Biology subjects

Zheng, W. J.

Publications and source records attributed to Zheng, W. J..

9 recordsLinked to original sources

bioRxiv2026

A Structural Antibody Benchmark of AlphaFold3 reveals Hallucinated Epitopes and a Bias for Orderness

AlphaFold3 has shown promise as a tool for predicting antibody-antigen binding, yet its performance across large datasets has not been fully characterized. In this study, 3401 experimentally validated antibody-antigen complexes were sourced from the Structural Antibody Database and screened alongside 23798 negative controls to benchmark AlphaFold3s binding prediction capabilities. Confidence metrics including Predicted Aligned Error and Interface Predicted Template Modeling score were used to achieving a maximum recall of 53% at 100 inference seeds. Several factors were found to influence prediction accuracy: a notable bias was observed toward antibodies derived from X-ray crystallography structures versus those from electron microscopy, and positive prediction rates were found to decrease with increasing target protein size and surface area. In contrast, neither the amino acid composition or lengths of the complementarity determining regions, nor training data leakage were found to introduce significant bias. An innate false positive rate of approximately 3% was identified, with AF3 shown to hallucinate plausible binding interfaces across the surface of decoy targets while avoiding disordered regions. Epitope mapping using DockQ, epitope shift, and antibody displacement revealed that approximately 34% of false negatives retained the correct epitope location despite poor structural alignment, suggesting that conformation refinement tools could recover additional true binding predictions. These findings provide a comprehensive characterization of AlphaFold3s strengths and limitations for antibody screening in computational drug discovery. Key MessagesO_LIAlphaFold3 has a recall of 50% and an innate false positive prediction rate of 3%. C_LIO_LIFalse negative predictions can still feature the correct epitope despite poor RMSD. C_LIO_LIFactors such as disorder and target size impact accuracy. C_LI

bioinformatics↗

bioRxiv2026

A layered standards framework for integrating single-cell and spatial omics data into brain cell atlases

The BRAIN Initiative Cell Atlas Network (BICAN) is generating large-scale multimodal datasets to profile cell types in the human, non-human primate, and mouse brain. The diversity of single-cell and spatial transcriptomic and epigenomic assays, combined with varied experimental contexts, multiple data-generating laboratories and distributed infrastructure, poses substantial challenges for data integration and reuse in BICAN. To address this, we implemented a standards framework that enables layered integration of these data into knowledge-ready products for interoperable brain cell atlases. This framework organizes data based on three progressively structured layers. First, we introduced an assay-agnostic modeling layer that unifies the representation of single-cell and spatial omics data using a common set of biological entities and processes assessed by diverse experimental techniques. Second, we implemented harmonized metadata standards that capture key experimental features linked to biospecimen provenance across heterogeneous tissue sources, species, and preparations, supporting integration and validation while minimizing burden on data contributors. Third, we present an extensible representation for data-driven cell type taxonomies that integrates molecular data with annotations, ontology mappings, and evidence. Together, these contributions represent an end-to-end framework that transforms heterogeneous datasets into structured, interoperable resources that support broad community reuse via mapping algorithms, annotation systems, and visualization platforms. This approach links biospecimen provenance with cell-level outputs and embeds these in a standardized taxonomy format, enabling downstream applications such as cross-dataset integration, reference mapping, and knowledge-driven analysis. More broadly, our work demonstrates a generalizable strategy for enabling an efficient data-to-knowledge pipeline in a large-scale consortium setting.

genomics↗

bioRxiv2026

eSIG-Net: Accurate prediction of single-mutation induced perturbations on protein interactions using a language model

Most proteins exert their functions in complex with other interactors. Single mutations can exhibit a profound impact on perturbing protein interactions, leading to human disease. However, predicting the effect of single mutations on protein interactions remains a major computational challenge. Deep learning, particularly protein language models or transformers, has become an effective tool in bioinformatics for protein structure prediction. However, the functional divergence of mutations makes it difficult to predict their interaction perturbation profiles. To address this fundamental challenge, we present eSIG-Net (edgetic mutation Sequence-based Interaction Grammar Network), a novel sequence-based "Interaction Language Model" for predicting protein interaction alterations caused by single mutations. eSIG-Net combines various protein sequence embeddings, introduces a mutation-encoding module with syntax and evolutionary insights, and employs contrastive learning to evaluate mutation-induced interaction changes. eSIG-Net significantly outperforms current state-of-the-art sequence-based and structure-based prediction methods at predicting mutational impact on protein interactions. We highlight examples where eSIG-Net nominates causal variants with high confidence and elucidates their functional role under relevant biological contexts. Together, eSIG-Net is a first-in-kind "interaction language model" that can accurately predict interaction-specific rewiring by single mutations with only sequence information, and exhibits generalizability across biological contexts.

bioinformatics↗

bioRxiv2024

Deletion of adipocyte Sine Oculis Homeobox Homolog 1 prevents lipolysis and attenuates skin fibrosis.

Dermal fibrosis is a cardinal feature of systemic sclerosis (SSc) for which there are limited treatment strategies. This is in part due to our fragmented understanding of how dermal white adipose tissue (DWAT) contributes to skin fibrosis. We identified elevated sine oculis homeobox homolog 1 (SIX1) expression in SSc skin samples from the GENISOS and PRESS cohorts, the expression of which correlated with adipose-associated genes and molecular pathways. SIX1 localization studies identified increased signals in the DWAT area in SSc and in experimental models of skin fibrosis. Global and adipocyte specific Six1 deletion abrogated end-stage fibrotic gene expression and dermal adipocyte shrinkage induced by SQ bleomycin treatment. Further studies revealed a link between elevated SIX1 and increased expression of SERPINE1 and its protein PAI-1 which are known pro-fibrotic mediators. However, SIX1 deletion did not appear to affect cellular trans differentiation. Taken together these results point at SIX1 as a potential target for dermal fibrosis in SSc. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSSkin thickening and tightening are leading causes of morbidity in systemic sclerosis (SSc). The authors previously reported that the aberrantly expressed developmental transcription factor sine oculis homeobox homology 1 (SIX1) drives pulmonary fibrosis. However, the contribution of SIX1 to skin fibrosis and associated dermal fat loss remains unknown. Added value of this studyThe role of dermal fat loss in skin fibrosis is not fully understood. Studies have shown that adipocytes can transition to mesenchymal cells promoting fibrosis, consistent with loss of the dermal white adipose layer. Our research provides insight into a novel molecular mechanism of lipodystrophy important for skin fibrosis in SSc. We identified the upregulation of SIX1 in adipocytes in skin from patients with SSc which was associated with the progression of skin fibrosis. We found elevated Six1 in mouse dermal adipocytes of early fibrotic skin. Ubiquitous and adipose-specific loss of Six1 decreased markers of experimental skin fibrosis in mice which recapitulate cardinal features of SSc dermal fibrosis. Increased SIX1 expression is linked with elevated levels of Serpine1 the gene that codes for the protein plasminogen activator inhibitor (PAI)-1. This is important since PAI-1 is a known pro-fibrotic agent in the skin that contributes to the deposition of extracellular matrix (ECM) products. Implications of all the available evidenceFat atrophy may represent a targetable contributor to early systemic sclerosis manifestations. This is as it precedes skin fibrosis and the use of topical agent which are usually lipophilic can help us target dermal adipocytes. Our results show that SIX1 could be an important early marker for skin fibrosis in SSc that can also be targeted therapeutically.

molecular biology↗

bioRxiv2024

Systematic characterization of protein structural features of alternative splicing isoforms using AlphaFold 2

Alternative splicing is an important cellular process in eukaryotes, altering pre-mRNA to yield multiple protein isoforms from a single gene. However, our understanding of the impact of alternative splicing events on protein structures is currently constrained by a lack of sufficient protein structural data. To address this limitation, we employed AlphaFold 2, a cutting-edge protein structure prediction tool, to conduct a comprehensive analysis of alternative splicing for approximately 3,000 human genes, providing valuable insights into its impact on the protein structural. Our investigation employed state of the art high-performance computing infrastructure to systematically characterize structural features in alternatively spliced regions and identified changes in protein structure following alternative splicing events. Notably, we found that alternative splicing tends to alter the structure of residues primarily located in coils and beta-sheets. Our research highlighted a significant enrichment of loops and highly exposed residues within human alternatively spliced regions. Specifically, our examination of the Septin-9 protein revealed potential associations between loops and alternative splicing, providing insights into its evolutionary role. Furthermore, our analysis uncovered two missense mutations in the Tau protein that could influence alternative splicing, potentially contributing to the pathogenesis of Alzheimers disease. In summary, our work, through a thorough statistical analysis of extensive protein structural data, sheds new light on the intricate relationship between alternative splicing, evolution, and human disease.

bioinformatics↗

bioRxiv2022

Sequence-to-expression approach to identify etiological non-coding DNA variations in P53 and cMYC-driven diseases

Disease risk prediction based on DNA sequence and transcriptional profile can improve disease screening, prevention, and potential therapeutic approaches by revealing contributing genetic factors and altered networks. Despite identifying many disease-associated DNA variants through genome-wide association studies, distinguishing deleterious non-coding DNA variations remains poor for most common diseases. We previously reported that non-coding variations disrupting cis-overlapping motifs (CisOMs) of opposing transcription factors significantly affect enhancer activity. Analyzing publicly available ChIP-seq data for P53 and cMYC in human embryonic stem cells and mouse embryonic cells showed that [~]344-366 genomic regions are co-occupied by P53 and cMYC. We identified, on average, two CisOMs per region, suggesting that co-occupancy is evolutionarily conserved in vertebrates. Therefore, we designed in vitro experiments to uncover the significance of the co-occupancy and competitive binding and inhibition between P53 and cMYC on target gene expression. We found that treating U2OS cells with doxorubicin increased P53 protein level while reducing cMYC level. In contrast, no change in protein levels was observed in Raji cells. ChIP-seq analysis showed that 16-922 genomic regions were co-occupied by P53 and cMYC before and after treatment, and substitutions of cMYC signals by P53 were detected after doxorubicin treatment in U2OS. Around 187 expressed genes near co-occupied regions were altered at mRNA level according to RNA-seq data. We utilized a computational motif-matching approach to determine that changes in predicted P53 binding affinity by DNA variations in CisOMs of co-occupied elements significantly correlate with alterations in reporter gene expression. We performed a similar analysis using SNPs mapped in CisOMs for P53 and cMYC from ChIP-seq data in U2OS and Raji, and expression of target genes from the GTEx portal. We found a significant correlation between change in motif-predicted cMYC binding affinity by SNPs in CisOMs and gene expression. In conclusion, our study suggests a generally applicable approach to filter etiological non-coding variations associated with P53 and cMYC-dependent diseases. Author SummaryMost DNA variants associated with common complex diseases fall outside the protein-coding regions of the genome, making them hard to detect and relate to a function. Although many computational tools are available for prioritizing functional disease risk variants outside the protein-coding regions of the genome, the precision of prediction of these tools is mostly unreliable and hence not close to cancer risk prediction. This study brings to light a novel way to improve prediction accuracy of publicly available tools by integrating the impact of cis-overlapping binding sites of opposing cancer proteins, such as P53 and cMYC, in their analysis to filter out deleterious DNA variants outside the protein-coding regions of the human genome. Using a biology-based statistical approach, DNA variants within cis-overlapping motifs impacting the binding affinity of opposing transcription factors can significantly alter the expression of target genes and regulatory networks. This study brings us closer to developing a generally applicable approach capable of filtering etiological non-coding variations in co-occupied genomic regions of P53 and cMYC family members to improve disease risk assessment.

genetics↗

bioRxiv2022

Unifying Gene Expression Complexity and Cellular Operational Efficiency: Application of the Locality and Caching Principles via a Cell-to-Computer Analogy

Gene expression is time-consuming and sequentially more so from bacteria to yeast to human, rendering human cells vulnerable to proteomic-response and operational latency. Computers once suffered such latency, imposed by much-slower information retrieval (hard-drive (HD) to memory to central-processing-unit (CPU)) than CPU execution. Optimization principles, namely, spatiotemporal-locality-principles that control specialized user-programs and caching that controls operating system (OS) kernel (the HD-CPU information flow channel), successfully mitigated the latency by gearing the memory towards near-future or high-priority CPU needs. We report evidence that the principles similarly act in cellular latency-mitigation via analogizing genome-mRNA-protein gene-expression to HD-memory-CPU information-retrieval, and transcriptome to memory. First, temporal-locality-principle is equivalent to mRNA stabilization-by-translation regulation and controls specialized cellular functions. Second, caching is equivalent to cytoplasmic mRNA sequestration. Highly sequestered mRNAs defy the locality-principle. In both cells and computers, caching controls the information channels; gene expression machinery and their regulators, i.e., the cellular channel (OS-kernel equivalent) that regulates arguably all cellular processes, are top sequestered mRNAs. Third, mRNA-caching contributes to the mRNA-protein expression discrepancy. Thus, locality and caching principles control specialized and core cellular functions, respectively, orchestrating transcriptome regulation and bridging it to cellular operational efficiency.

systems biology↗

bioRxiv2021

HSC-independent definitive hematopoietic cells persist into adult life

The stem cell theory that all blood cells are derived from hematopoietic stem cell (HSC) is a central dogma in hematology. However, various types of blood cells are already produced from hemogenic endothelial cells (HECs) before the first HSCs appear at embryonic day (E)11 in the mouse embryo. This early blood cell production from HECs, called HSC-independent hematopoiesis, includes primitive and definitive erythromyeloid progenitors that transiently support fetal blood homeostasis until HSC-derived hematopoiesis is established. Lymphoid potential has traditionally been detected in the extra-embryonic yolk sac (YS) and/or embryos before HSC emergence, but the actual presence of lymphoid progenitors at this stage remains unknown. In addition, whether HSCs in the fetal liver are the main source of innate-like B-1a cells has been controversial. Here, using complementary lineage tracing mouse models, we show that HSC-independent multipotent progenitors (MPPs) and HSC-independent adoptive B-lymphoid progenitors persist into adult life. Furthermore, HSCs minimally contribute to the peritoneal B-1a cell pool; most B-1a cells are originated directly from ECs in the YS and embryo and HSC-independent for life. Our discovery of extensive HSC-independent MPP and B-lymphoid progenitors in adults attests to the complex blood developmental dynamics through embryo to adult that underpin the immune system and challenges the paradigm of HSC theory in hematology.

developmental biology↗

bioRxiv2017

A Multi-Parameter Analysis of Cellular Coordination of Multiple Transcriptome Regulation Mechanisms

To understand cellular coordination of multiple transcriptome regulation mechanisms, we simultaneously measured transcription rate (TR), mRNA abundance (RA) and translation activity (TA). This revealed multiple quantitative insights. First, the genomic profiles of the three parameters are systematically different in key statistical features. Sequentially more genes exhibit extreme low or high expression values from TR to RA, then to TA. That is, because of cellular coordination of these regulatory mechanisms, sequentially higher levels of gene expression selectivity are achieved as genetic information flow from the genome to the proteome. Second, the contribution of the stabilization-by-translation regulatory mechanism to the cellular coordination process was assessed. The data enabled an estimation of mRNA stability, revealing a moderate but significant positive correlation between the estimated mRNA stability and translation activity. Third, the proportion of a mRNA occupied by un-translated regions (UTR) exhibits a negative relationship with the level of this correlation, and is thus a major determinant of the mode of regulation of the mRNA. High-UTR-proportion mRNAs tend to defy the stabilization-by-translation regulatory mechanism, staying out of the polysome but remaining stable; mRNAs with little UTRs largely follow this regulation. In summary, we quantitatively delineated the relationship among multiple transcriptome regulation parameters, i.e., cellular coordination of corresponding regulatory mechanisms.

systems biology↗