Biology subjects

Zhou, E.

Publications and source records attributed to Zhou, E..

17 recordsLinked to original sources

bioRxiv2026

LifeSciBench: Evaluating Language Models on Realistic, Expert-Level Tasks in the Life Sciences

We introduce LifeSciBench, a benchmark of 750 expert-authored tasks designed to evaluate whether language models can handle realistic life science research work. The majority of existing life sciences benchmarks have a narrow scope or are purely knowledge-based, and therefore fail to capture the complexity of real-world research, which often involves ambiguities and requires the accurate execution of multiple dependent judgment calls. Additionally, almost all existing benchmarks span at best a small collection of subdomains within the life sciences; there is at present no existing life sciences benchmark with both the requisite breadth and depth required to convincingly measure proficiency in real-world professional research settings. LifeSciBench addresses this gap by spanning seven representative scientific workflows and seven life science domains, with each constituent task paired with a human expert-written rubric. Across five frontier and domain-specialized models, GPT-Rosalind performs best, with a task-weighted mean normalized rubric score of 0.576 and a task-weighted response pass rate of 36.1% (response-level values are first averaged within each task, and the resulting task-level values are then averaged with equal weight). LifeSciBench remains unsaturated, with 171 tasks (22.8%) having no observed passing response from any evaluated model and 261 tasks (34.8%) having a best-model pass rate below 20%. LifeSciBench therefore serves as a high-resolution evaluation of practical scientific reasoning and operational decision-making in the life sciences.

bioinformatics↗

bioRxiv2026

A Germinal Center-Independent Innate-Like Memory B Cell Compartment of B1 Origin

Memory B cells (MBCs) are a critical cellular reservoir for long-term humoral immunity. MBCs display heterogeneous isotypes and surface markers and can arise through both germinal center (GC)-dependent and GC-independent extrafollicular (EF) pathways. Both the mechanisms controlling EF MBC differentiation and the identities of the B cell populations from which EF MBCs derive remain poorly understood. To capture MBC diversity, we applied a broad selection strategy followed by transcriptional profiling, identifying a subset of MBCs characterized by minimal class switching, limited somatic hypermutation, and an innate-like gene signature. Using genetic models, we demonstrated that this subset arises independently of GC responses and derives from innate B1 cells. These innate-like MBCs differentiate into antigen-specific antibody-secreting cells and confer protection against lethal viral infection. Together, our findings define a previously unrecognized arm of MBC responses, demonstrating that innate B1 cells contribute a non-redundant antibody repertoire to protective immunological memory.

immunology↗

bioRxiv2025

Mechanistically explainable AI model for predicting synergistic cancer therapy combinations

This study introduces a Large Language Model (LLM)-based framework that combines drug combination data with a knowledge graph to predict synergistic oncology drug combinations with mechanistic insights. Using a retrieval-augmented generation (RAG) approach, over 50,000 in vitro drug pair assay results and 1,631 human clinical trial or preclinical test entries were integrated to enhance predictive accuracy and explainability. Validation achieved an F1 score of 0.80, demonstrating the frameworks potential to streamline drug discovery and improve translational strategies in cancer treatment.

cancer biology↗

bioRxiv2025

High glucose levels impair the body's renal protective function during water deprivation by blunting the FXR-TonEBP axis

BackgroundDehydration often leads to kidney-related complications, yet its impact on diabetic patients remains underexplored. This study aimed to investigate the effects of dehydration on diabetic nephropathy(DN) and the underlying mechanisms of injury. MethodsWe analyzed clinical outcomes in 81 dehydration-induced acute kidney injury patients (35 with diabetes) and validated findings in streptozotocin-induced diabetic mice subjected to intermittent water deprivation (WD). Renal function, histopathology, and molecular markers (FXR, TonEBP, AQP2, gp91, apoptosis regulators) were assessed. ResultsKidney injury was aggravated after water deprivation in patients with diabetics and it is difficult to recover after water supplementation, with poor prognosis. Animal experiments have shown that compared with mice in a normal state, dehydration challenge aggravated renal damage in diabetic mice with increased serum creatinine, pathological damage, and renal interstitial cell apoptosis. Mechanistically, in a normal state FXR-TonEBP axis was enhanced to protect the kidney after water deprivation challenge, but this compensatory protective response was blunted in the diabetic state. ConclusionHyperglycemia blunts the FXR-TonEBP-osmolyte pathway during dehydration, compromising renal medullary protection. Targeting this axis may mitigate dehydration risks in diabetic nephropathy.

physiology↗

bioRxiv2025

Miniaturized and accessible functional ultrasound imaging system for freely moving mice

Functional ultrasound (fUS) imaging provides brain-wide activity maps with high spatiotemporal resolution and deep penetration, positioning it as a key technology for future non-invasive brain-computer interfaces (BCIs). Realizing this potential, particularly for chronic BCI applications requiring long-term monitoring in naturalistic settings, critically depends on significant system miniaturization to overcome the cost and complexity limitations of current platforms. Addressing this challenge, we present mini-fUS, a miniaturized, cost-effective fUS platform engineered for accessibility without compromising core performance for demanding neuroscience research. The system features a compact transmit-receiver, low-noise power supply, and high-speed data transfer, achieving pulse repetition frequencies up to 5 kHz with negligible jitter. Real-time GPU-accelerated beamforming and singular value decomposition (SVD) enable whole-brain activity mapping, demonstrated here in freely moving mice at up to 3.57 Hz with [~]100 {micro}m spatial resolution and 15 mm penetration depth. Validated through recordings of brain activity during sensory stimulation, anesthesia, and behavior, this design defines a practical hardware-software framework for fUS. By significantly improving accessibility and enabling robust monitoring in mobile subjects, this work advances the development of fUS for both fundamental research and future BCI technologies, while clarifying essential fUS operational principles.

neuroscience↗

bioRxiv2025

DARG: a Database of Alzheimer Related Genes in Model Organisms

Alzheimer disease is affecting a significant portion of the aging population worldwide. Despite extensive research over the years, a comprehensive understanding of its underlying mechanisms and effective treatments remains elusive. Meanwhile, animal models, such as mouse, zebrafish, Drosophila, and C. elegans, have proven invaluable in studying human diseases. In response, we have developed DARG, a Database of Alzheimer disease-related genes in model organisms, designed to bridge the gap between human geneticists investigating the molecular mechanisms of the disease and the model organisms that can be used to explore the functions of disease-associated genes. DARG allows users to search and browse Alzheimer-related human genes from various resources and datasets, identify orthologs in model organisms, and access data on their gene expression and related phenotypes in Drosophila.

bioinformatics↗

bioRxiv2024

P3 site-directed mutagenesis: An efficient method based on primer pairs with 3-prime overhangs

Site-directed mutagenesis is a fundamental tool indispensable for protein and plasmid engineering. An important technological question is how to achieve the efficiency at the ideal level of 100%. Based on complementary primer pairs, the QuickChange method has been widely used, but it requires significant improvements due to its low efficiency and frequent unwanted mutations. An alternative and innovative strategy is to utilize primer pairs with 3-prime overhangs, but this approach has not been fully developed. As the first step towards reaching the efficiency of 100%, we have optimized this approach systematically (such as use of newly designed short primers, test of different Pfu DNA polymerases and modification of PCR parameters) and evaluated the resulting method extensively with over 100 mutations on 12 mammalian expression vectors, ranging from 7.0-13.4 kb in size and encoding ten epigenetic regulators with links to cancer and neurodevelopmental disorders. We have also tested the new method with two expression vectors for the SARS-COV-2 spike protein. Compared to the QuickChange method, the success rate has increased substantially, with an average efficiency of [~]50%, with some at or close to 100 percent, and requiring much less time for engineering various mutations. Therefore, we have developed a new site-directed mutagenesis method for efficient and economical generation of various mutations. Notably, the method failed with a human KAT2B expression plasmid that possesses extremely GC-rich sequences. Thus, this study also sheds light on how to improve the method for developing ideal mutagenesis methods with the efficiency at [~]100% for a wide spectrum of plasmids.

molecular biology↗

bioRxiv2023

BioKG: a comprehensive, high-quality biomedical knowledge graph for AI-powered, data-driven biomedical research

To address the rapid growth of scientific publications and data in biomedical research, knowledge graphs (KGs) have become a critical tool for integrating large volumes of heterogeneous data to enable efficient information retrieval and automated knowledge discovery (AKD). However, transforming unstructured scientific literature into KGs remains a significant challenge, with previous methods unable to achieve human-level accuracy. In this study, we utilized an information extraction pipeline that won first place in the LitCoin NLP Challenge (2022) to construct a large-scale KG named iKraph using all PubMed abstracts. The extracted information matches human expert annotations and significantly exceeds the content of manually curated public databases. To enhance the KGs comprehensiveness, we integrated relation data from 40 public databases and relation information inferred from high-throughput genomics data. This KG facilitates rigorous performance evaluation of AKD, which was infeasible in previous studies. We designed an interpretable, probabilistic-based inference method to identify indirect causal relations and applied it to real-time COVID-19 drug repurposing from March 2020 to May 2023. Our method identified 600-1400 candidate drugs per month, with one-third of those discovered in the first two months later supported by clinical trials or PubMed publications. These outcomes are very challenging to attain through alternative approaches that lack a thorough understanding of the existing literature. A cloud-based platform (https://biokde.insilicom.com) was developed for academic users to access this rich structured data and associated tools.

bioinformatics↗

bioRxiv2022

ppGpp is Present in and Functions to Regulate Sleep in Drosophila

Sleep is essential for animals, and receives inputs from circadian, homeostasis, and environment, yet the mechanisms of sleep regulation remain elusive. Discovery of molecules in living systems and demonstration of their functional roles are pivotal in furthering our understanding of the molecular basis of biology. Here we report that ppGpp (guanosine-5-diphosphate, 3-diphosphate), a molecule that has been detected in prokaryotes for more than five decades, is present in Drosophila, and plays an important role in regulation of sleep and SISL (starvation induced sleep loss). ppGpp is detected in germ-free Drosophila and hydrolyzed by an enzyme encoded by the mesh1 gene in Drosophila. Nighttime sleep and SISL were defected in mesh1 mutant flies, and rescued by expression of wildtype Mesh1, but not the enzymatically defective mutant Mesh1E66A. Ectopic expression of RelA, the E. coli synthetase for ppGpp, phenocopied mesh1 knockout mutants, whereas overexpression of Mesh1 resulted in the opposite phenotypes, supporting that ppGpp is both necessary and sufficient in sleep regulation. A chemoconnectomic screen followed by genetic intersection experiments implicate the Dilp2 neurons in the pars intercerebralis (PI) brain region as the site of ppGpp function. Our results have thus supported that ppGpp is present in animals after long lag since its discovery in bacteria, and revealed a physiological role of ppGpp in sleep regulation for the first time.

genetics↗

bioRxiv2022

Systematic analysis of transcriptional and epigenetic effects of genetic variation in Kupffer cells enables discrimination of cell intrinsic and environment-dependent mechanisms

Noncoding genetic variation is a major driver of phenotypic diversity but determining the underlying mechanisms and the cell types in which it acts remain challenging problems. Here, we investigate the impact of natural genetic variation provided by phenotypically diverse inbred strains of mice on gene expression and epigenetic landscapes of Kupffer cells. Analysis of gene expression in Kupffer cells and other liver cell types derived from C57BL/6J, BALB/cJ and A/J mice provided evidence for strain-specific differences in environmental factors influencing Kupffer cell phenotypes, including preferential Leptin signaling in BALB/cJ Kupffer cells. Systematic analysis of transcriptomic and epigenetic data from F1 hybrids of these mice, and transcriptomic data from strain-specific Kupffer cells engrafted into a common host enabled quantitative assessment of cis versus trans effects of genetic variation on gene expression and an estimate of cell autonomous versus non cell autonomous effects. Under homeostatic conditions, trans effects of genetic variation were dominant, with the majority of trans regulation being non cell autonomous. In contrast, strain specific responses to acutely administered LPS were primarily associated with genetic variation acting in cis to modify response elements for lineage determining and signal dependent transcription factors. Collectively, these findings reveal cell intrinsic and environmental effects of natural genetic variation on gene expression, demonstrate the use of enhancers as detectors of trans effects of genetic variation, and provide a new resource for understanding the impact of genetic variation on gene expression in Kupffer cells.

genomics↗

bioRxiv2022

FGF21 protects against hepatic lipotoxicity and macrophage activation to attenuate fibrogenesis in nonalcoholic steatohepatitis

Analogues of the hepatokine FGF21 are in clinical development for type 2 diabetes and nonalcoholic steatohepatitis (NASH) treatment. Although their glucose-lowering and insulin-sensitizing effects have been largely unraveled, the mechanisms by which they alleviate liver injury have only been scarcely addressed. Here, we aimed to unveil the mechanisms underlying the protective effects of FGF21 on NASH using APOE*3-Leiden.CETP mice, a well-established model for human-like metabolic diseases. Liver-specific FGF21 overexpression was achieved in mice, followed by administration of a high-fat high-cholesterol diet for 23 weeks. FGF21 prevented hepatic lipotoxicity, accompanied by activation of thermogenic tissues and attenuation of adipose tissue inflammation, improvement of hyperglycemia and hypertriglyceridemia, and upregulation of hepatic programs involved in fatty acid oxidation and cholesterol removal. Furthermore, FGF21 inhibited hepatic inflammation, as evidenced by reduced Kupffer cell (KC) activation, diminished monocyte infiltration and lowered accumulation of monocyte-derived macrophages. Moreover, FGF21 decreased lipid- and scar-associated macrophages, which correlated with less hepatic fibrosis as demonstrated by reduced collagen accumulation. Collectively, hepatic FGF21 overexpression limits hepatic lipotoxicity, inflammation and fibrogenesis. Mechanistically, FGF21 blocks hepatic lipid influx and accumulation through combined endocrine and autocrine signaling, respectively, which prevents KC activation and lowers the presence of lipid- and scar-associated macrophages to inhibit fibrogenesis.

pathology↗

bioRxiv2022

Polycomb-lamina antagonism partitions heterochromatin at the nuclear periphery

The genome can be divided into two spatially segregated compartments, A and B,1,2 which broadly partition active and inactive chromatin states, respectively. Constitutive heterochromatin is predominantly located within the B compartment and comprises chromatin that is in close contact with the nuclear lamina.3-5 By contrast, facultative heterochromatin marked by H3K27me3 can span both compartments.2-5 How epigenetic modifications, A/B compartmentalization, and lamina association collectively maintain heterochromatin architecture and function remains unclear.6,7 Here we developed an approach termed Lamina-Inducible Methylation and Hi-C (LIMe-Hi-C) that jointly measures chromosome conformation, DNA methylation, and nuclear lamina positioning. Through this approach, we identified topologically distinct A/B sub-compartments characterized by high levels of H3K27me3 and differing degrees of lamina association. To study the regulation of these sub-compartments, we inhibited Polycomb repressive complex 2 (PRC2), revealing that H3K27me3 is an essential factor in sub-compartment segregation. Unexpectedly, PRC2 inhibition also elicited broad gains in lamina association and constitutive heterochromatin spreading into H3K27me3-marked B sub-compartment regions. Consistent with repositioning to the lamina, genes originally marked with H3K27me3 in the B compartment, but not in the A compartment, remained largely repressed, suggesting that constitutive heterochromatin spreading can compensate for loss of H3K27me3 at a transcriptional level. These findings demonstrate that Polycomb sub-compartments and their antagonism with nuclear lamina association are fundamental organizational features of genome structure. More broadly, by jointly measuring nuclear position and Hi-C contacts, our study demonstrates how dynamic changes in compartmentalization and nuclear lamina association represent distinct but interdependent modes of heterochromatin regulation.

genomics↗

bioRxiv2021

Sleep Need, the Key Regulator of Sleep Homeostasis, Is Indicated and Controlled by Phosphorylation of Threonine 221 in Salt Inducible Kinase 3

Sleep need drives sleep and plays a key role in homeostatic regulation of sleep. So far sleep need can only be inferred by animal behaviors and indicated by electroencephalography (EEG). Here we report that threonine 221 (T221) of the salt inducible kinase 3 (SIK3) was important for the catalytic activity and stability of SIK3. T221 phosphorylation in the mouse brain indicates sleep need: more sleep resulting in less phosphorylation and less sleep more phosphorylation during daily sleep/wake cycle and after sleep deprivation (SD). Sleep need was reduced in SIK3 loss of function (LOF) mutants and by T221 mutation to alanine (T221A). Sleep rebound after SD was also decreased in SIK3 LOF and T221A mutant mice. Other kinases such as SIK1 and SIK2 or other sites in SIK3 do not fulfil criteria to be both an indicator and a controller of sleep need. Our results reveal SIK3 T221 phosphorylation as the first and only chemical modification which indicates and controls sleep need.

neuroscience↗

bioRxiv2021

ppGpp Is Present in, and Regulates Sleep of, Drosophila

Discovery of molecules in living systems and demonstration of their functional roles are pivotal in furthering our understanding of the molecular basis of biology. ppGpp (guanosine-5-diphosphate, 3-diphosphate) has been detected in prokaryotes for more than five decades. Here we report that a genetic screen followed by chemical analysis revealed the presence of ppGpp in Drosophila. It can be detected in germ-free Drosophila and its level is controlled by an enzyme encoded by the mesh1 gene in Drosophila. Loss of function mutations in mesh1, which encoded the ppGpp degrading enzyme led to longer sleep latency and less total sleep. These phenotypes could be rescued by wild type mesh1, but not by the enzymatically defective mutant Mesh1E66A, functionally implicating ppGpp. Ectopic expression of RelA, the E. coli synthetase for ppGpp, phenocopied mesh1 knockout mutants, whereas overexpression of mesh1 resulted in the opposite phenotypes, supporting that ppGpp is both necessary and sufficient in sleep regulation. mesh1 is expressed in a specific population of neurons, and a chemoconnectomic screen followed by genetic intersection experiments implicate the pars intercerebralis (PI) as the site of ppGpp function. Our results have thus revealed that ppGpp is present in animals after long lag since its discovery in bacteria. Furthermore, we have demonstrated that ppGpp in a specific subset of neurons plays a physiological role in regulating sleep. We speculate that ppGpp may play function in mammals.

neuroscience↗

bioRxiv2019

A complex structure of arrestin-2 bound to a G protein-coupled receptor

Arrestins comprise a family of signal regulators of G-protein-coupled receptors (GPCRs), which include arrestins 1 to 4. While arrestins 1 and 4 are visual arrestins dedicated to rhodopsin, arrestins 2 and 3 (Arr2 and Arr3) are {beta}-arrestins known to regulate many nonvisual GPCRs. The dynamic and promiscuous coupling of Arr2 to nonvisual GPCRs has posed technical challenges to tackle the basis of arrestin binding to GPCRs. Here we report the structure of Arr2 in complex with neurotensin receptor 1 (NTSR1), which reveals an overall assembly that is strikingly different from the visual arrestin-rhodopsin complex by a 90{degrees} rotation of Arr2 relative to the receptor. In this new configuration, intracellular loop 3 (ICL3) and transmembrane helix 6 (TM6) of the receptor are oriented toward the N-terminal domain of the arrestin, making it possible for GPCRs that lack the C-terminal tail to couple Arr2 through their ICL3. Molecular dynamics simulation and crosslinking data further support the assembly of the Arr2-NTSR1 complex. Sequence analysis and homology modeling suggest that the Arr2-NTSR1 complex structure may provide an alternative template for modeling arrestin-GPCR interactions.

biophysics↗

bioRxiv2018

Decoys provide a scalable platform for the genetic analysis of plant E3 ubiquitin ligases that regulate circadian clock function

The circadian clock in all eukaryotes relies on the regulated degradation of clock proteins to maintain 24-hour rhythmicity. Despite this knowledge, we know very few of the components that mediate degradation of proteins to control clock function. This is likely due to high levels of gene duplication and functional redundancy within plant E3 ubiquitin ligase gene families. In order to overcome this issue and discover E3 ubiquitin ligases that control circadian clock function, we generated a library of transgenic Arabidopsis lines expressing dominant-negative \"decoy\" E3 ubiquitin ligases. We determined their effects on the plant circadian clock and identified dozens of new potential regulators of circadian clock function. To demonstrate the potency of the decoy screening methodology to overcome genetic redundancy and identify bona fide clock regulators, we performed follow-up studies on PUB59 and PUB60. Using knock-out studies, we show that they redundantly control circadian clock period by regulating gene splicing. Furthermore, we confirm that they are part of a conserved protein complex that mediates splicing in eukaryotes. This work demonstrates the viability of E3 ubiquitin ligase decoys as a scalable screening platform to overcome traditional genetic challenges and discover E3 ubiquitin ligases that regulate plant developmental processes.

plant biology↗

bioRxiv2018

GLaMST: Grow Lineages along Minimum Spanning Tree for B Cell Receptor Sequencing Data

B cell affinity maturation enables B cells to generate high-affinity antibodies. This process involves somatic hypermutation of B cell immunoglobulin receptor (BCR) genes and selection by their ability to bind antigens. Lineage trees are used to describe this microevolution of B cell immunoglobulin genes. In a lineage tree, each node is one BCR sequence that mutated from the germinal center and each directed edge represents a single base mutation, insertion or deletion. In BCR sequencing data, the observed data only contains a subset of BCR sequences in this microevolution process. Therefore, reconstructing the lineage tree from experimental data requires algorithms to build the tree based on partially observed tree nodes. We developed a new algorithm named Grow Lineages along Minimum Spanning Tree (GLaMST), which efficiently reconstruct the lineage tree given observed BCR sequences that correspond to a subset of the tree nodes. GLaMST constructs the minimum-spanning-tree (MST) to approximate the landscape of how observed BCR sequences are related, uses the MST to guide the interpolation of the closest unobserved sequence, updates the MST for the interpolation of additional unobserved sequences, and iterates until a full lineage tree is completed, where all observed sequences are connected by interpolated unobserved sequences and single base operations of mutations, insertions and deletions. Through comparison using simulated and real data, GLaMST outperforms existing algorithms in simulations with high rates of mutation, insertion and deletion, and generates lineage trees with smaller size and closer to ground truth according to tree features that highly correlated with selection pressure.

bioinformatics↗