Biology subjects

Huang, J. Z.

Publications and source records attributed to Huang, J. Z..

9 recordsLinked to original sources

bioRxiv2026

Small molecule structure elucidation from NMR spectra with generative and reasoning models

Elucidating the structures of unknown small molecules remains a major challenge in chemistry, pharmacology, and drug discovery. While nuclear magnetic resonance (NMR) spectroscopy is the gold standard for structure elucidation, interpreting spectra remains labor-intensive and requires expert knowledge to navigate the vast space of possible structures. Here, we present an AI system for rapid, automated de novo structure elucidation that combines a molecular generative model with a scientific reasoning agent. Its generative component, ChefNMR2, is an atomic flow-matching model that predicts small-molecule structures from molecular formulae and one-dimensional 1H and 13C NMR spectra. ChefNMR2-Agent augments our generative model with large language models (LLMs) to reason over and refine candidate structures. Across diverse benchmarks, ChefNMR2 substantially outperforms existing tools, offering calibrated confidence estimates and two to three orders of magnitude faster inference than frontier LLMs. We use ChefNMR2-Agent to elucidate the structures of four novel natural products, matching human-chemist accuracy in a fraction of the time. Our work demonstrates how specialized generative models and general-purpose reasoning agents can unite to solve challenging scientific inverse problems and accelerate NMR structure elucidation, thus opening new possibilities for AI-driven small-molecule discovery across diverse disciplines.

bioinformatics↗

bioRxiv2026

Caenorhabditis elegans avoids Todstoff, a novel nociceptive necrotaxis cue

To survive in hazardous environments, organisms must navigate and avoid myriad threats. Many species across phyla detect warning cues in the remains of dead or injured conspecifics, allowing them to enact defensive strategies and avoid a similar fate. Here, we demonstrate that Caenorhabditis elegans senses a novel aversive cue (which we term Todstoff, or death substance) present in the remains of dead worms that induces living conspecifics to perform negative necrotaxis behaviors. Todstoff is distinct from previously-identified social signals in C. elegans, including ascaroside and alarm pheromones, and our biochemical analysis has revealed that Todstoff may be a protein-associated amino acid derivative less than 1.35 kDa in size. We determined that Todstoff is sensed by the ASH polymodal nociceptive neurons, and further, that the necrotaxis signal is transduced via the activity of G-protein coupled receptor (GPCR) signaling, the Gi/o-like protein ODR-3, TRPV channels OSM-9 and OCR-2, and glutamatergic synaptic transmission to downstream AIB interneurons. Taken together, our work illuminates a post-mortem inter-animal chemical signaling pathway that promotes death avoidance, and thus, survival.

genetics↗

bioRxiv2026

Discovery of a Human Metabolite that Mimics the Bacterial Quorum-Sensing Autoinducer AI-2

Bacteria use small molecules to orchestrate collective behaviors in a process called quorum sensing (QS), which relies on the production, release, and group-wide detection of extracellular signal molecules referred to as autoinducers. One QS autoinducer, termed AI-2, is broadly used for inter-species bacterial communication, including in the mammalian gut. AI-2 consists of a family of interconverting compounds and adducts originating from 4,5-hydroxy-2,3-pentanedione. This complex speciation, coupled with the inherent instability of AI-2 congeners, have complicated isolation efforts. It has been known that mammalian epithelial cells produce an AI-2 mimic to which bacteria respond. However, the identity of the AI-2 mimic has remained elusive, presumably due to its instability, similar to that of known AI-2 compounds. Here, we developed a reactivity-based metabolomics approach to capture and identify a mammalian AI-2 mimic. Using a chemical strategy targeted at the -diketone moiety of known AI-2s, we identify the unusual sugar L-xylosone, as well as the related metabolite L-xylulose, as AI-2 mimics. While L-xylulose is a common and naturally occurring sugar known in human metabolism, L-xylosone is a rare and highly reactive oxidation product. We established a facile synthetic route to access pure enantiomers of xylosone and confirmed that, like AI-2, the L-configuration is required for recognition by the bacterial AI-2 receptor, LuxP, whereas D-xylosone is inactive. L-xylosone is new to the human metabolome, suggesting that other chemically reactive small molecules that mediate host-microbe interactions await discovery. The identification of L-xylosone expands the AI-2 family of molecules and adds a new word to the lexicon of host-bacterial interactions.

biochemistry↗

bioRxiv2025

A generalized and efficient approach for complete mRNA design improves translation, stability and specificity

End-to-end, machine-learning-based design of mRNA molecules offers a powerful means to tailor their properties for specific tasks. mRNA expression level, immunogenicity, tissue specificity, stability, and localization strongly depend on sequence, providing a rich set of properties amenable to optimization. Despite this potential, the components of mRNA are governed by distinct grammatical and functional rules that hinder a unified approach to complete mRNA design. While machine learning and generative AI techniques can excel on individual sequence design tasks, out-of-distribution design, where the biological objective shifts substantially from the original training data, remains difficult. Moreover, there is a disconnect between available sequence generation technologies and the diverse body of biological datasets needed to form and test mechanistic hypotheses. In this work, we describe a simple and powerful alteration to integrated gradients (Design by Integrated Gradients or DIGs) that serves as the foundation for several mRNA design tasks and an agentic hypothesis engine, the Structured RNA Evidence Aggregation Module (STREAM), which enables rapid adaptation of this technique to new contexts. Using this framework, we demonstrate complete model-informed mRNA design and reveal the underexplored rules governing the assembly of mRNA components into high-performance transcripts. By linking neural-network-based design to independent datasets, we design complete mRNA sequences in shifted settings, culminating in up to 6-fold increases in intramuscular expression compared to state-of-the-art methods in vivo. Together, DIGs and STREAM enable automated mRNA design in increasingly complex settings.

synthetic biology↗

bioRxiv2022

Genetic or transcranial magnetic stimulation of B-RAF-MEK signaling promotes CST axon sprouting and functional regeneration

Facilitating axon regeneration in the injured central nervous system remains a challenging task. RAF-MEK signaling plays an important role in axon elongation during nervous system development. Here we show that activation of B-RAF in mature corticospinal neurons elicited the expression of a discrete set of transcription factors previously implicated in the regeneration of zebrafish optic nerve axons. Genetic activation of B-RAF-MEK signaling promoted robust regeneration and sprouting of corticospinal tract axons after injury. Newly sprouting axon collaterals formed synaptic connections with spinal interneurons, correlating with the recovery of skilled motor function. Seeking a non-invasive way to stimulate axon regeneration, we found that suprathreshold high-frequency repetitive transcranial magnetic stimulation activates the B-RAF canonical effectors MEK1/2 and requires MEK1/2 activity to promote corticospinal axon regeneration and sprouting after injury. These data demonstrate a central role of neuron-intrinsic RAF-MEK signaling in enhancing the growth capacity of mature corticospinal neurons and propose HF-rTMS as a potential therapy for spinal cord injury. One Sentence SummaryGenetic or HF-rTMS-mediated activation of B-RAF- MEK signaling promotes CST axon sprouting and functional regeneration after a spinal cord injury.

neuroscience↗

bioRxiv2021

Cortical glutamatergic projection neuron types contribute to distinct functional subnetworks

The cellular basis of cerebral cortex functional architecture remains not well understood. A major challenge is to monitor and decipher neural network dynamics across broad cortical areas yet with projection neuron (PN)-type resolution in real time during behavior. Combining genetic targeting and wide-field imaging, we monitored activity dynamics of subcortical-projecting (PTFezf2) and intratelencephalic-projecting (ITPlxnD1) types across dorsal cortex of mice during different brain states and behaviors. ITPlxnD1 and PTFezf2 neurons showed distinct activation patterns during wakeful resting, spontaneous movements, and upon sensory stimulation. Distinct ITPlxnD1 and PTFezf2 subnetworks were dynamically tuned to different sensorimotor components of a naturalistic feeding behavior, and optogenetic inhibition of ITsPlxnD1 and PTsFezf2 in subnetwork nodes disrupted distinct components of this behavior. Lastly, ITPlxnD1 and PTFezf2 projection patterns are consistent with their subnetwork activation patterns. Our results show that, in addition to the concept of columnar organization, dynamic areal and PN type-specific subnetworks are a key feature of cortical functional architecture linking microcircuit components with global brain networks.

neuroscience↗

bioRxiv2021

scTenifoldKnk: a machine learning workflow performing virtual knockout experiments on single-cell gene regulatory networks

Bigger PictureO_LIGene knockout (KO) experiments, using genetically altered animals, are a proven powerful approach to elucidate the role of a gene in a biological process. However, systematic KO experiments targeting many genes are usually prohibitive due to limited experimental and animal resources. Here, we present scTenifoldKnk, an efficient virtual KO tool that allows the systematic deletion of many genes individually. scTenifoldKnk uses single-cell RNA sequencing (scRNAseq) data from wild-type (WT) samples to predict gene function in a cell type-specific manner. We show that predictions made by scTenifoldKnk recapitulate findings from real-animal KO experiments. scTenifoldKnk has proven to be a powerful and effective approach for elucidating gene function, prioritizing KO targets, predicting experimental outcomes before real-animal KO experiments are conducted. C_LI HighlightsO_LIscTenifoldKnk performs virtual KO experiments using scRNAseq data. C_LIO_LIscTenifoldKnk only requires data from WT samples; no data is needed from KO samples. C_LIO_LIPredictions made by scTenifoldKnk recapitulate findings from real-animal KO experiments. C_LI Data Science Maturity Level O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=14 SRC="FIGDIR/small/436484v2_ufig1.gif" ALT="Figure 1"> View larger version (8K): org.highwire.dtl.DTLVardef@157126aorg.highwire.dtl.DTLVardef@17a2e76org.highwire.dtl.DTLVardef@805c4dorg.highwire.dtl.DTLVardef@12bdc1b_HPS_FORMAT_FIGEXP M_FIG C_FIG eTOC blurbscTenifoldKnk is a machine learning workflow performing virtual KO experiments to predict gene function. It constructs gene regulatory networks using single-cell RNA sequencing data from wild-type samples and then computationally deletes target genes. Real-data applications demonstrate that scTenifoldKnk recapitulates findings of real-animal KO experiments and accurately predicts gene function in analyzed cells. SummaryGene knockout (KO) experiments are a proven, powerful approach for studying gene function. However, systematic KO experiments targeting a large number of genes are usually prohibitive due to the limit of experimental and animal resources. Here, we present scTenifoldKnk, an efficient virtual KO tool that enables systematic KO investigation of gene function using data from single-cell RNA sequencing (scRNAseq). In scTenifoldKnk analysis, a gene regulatory network (GRN) is first constructed from scRNAseq data of wild-type samples, and a target gene is then virtually deleted from the constructed GRN. Manifold alignment is used to align the resulting reduced GRN to the original GRN to identify differentially regulated genes, which are used to infer target gene functions in analyzed cells. We demonstrate that the scTenifoldKnk-based virtual KO analysis recapitulates the main findings of real-animal KO experiments and recovers the expected functions of genes in relevant cell types.

bioinformatics↗

bioRxiv2020

scTenifoldNet: a machine learning workflow for constructing and comparing transcriptome-wide gene regulatory networks from single-cell data

Constructing and comparing gene regulatory networks (GRNs) from single-cell RNA sequencing (scRNAseq) data has the potential to reveal critical components in the underlying regulatory networks regulating different cellular transcriptional activities. Here, we present a robust and powerful machine learning workflow--scTenifoldNet--for comparative GRN analysis of single cells. The scTenifoldNet workflow, consisting of principal component regression, low-rank tensor approximation, and manifold alignment, constructs and compares transcriptome-wide single-cell GRNs (scGRNs) from different samples to identify gene expression signatures shifting with cellular activity changes such as those associated with pathophysiological processes and responses to environmental perturbations. We used simulated data to benchmark scTenifoldNets performance, and then applied scTenifoldNet to several real data sets. In real-data applications, scTenifoldNet identified highly specific changes in gene regulation in response to acute morphine treatment, an antibody anticancer drug, gene knockout, double-stranded RNA stimulus, and amyloid-beta plaques in various types of mouse and human cells. We anticipate that scTenifoldNet can help achieve breakthroughs through constructing and comparing scGRNs in poorly characterized biological systems, by deciphering the full cellular and molecular complexity of the data. HighlightsO_LIscTenifoldNet is a machine learning workflow built upon principal component regression, low-rank tensor approximation, and manifold alignment C_LIO_LIscTenifoldNet uses single-cell RNA sequencing (scRNAseq) data to construct single-cell gene regulatory networks (scGRNs) C_LIO_LIscTenifoldNet compares scGRNs of different samples to identify differentially regulated genes C_LIO_LIReal-data applications demonstrate that scTenifoldNet accurately detects specific signatures of gene expression relevant to the cellular systems tested. C_LI Short abstractWe present scTenifoldNet--a machine learning workflow built upon principal component regression, low-rank tensor approximation, and manifold alignment--for constructing and comparing single-cell gene regulatory networks (scGRNs) using data from single-cell RNA sequencing (scRNAseq). scTenifoldNet reveals regulatory changes in gene expression between samples by comparing the constructed scGRNs. With real data, scTenifoldNet identifies specific gene expression programs associated with different biological processes, providing critical insights into the underlying mechanism of regulatory networks governing cellular transcriptional activities.

bioinformatics↗

bioRxiv2019

Loss of the Krppel-like factor 4 tumor suppressor is associated with epithelial-mesenchymal transition in colorectal cancer

Colorectal cancer (CRC) is the third leading cancer-related cause of death due to its propensity to metastasize. Epithelial-mesenchymal transition (EMT) is a multistep process important for invasion and metastasis of CRC. Kruppel-like factor 4 (KLF4) is a zinc finger transcription factor highly expressed in differentiated cells of the intestinal epithelium. KLF4 has been shown to play a tumor suppressor role during CRC tumorigenesis - its loss accelerates development and progression of cancer. The present study examines the relationship between KLF4 and markers of EMT in CRC.\n\nMethodsImmunofluorescence staining for KLF4 and EMT markers was performed on archived patient samples after colorectal cancer resection and on colonic tissues of mice with colitis-associated cancer.\n\nResultsWe found that KLF4 expression is lost in tumor sections obtained from CRC patients and in those of mouse colon following azoxymethane and dextran sodium sulfate (AOM/DSS) treatment when compared to their respective normal appearing mucosa. Importantly, in CRC patient tumor sections we observed a negative correlation between KLF4 levels and mesenchymal markers including TWIST, {beta}-catenin, claudin-1, N-cadherin, and vimentin. Similarly, in tumor tissues from AOM/DSS-treated mice KLF4 levels were negatively correlated with mesenchymal markers including SNAI2, {beta}-catenin, and vimentin and positively correlated with the epithelial marker E-cadherin.\n\nConclusionThese findings suggest that the loss of KLF4 expression is a potentially significant indicator of EMT in CRC.

cancer biology↗