexts. Common variants associated with LOAD tended to exert stronger effects in immune contexts, whereas rare variants showed more pronounced impacts in brain contexts. Interpretable sequence-to-function deep-learning models elucidated how genetic variation leads to cell-type-specific differences in regulatory activity, pinpointing both direct transcription-factor motif disruptions and subtler tuning of motif context. To probe the broader functional consequences of a locus prioritized by our reporter assays and models, we used CRISPR interference to silence an enhancer within the SEC63-OSTM1 locus that harbors four functional rare variants, revealing its gatekeeper role in inflammation and amyloidogenesis. These findings underscore the context-dependent nature of noncoding variant effects in LOAD and provide a generalizable framework for the mechanistic interpretation of risk alleles in complex diseases.
- 25 June 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Translational Medicine
- Vol. 17 (804) , eadl5623
A cationic polymer drives glycosaminoglycan assembly and secretion for preclinical osteoarthritis therapy
Osteoarthritis (OA) affects nearly 500 million people worldwide and is characterized by an irreversible loss of glycosaminoglycans (GAGs) at articular cartilage surfaces, which are essential in maintaining cartilage mechanical properties and chondrocyte phenotypes. Despite advances, preserving cartilage GAGs and controlling their turnover in living cells remain challenging. On the basis of the hypothesis that GAGs can interact with cationic molecules, we demonstrated a cost-effective strategy to increase human cartilage GAGs using a cationic polymer hexadimethrine bromide (HDMBr). HDMBr promoted chondrogenesis of mesenchymal stem cells by attracting pericellular GAGs and up-regulating vesicle formation, leading to increased matrix secretion. HDMBr also acted like a molecular assembler to promote the assembly of chondroitin sulfate (CS) into highly concentrated condensates during intracellular trafficking, resulting in more efficient GAG secretion. HDMBr was then evaluated as a potential therapeutic in two animal models. In a rabbit model of large cartilage defects, HDMBr promoted the intrinsic regeneration of GAG-rich hyaline-like cartilage and improved tissue integration. In a rat model of OA, low-dose HDMBr treatment increased cartilage thickness, supported cartilage matrix homeostasis, and supported cell-based therapy, reducing OA damage as compared with other tested clinical treatments. Overall, this study introduces a cost-effective GAG manipulation approach to cartilage repair and joint preservation, offering insights into the mechanisms of cell-material interactions.
- 25 June 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Translational Medicine
- Vol. 17 (804) , eadl5623
A cationic polymer drives glycosaminoglycan assembly and secretion for preclinical osteoarthritis therapy
Osteoarthritis (OA) affects nearly 500 million people worldwide and is characterized by an irreversible loss of glycosaminoglycans (GAGs) at articular cartilage surfaces, which are essential in maintaining cartilage mechanical properties and chondrocyte phenotypes. Despite advances, preserving cartilage GAGs and controlling their turnover in living cells remain challenging. On the basis of the hypothesis that GAGs can interact with cationic molecules, we demonstrated a cost-effective strategy to increase human cartilage GAGs using a cationic polymer hexadimethrine bromide (HDMBr). HDMBr promoted chondrogenesis of mesenchymal stem cells by attracting pericellular GAGs and up-regulating vesicle formation, leading to increased matrix secretion. HDMBr also acted like a molecular assembler to promote the assembly of chondroitin sulfate (CS) into highly concentrated condensates during intracellular trafficking, resulting in more efficient GAG secretion. HDMBr was then evaluated as a potential therapeutic in two animal models. In a rabbit model of large cartilage defects, HDMBr promoted the intrinsic regeneration of GAG-rich hyaline-like cartilage and improved tissue integration. In a rat model of OA, low-dose HDMBr treatment increased cartilage thickness, supported cartilage matrix homeostasis, and supported cell-based therapy, reducing OA damage as compared with other tested clinical treatments. Overall, this study introduces a cost-effective GAG manipulation approach to cartilage repair and joint preservation, offering insights into the mechanisms of cell-material interactions.
- Open Access
Combining Machine Learning and Multiplexed, In Situ Profiling to Engineer Cell Type and Behavioral Specificity
Summary: A promising strategy for the precise control of neural circuits is to use cis -regulatory enhancers to drive transgene expression in specific cells. However, enhancer discovery faces key challenges: low in vivo success rates, species-specific differences in activity, challenges with multiplexing adeno-associated viruses (AAVs), and the lack of spatial detail from single-cell sequencing. In order to accelerate enhancer discovery for the dorsal spinal cord—a region critical for pain and itch processing—we developed an end-to-end platform, ESCargoT ( Engineered Specificity of Cargo Transcription ), combining machine learning (ML)–guided enhancer prioritization, modular AAV assembly, and multiplexed, in situ screening. Using cross-species chromatin accessibility data, we trained ML models to predict enhancer activity in oligodendrocytes and in 15 dorsal horn neuronal subtypes. We first demonstrated that an initial enhancer, Excit-1, targeted excitatory dorsal horn neurons and drove reversal of mechanical allodynia in an inflammatory pain model. To enable parallel profiling of a 27-enhancer-AAV library delivered intraspinally in mice, we developed a Spatial Parallel Reporter Assay (SPRA) by integrating a novel Golden-Gate assembly pipeline with multiplexed, in situ screening. Regression adjustment for spatial confounding enabled specificity comparisons between enhancers, demonstrating the ability to screen enhancers targeting diverse cell types (oligodendrocytes, motoneurons, dorsal neuron subtypes) in one experiment. We then validated two candidates, targeting Exc-LMO3 and Exc-SKOR2 neurons, respectively. In a companion paper by Noh et al , our colleagues show that the functional specificity of the Exc-SKOR2-targeting enhancer, unlike Excit-1, is capable of blocking the sensation of chemical itch in mice. These enhancers were derived from the macaque genome but displayed functional sensitivity in mice. This platform enables spatially resolved, multiplexed in vivo enhancer profiling to accelerate discovery of cell-targeting tools and gene therapy development.
Open AccessCombining Machine Learning and Multiplexed, In Situ Profiling to Engineer Cell Type and Behavioral Specificity
Summary: A promising strategy for the precise control of neural circuits is to use cis -regulatory enhancers to drive transgene expression in specific cells. However, enhancer discovery faces key challenges: low in vivo success rates, species-specific differences in activity, challenges with multiplexing adeno-associated viruses (AAVs), and the lack of spatial detail from single-cell sequencing. In order to accelerate enhancer discovery for the dorsal spinal cord—a region critical for pain and itch processing—we developed an end-to-end platform, ESCargoT ( Engineered Specificity of Cargo Transcription ), combining machine learning (ML)–guided enhancer prioritization, modular AAV assembly, and multiplexed, in situ screening. Using cross-species chromatin accessibility data, we trained ML models to predict enhancer activity in oligodendrocytes and in 15 dorsal horn neuronal subtypes. We first demonstrated that an initial enhancer, Excit-1, targeted excitatory dorsal horn neurons and drove reversal of mechanical allodynia in an inflammatory pain model. To enable parallel profiling of a 27-enhancer-AAV library delivered intraspinally in mice, we developed a Spatial Parallel Reporter Assay (SPRA) by integrating a novel Golden-Gate assembly pipeline with multiplexed, in situ screening. Regression adjustment for spatial confounding enabled specificity comparisons between enhancers, demonstrating the ability to screen enhancers targeting diverse cell types (oligodendrocytes, motoneurons, dorsal neuron subtypes) in one experiment. We then validated two candidates, targeting Exc-LMO3 and Exc-SKOR2 neurons, respectively. In a companion paper by Noh et al , our colleagues show that the functional specificity of the Exc-SKOR2-targeting enhancer, unlike Excit-1, is capable of blocking the sensation of chemical itch in mice. These enhancers were derived from the macaque genome but displayed functional sensitivity in mice. This platform enables spatially resolved, multiplexed in vivo enhancer profiling to accelerate discovery of cell-targeting tools and gene therapy development.
- Open Access
- 26 August 2024
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 20 (8) , e1012356
Regression convolutional neural network models implicate peripheral immune regulatory variants in the predisposition to Alzheimer’s disease
Alzheimer’s disease (AD) involves aggregation of amyloid β and tau, neuron loss, cognitive decline, and neuroinflammatory responses. Both resident microglia and peripheral immune cells have been associated with the immune component of AD. However, the relative contribution of resident and peripheral immune cell types to AD predisposition has not been thoroughly explored due to their similarity in gene expression and function. To study the effects of AD-associated variants on cis-regulatory elements, we train convolutional neural network (CNN) regression models that link genome sequence to cell type-specific levels of open chromatin, a proxy for regulatory element activity. We then use in silico mutagenesis of regulatory sequences to predict the relative impact of candidate variants across these cell types. We develop and apply criteria for evaluating our models and refine our models using massively parallel reporter assay (MPRA) data. Our models identify multiple AD-associated variants with a greater predicted impact in peripheral cells relative to microglia or neurons. Our results support their use as models to study the effects of AD-associated variants and even suggest that peripheral immune cells themselves may mediate a component of AD predisposition. We make our library of CNN models and predictions available as a resource for the community to study immune and neurological disorders.
- 26 August 2024
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 20 (8) , e1012356
Open AccessRegression convolutional neural network models implicate peripheral immune regulatory variants in the predisposition to Alzheimer’s disease
Alzheimer’s disease (AD) involves aggregation of amyloid β and tau, neuron loss, cognitive decline, and neuroinflammatory responses. Both resident microglia and peripheral immune cells have been associated with the immune component of AD. However, the relative contribution of resident and peripheral immune cell types to AD predisposition has not been thoroughly explored due to their similarity in gene expression and function. To study the effects of AD-associated variants on cis-regulatory elements, we train convolutional neural network (CNN) regression models that link genome sequence to cell type-specific levels of open chromatin, a proxy for regulatory element activity. We then use in silico mutagenesis of regulatory sequences to predict the relative impact of candidate variants across these cell types. We develop and apply criteria for evaluating our models and refine our models using massively parallel reporter assay (MPRA) data. Our models identify multiple AD-associated variants with a greater predicted impact in peripheral cells relative to microglia or neurons. Our results support their use as models to study the effects of AD-associated variants and even suggest that peripheral immune cells themselves may mediate a component of AD predisposition. We make our library of CNN models and predictions available as a resource for the community to study immune and neurological disorders.
- Top Cited Papers
- 12 April 2023
- Published by Springer Nature in Nature
- Vol. 617 (7960) , 395-402
Noncoding translation mitigation
Translation is pervasive outside of canonical coding regions, occurring in long noncoding RNAs, canonical untranslated regions and introns1,2,3,4, especially in ageing4,5,6, neurodegeneration5,7 and cancer8,9,10. Notably, the majority of tumour-specific antigens are results of noncoding translation11,12,13. Although the resulting polypeptides are often nonfunctional, translation of noncoding regions is nonetheless necessary for the birth of new coding sequences14,15. The mechanisms underlying the surveillance of translation in diverse noncoding regions and how escaped polypeptides evolve new functions remain unclear10,16,17,18,19. Functional polypeptides derived from annotated noncoding sequences often localize to membranes20,21. Here we integrate massively parallel analyses of more than 10,000 human genomic sequences and millions of random sequences with genome-wide CRISPR screens, accompanied by in-depth genetic and biochemical characterizations. Our results show that the intrinsic nucleotide bias in the noncoding genome and in the genetic code frequently results in polypeptides with a hydrophobic C-terminal tail, which is captured by the ribosome-associated BAG6 membrane protein triage complex for either proteasomal degradation or membrane targeting. By contrast, canonical proteins have evolved to deplete C-terminal hydrophobic residues. Our results reveal a fail-safe mechanism for the surveillance of unwanted translation from diverse noncoding regions and suggest a possible biochemical route for the preferential membrane localization of newly evolved proteins.
- 12 April 2023
- Published by Springer Nature in Nature
- Vol. 617 (7960) , 395-402
Top Cited PapersNoncoding translation mitigation
Translation is pervasive outside of canonical coding regions, occurring in long noncoding RNAs, canonical untranslated regions and introns1,2,3,4, especially in ageing4,5,6, neurodegeneration5,7 and cancer8,9,10. Notably, the majority of tumour-specific antigens are results of noncoding translation11,12,13. Although the resulting polypeptides are often nonfunctional, translation of noncoding regions is nonetheless necessary for the birth of new coding sequences14,15. The mechanisms underlying the surveillance of translation in diverse noncoding regions and how escaped polypeptides evolve new functions remain unclear10,16,17,18,19. Functional polypeptides derived from annotated noncoding sequences often localize to membranes20,21. Here we integrate massively parallel analyses of more than 10,000 human genomic sequences and millions of random sequences with genome-wide CRISPR screens, accompanied by in-depth genetic and biochemical characterizations. Our results show that the intrinsic nucleotide bias in the noncoding genome and in the genetic code frequently results in polypeptides with a hydrophobic C-terminal tail, which is captured by the ribosome-associated BAG6 membrane protein triage complex for either proteasomal degradation or membrane targeting. By contrast, canonical proteins have evolved to deplete C-terminal hydrophobic residues. Our results reveal a fail-safe mechanism for the surveillance of unwanted translation from diverse noncoding regions and suggest a possible biochemical route for the preferential membrane localization of newly evolved proteins.
- Open Access
- 28 March 2023
- Published by Springer Nature in Communications Biology
- Vol. 6 (1) , 1-8
Collateral activity of the CRISPR/RfxCas13d system in human cells
CRISPR/Cas13 systems are increasingly used for programmable targeting of RNAs. While Cas13 nucleases are capable of degrading both target RNAs and bystander RNAs in vitro and in bacteria, initial studies fail to detect collateral degradation of non-target RNAs in eukaryotic cells. Here we show that RfxCas13d, also known as CasRx, a widely used Cas13 system, can cause collateral transcriptome destruction when targeting abundant reporter RNA and endogenous RNAs, resulting in proliferation defect in target cells. While these results call for caution of using RfxCas13d for targeted RNA knockdown, we demonstrated that the collateral activity can be harnessed for selective depletion of a specific cell population defined by a marker RNA in an in vitro setting.
- 28 March 2023
- Published by Springer Nature in Communications Biology
- Vol. 6 (1) , 1-8
Open AccessCollateral activity of the CRISPR/RfxCas13d system in human cells
CRISPR/Cas13 systems are increasingly used for programmable targeting of RNAs. While Cas13 nucleases are capable of degrading both target RNAs and bystander RNAs in vitro and in bacteria, initial studies fail to detect collateral degradation of non-target RNAs in eukaryotic cells. Here we show that RfxCas13d, also known as CasRx, a widely used Cas13 system, can cause collateral transcriptome destruction when targeting abundant reporter RNA and endogenous RNAs, resulting in proliferation defect in target cells. While these results call for caution of using RfxCas13d for targeted RNA knockdown, we demonstrated that the collateral activity can be harnessed for selective depletion of a specific cell population defined by a marker RNA in an in vitro setting.
- Open Access
An in vivo massively parallel platform for deciphering tissue-specific regulatory function
Genetic studies are rapidly identifying non-protein-coding human disease-associated loci. Understanding the regulatory mechanisms underlying these loci remains a challenge because the causal variants and the tissues in which they act are often unclear. Massively parallel reporter assays (MPRAs) have the potential to link differences in genome sequence, including genetic variants, to tissue-specific regulatory function. Although MPRA and similar technologies have been widely adopted in cell culture, there have been several barriers to widespread use in animals. We overcome these challenges with a new whole-animal MPRA (WhAMPRA), where systemic intravenous AAV effectively transduces the plasmid MPRA library to mouse tissues. Our WhAMPRA approach revealed models of tissue-specific regulation that generally match machine learning model predictions. In addition, we measured the regulatory effects of disrupting MEF2C transcription factor binding sites and impacts of late onset Alzheimer’s disease-associated genetic variations. Overall, our WhAMPRA technology simultaneously determines the transcriptional functions of hundreds of enhancers in vivo across multiple tissues.
Open AccessAn in vivo massively parallel platform for deciphering tissue-specific regulatory function
Genetic studies are rapidly identifying non-protein-coding human disease-associated loci. Understanding the regulatory mechanisms underlying these loci remains a challenge because the causal variants and the tissues in which they act are often unclear. Massively parallel reporter assays (MPRAs) have the potential to link differences in genome sequence, including genetic variants, to tissue-specific regulatory function. Although MPRA and similar technologies have been widely adopted in cell culture, there have been several barriers to widespread use in animals. We overcome these challenges with a new whole-animal MPRA (WhAMPRA), where systemic intravenous AAV effectively transduces the plasmid MPRA library to mouse tissues. Our WhAMPRA approach revealed models of tissue-specific regulation that generally match machine learning model predictions. In addition, we measured the regulatory effects of disrupting MEF2C transcription factor binding sites and impacts of late onset Alzheimer’s disease-associated genetic variations. Overall, our WhAMPRA technology simultaneously determines the transcriptional functions of hundreds of enhancers in vivo across multiple tissues.