Biology subjects

Xiong, P.

Publications and source records attributed to Xiong, P..

15 recordsLinked to original sources

bioRxiv2026

LigRNA: High-accuracy prediction of RNA-ligand binding sites via dynamic chemical complementarity and language model encoding

RNA-binding small molecules represent a validated therapeutic modality, yet pinpointing their exact RNA binding sites remains critical for mechanistic understanding and drug optimization. Experimental methods, while precise, are costly and low-throughput, necessitating robust computational alternatives. Approaches based on static feature concatenation may overlook input-dependent relationships between RNA residues and ligand functional groups. Here we present LigRNA, a graph neural network framework that predicts RNA-ligand binding sites through joint cross-modal encoding of both molecules. LigRNA integrates structure-guided language model embeddings with physicochemical features-solvent-accessible surface area and secondary structure types, to construct comprehensive RNA representations. Ligands are encoded as attributed molecular graphs via graph attention networks, preserving atom-level chemical topology rather than compressing molecules into fixed fingerprints. A cross-attention mechanism then enables dynamic, ligand-dependent modulation of RNA nucleotide attention, capturing context-specific interaction patterns. Benchmarked on TE9 and TE18 datasets, LigRNA achieves F1 scores of 0.73 and 0.58, improving on GATRsite by 7.4% and 18.4%, respectively. Ablation studies identify the RNA residual encoder as the most consistent contributor across the two benchmarks. The source code is freely available at https://github.com/WillowWang0216/LigRNA.

bioinformatics↗

bioRxiv2026

ReTIF: Granularity-Aware Multitask Interaction Routing for RNA-Compound Interaction Prediction and Binding-Site Localization

RNA-targeted drug discovery requires both RNA-compound interaction prediction and nucleotide-level binding-site (BS) localization. These two tasks rely on different levels of interaction information: DTI prediction summarizes overall RNA-compound compatibility, whereas BS localization requires preserving nucleotide-level compound-associated signals. However, existing multitask models often use shared cross-modal representations for both tasks before task-specific prediction layers, which may limit their ability to preserve task-dependent interaction patterns. We therefore propose ReTIF (Relation-enhanced Task-specific Interaction Framework), which constructs separate cross-modal interaction representations for DTI prediction and BS localization before aggregation. ReTIF integrates multi-source RNA-compound representations from frozen RNA-FM, StructRFM, Mole-BERT, and MolFormer encoders, and builds separate interaction representations for DTI scoring and BS localization. The DTI branch captures global compatibility through aggregation, whereas the BS branch preserves nucleotide-compound resolution and enhances local evidence through relation-guided propagation with RNA structural and compound topological priors. An asymmetric DTI-derived compatibility signal provides global context to BS prediction while maintaining local evidence. Across five-fold evaluations under unseen pair, RNA, compound, and joint shifts with 13 baselines, ReTIF has the highest mean in 13 of 16 scenario-metric combinations, including BS AUPR in all four settings.

bioinformatics↗

bioRxiv2025

Acoustoluminescence in Transition Metal and Rare Earth Oxides Beyond 1800 nm for In Vivo Imaging

Acoustoluminescence (AL) is promising for molecular imaging in living tissue, but efficient acousto-optic conversion remains challenging due to the [~]108 times difference in quantum energy between phonons (1 MHz ultrasound) and photons (visible or near-infrared light). Here, we report AL in transition metal oxides (TMOs) and rare earth oxides (REOs) at wavelengths beyond 1800 nm in the short-wave infrared (SWIR) or near-infrared II (NIR-II, 1000-3000 nm) window, under ultrasound excitation at power densities 100-150-fold lower than those required for sonoluminescence in liquids. High-temperature N2/H2-mixed gas reduction was demonstrated as a safe and efficient method to regulate the AL spectra and brightness of TMOs and REOs. TMOs exhibited broadband NIR-II AL emission. Intrinsic emission peaks of rare earth ions and non-conventional luminescence were observed in the AL spectra of REOs under ultrasound excitation. NIR-II AL imaging enabled twice the penetration depth of fluorescence imaging. We developed a scanning focused ultrasound AL imaging system for in vivo tumor imaging through the intact hindlimb, achieving acoustic resolution and penetration depths exceeding one centimeter.

bioengineering↗

bioRxiv2025

Toward Accurate RNA Non-Canonical Structure Prediction: The NC-Bench Benchmark and the NCfold Framework

AO_SCPLOWBSTRACTC_SCPLOWAccurate prediction of non-canonical (NC) base pairs is a pivotal step toward un-raveling the full functional landscape of RNA. This goal requires both a standardized dataset for evaluation and powerful models that overcome data limitations. Here, we make a dual contribution to this endeavor: the NC-Bench benchmark and the NCfold framework. NC-Bench offers a trusted ground for comparison with 925 sequences and 6,708 annotated NC pairs, defining tasks for edge and orientation prediction. NCfold provides a powerful analytical engine, a closed-loop dual-branch network that synergizes sequence features with structure-aware embeddings from RNA foundation models via our proposed REFweighted self-attention. This iterative design enables robust learning from scarce NC data. On the NC-Bench benchmark, NCfold establishes new state-of-the-art results, significantly outperforming existing methods. Comprehensive analyses validate its design and affirm the critical role of NC-Bench in driving future progress. NC-Bench and NCfold together form a cornerstone for the next generation of RNA structure prediction. The datasets and codes are publicly available at https://github.com/heqin-zhu/NCBench

bioinformatics↗

bioRxiv2025

Identification and quantification of alternative polyadenylation sites in single cell RNA-seq data using scPAISO

AbstractsAlternative polyadenylation (APA) is a critical posttranscriptional mechanism that generates transcriptomic diversity through the production of mRNA isoforms with distinct 3 UTRs or coding sequences. Current APA analysis based on single-cell RNA sequencing (scRNA-seq) for establishing cell type-specific APA landscapes primarily rely on Read2 data, which lacks precise cleavage site (CS) information. This limitation restricts their ability to achieve precise de novo mapping of polyadenylation sites (PASs). Here, we present single-cell PolyAdenylation ISOform quantification (scPAISO), a computational pipeline designed for de novo identification of PAS and quantification of PAS isoforms in scRNA-seq data, by leveraging the often-discarded Read1 from 3 tag-based scRNA-seq protocols. Unlike existing tools, scPAISO directly captures mRNA 3 end cleavage sites, enabling superior performance in motif enrichment (stronger AAUAAA signal) and peak precision (sharper PAS peaks). Moreover, the smaller peak widths enhance the spatial resolution, enabling more accurate detection of closely spaced PASs in the genome. By integrating Read1 and Read2 data, scPAISO achieves isoform-level quantification with an assignment accuracy exceeding 95%. We demonstrate the robustness of scPAISO in identifying PASs and quantifying APA events across diverse biological contexts, including hematopoiesis, systemic sclerosis (SSc), and mouse tissues. We identified stage-specific 3 UTR lengthening in hematopoietic progenitors, global 3 UTR remodeling in SSc and tissue-specific polyadenylation (PA) preference along with RNA-binding proteins in mice. Overall, scPAISO represents a significant advancement in the analysis of APA at single-cell resolution and provides a powerful tool for exploring the regulatory landscape of APA, offering new insights into transcriptome complexity and gene regulation in both health and disease.

bioinformatics↗

bioRxiv2025

A fully-open structure-guided RNA foundation model for robust structural and functional inference

RNA language models have achieved strong performances across diverse downstream tasks by leveraging large-scale sequence data. However, RNA function is fundamentally shaped by its hierarchical structure, making the integration of structural information into pre-training essential. Existing methods often depend on noisy structural annotations or introduce task-specific biases, limiting model generalizability. Here, we propose structRFM, a structure-guided RNA foundation model that is pre-trained on millions of RNA sequences and secondary structures data by integrating base pairing interactions into masked language modeling through a novel pair matching operation. We further introduce MUSES (multi-source ensemble of secondary structures) to mitigate model bias, and a dynamic masking ratio to balance the structure-guided mask and nucleotide-level mask. structRFM learns joint knowledge of sequential and structural data, producing versatile representations, including classification-level, sequence-level, and pairwise matrix features, that support a broad spectrum of downstream adaptations. structRFM ranks among the top models in zero-shot homology classification across seventeen biological language models, and sets new benchmarks for secondary structure prediction. structRFM further derives Zfold, which enables robust and reliable tertiary structure prediction, with consistent improvements in estimating 3D structures and their accordingly extracted 2D structures, achieving a pronounced about 20% performance gain compared with baselines and comparable performances with AlphaFold3 on CASP15-natural, CASP16, and RNA-Puzzles datasets. In functional tasks such as internal ribosome entry site identification, structRFM achieves a whopping 48% performance gain in F1 score. Furthermore, state-of-the-art performances in extensive experiments across novel RNA families and long non-coding RNAs indicate the robustness and generalizability of structRFM. These results demonstrate the effectiveness of structure-guided pre-training and highlight a promising direction for developing multi-modal RNA language models in computational biology. To support the broader scientific community, we have made the 21-million sequence-structure dataset and the pre-trained structRFM model fully open-source, facilitating the development of multimodal foundation models in biology.

bioinformatics↗

bioRxiv2025

Modeling human phenotypes in the nematode C. elegans during an undergraduate developmental biology course

Caenorhabditis elegans is a widely used model organism in biomedical research due to its genetic tractability, short life cycle, and conservation of many developmental processes with humans. In this study, undergraduate students conducted nine independent experiments during a Developmental Biology course to model human phenotypes using C. elegans. Each group selected a human phenotype of interest, identified a gene associated with the phenotype, and then determined the orthologous or homologous gene in C. elegans. By obtaining mutants and designing phenotypic assays, students investigated the extent to which the worm models could recapitulate aspects of the human condition. This collective work highlights both the potential and limitations of C. elegans as a model for human phenotypic variation and disease and demonstrates the value of undergraduate inquiry as a catalyst for scientific engagement and research-based learning.

scientific communication and education↗

bioRxiv2024

Deep generalizable prediction of RNA secondary structure via base pair motif energy

Deep learning methods have demonstrated great performance for RNA secondary structure prediction. However, generalizability is a common unsolved issue on unseen out-of-distribution RNA families, which hinders further improvement of the accuracy and robustness of deep learning methods. Here we construct a base pair motif library that enumerates the complete space of the locally adjacent three-neighbor base pair and records the thermodynamic energy of corresponding base pair motifs through de novo modeling of tertiary structures, and we further develop a deep learning approach for RNA secondary structure prediction, named BPfold, which learns relationship between RNA sequence and the energy map of base pair motif. Experiments on sequence-wise and family-wise datasets have demonstrated the great superiority of BPfold compared to other state-of-the-art approaches in accuracy and generalizability. We hope this work contributes to integrating physical priors and deep learning methods for the further discovery of RNA structures and functionalities.

bioinformatics↗

bioRxiv2024

Identification of P-type ATPase as a bacterial transporter for host-derived small RNA

Host-derived small RNAs represent a finely tuned host response to maintain the host-microbial homeostasis. Among these, an emerging class known as transfer RNA-derived small RNA (tsRNA) has been implicated in modulating microbial-host interaction. Our previous study showed that when challenged with an oral pathobiont, Fusobacterium nucleatum (Fn), an immortalized human oral keratinocyte cell line releases certain Fn-targeting tsRNAs that selectively inhibit the growth of Fn via their ribosome-targeting function. We also revealed the sequence- and energy-dependent uptake of tsRNAs by Fn. However, the mechanism underlying the tsRNA uptake at the molecular level remains elusive. In this study, using RNA affinity pull-down assay in combination with Mass Spectrometry, we identified a putative P-type ATPase transporter (PtaT) in Fn, which binds Fn-targeting tsRNAs in a sequence-specific manner. AlphaFold 3 simulation provides further evidence supporting the specific binding between PtaT and tsRNA compared to the scrambled control and the DNA counterpart. Through targeted mutagenesis and phenotypic characterization, we demonstrated the important role of PtaT in the uptake and antimicrobial capacity of tsRNAs against Fn in both ATCC 23726 type strain and a clinical tumor isolate (Fn CTI). Furthermore, global RNA sequencing and label-free Raman spectroscopy revealed the phenotypic differences between Fn wild type and PtaT-deficient mutant, highlighting the functional significance of PtaT in purine and pyrimidine metabolism. Collectively, our work identifies a bacterial transporter for tsRNAs and provides critical information for a fundamental understanding of how the host-generated tsRNAs specifically interact with its targeted bacteria at the molecular level.

microbiology↗

bioRxiv2024

Multi-step engineered adeno-associated virus enables whole-brain mRNA delivery

Adeno-associated viruses (AAVs) are commonly used vectors for DNA delivery in gene therapy. Here we developed a system that enables the AAV shell to package mRNAs by multi-step introduction of RNA-packaging components and modification of AAV Rep proteins. The resultant mRNA-carrying AAVs (RAAVs) retained most properties of conventional AAVs, including capsid composition, virus morphology, and tissue tropism. These RAAVs could mediate mRNA transfer into target cells and tissues, leading to transient expression of the functional protein. Importantly, intravenously injected RAAVs efficiently crossed the blood-brain barrier (BBB) and infected the whole mouse brain. Thus, the DNA viral vector could be modified for RNA delivery, and our RAAV represents the first highly efficient BBB-crossing mRNA delivery system that could be used for therapeutic purposes via whole-brain infection.

bioengineering↗

bioRxiv2023

RNA tertiary structure modeling with BRiQ potential in CASP15

We describe the modeling method for RNA tertiary structures employed by team AIchemy_RNA2 in the 15th Critical Assessment of Protein Structure Prediction (CASP15). The method consists of the following steps. Firstly, secondary structure information was derived from various manually-verified sources. With this information, the full length RNA was fragmented into structural motifs. The structures of each motif were predicted and then assembled into the full structure. To reduce the searching conformational space, a RNA structure was organized into an optimal base folding tree. And to further improve the sampling efficiency, the energy surface was smoothed at high temperatures during the Monte Carlo sampling to make it easier to move across the energy barrier. The statistical potential energy function BRiQ was employed during Monte Carlo energy optimization.

bioinformatics↗

bioRxiv2021

Metformin modulates microbiota-derived inosine and ameliorates methamphetamine-induced anxiety and depression-like withdrawal symptoms in mice

ObjectiveMetformin exhibits therapeutic potential in behavioural deficits induced by methamphetamine (METH) in rats. Emerging studies suggest gut microbiota may impact psychiatric symptoms, but there is no direct evidence supporting metformins participation in the pathophysiology of withdrawal symptoms via modulation of gut microbiota. MehodsIn order to define the functional contributions by gut microbiota and metformin to the behavioural deficits during METH withdrawal, we utilized a combination of fecal microbiota transplantation (FMT), high-throughput sequencing, and untargeted metabolomics technologies. ResultsFirst, METH addicts exhibited higher diversity and distinct microbial structures compared to heathy controls. In particular, the relative abundance of Rikenellaceae was positively correlated with the severity of anxiety and depression. Second, both human-to-mouse and mouse-to-mouse FMTs confirmed that METH-altered-microbiota transplantation is sufficient to promote anxiety and depression-like behaviours in recipient germ-free mice, and these behavioural disturbances could be ameliorated by metformin. In-depth analysis revealed that METH significantly altered the bacterial composition and structure as well as relative abundance of several bacterial taxa and metabolites, including Rikenellaceae and inosine, respectively, whereas add-on metformin could remodel these alterations. Finally, the inosine complementation successfully restored METH-induced anxiety and depression-like behaviours in mice. DiscussionThis study demonstrates that METH withdrawal-induced anxiety and depression-like behaviours are convertible and transmissible via gut microbiota in a mouse model. The therapeutic effects of metformin on psychiatric manifestations are associated with microbiota-derived metabolites, highlighting the role of the gut microbiota in substance use disorders and the pathophysiology of withdrawal symptoms. Study HighlightsO_ST_ABSWhat is known?C_ST_ABSO_LIThere are no targeted therapies for substance withdrawal syndrome, but there is considerable evidence that withdrawal-associated psychiatric manifestations contribute to the poor adherence to rehabilitation treatment as well as the relapse rates. C_LIO_LIMetformin has shown its therapeutic potential against METH-induced neurobehavioural changes and neurodegeneration in rats through CREB/BDNF and Akt/GSK3 signaling pathways in the anxiety-related brain nuclei. C_LI What is new here?O_LIMETH withdrawal-induced anxiety and depression-like behaviours are convertible and transmissible via gut microbiota in a mouse model. C_LIO_LIThe therapeutic effects of metformin on psychiatric manifestations are associated with microbiota derived metabolites. C_LIO_LIInosine complementation could restore METH withdrawal-induced anxiety and depression-like behaviours. C_LI

animal behavior and cognition↗

bioRxiv2021

Resurrecting self-cleaving mini-ribozymes from 40-million-year-old LINE-1 elements in human genome

Long Interspersed Nuclear Element (LINE) retrotransposons play an important role in genomic innovation as well as genomic instability in many eukaryotes including human. Random insertions and extinction through mutational inactivation make them perfectly time-stamped "DNA fossils". Here, we investigated the origin of a self-cleaving ribozyme in 5 UTR of LINE-1. We showed that this ribozyme only requires 35 nucleotides for self-cleavage with a simple but previously unknown secondary-structure motif that was determined by deep mutational scanning and covariation analysis. Structure-based homology search revealed the existence of this mini-ribozyme in anthropoids but not in prosimians. In human, the most homologs of this mini-ribozyme were found in lineage L1PA6-10 but essential none in more recent L1PA1-2 or more ancient L1PA13-15. We resurrected mini-ribozymes according to consensus sequences and confirmed that mini-ribozymes were active in L1PA10 and L1PA8 but not in L1PA7 and more recent lineages. The result paints a consistent picture for the emergence of the active ribozyme around 40 million years ago, just before the divergence of the new world monkeys (Platyrrhini) and old-world monkeys (Catarrhini). The ribozyme, however, subsequently went extinct after L1PA7 emerged around 30 million years ago with a deleterious mutation. This work uncovers the rise and fall of the mini-LINE-1 ribozyme recorded in the "DNA fossils" of our own genome. More importantly, this ancient, naturally trans-cleaving ribozyme (after removing the non-functional stem loop) may find its modern usage in bioengineering and RNA-targeting therapeutics.

molecular biology↗

bioRxiv2020

Characterization of miRNAs encoded by Autographa californica nucleopolyhedrovirus

Two Autographa californica nucleopolyhedrovirus (AcMNPV) encoded miRNAs, AcMNPV-miR-1 and -miR-3, have been reported in 2013 and 2019. Here, we present an integrated investigation of AcMNPV-encoded miRNAs. Six candidate miRNAs were predicted through small RNA deep sequencing and bioinformatics, of which, five validated by experiments. Three miRNAs perfectly matched the coding sequence of viral genes. The other two are located in coding sequences of viral genes. Targets in both virus and host were predicted and subsequently tested using dual-luciferase reporter assay. The validated targets were found mainly in AcMNPV, except for the targets of AcMNPV-miR-4, which are all host genes. Based on reporter assays, the five miRNAs predominantly function by down-regulating their targets, though individual target is slightly up-regulated. The transcription start sites of these miRNAs were analyzed. Our results suggest that AcMNPV-encoded miRNAs function as fine modulators of the interactions between host and virus by regulating viral and/or host genes. Author summaryVirus-encoded miRNAs have been widely reported as modulators participating in almost all biological processes. However, among Baculoviridae, which consists of a large family of dsDNA viruses that infect numerous beneficial insects and agricultural pests, only several have been reported encoding miRNAs. To clarify the roles of AcMNPV-encoded miRNAs in host-pathogen interactions, we employed RNA deep sequencing and series of experimental approaches identifying AcMNPV-encoded miRNAs, followed by target validation and function deduction. Among them, AcMNPV-miR-1 and AcMNPV-miR-3 have been reported in 2013 and 2019, respectively. This study reveals the sites of these miRNAs in the genome, both in coding sequences and complements, suggesting diverse functions. These miRNAs target genes in the virus itself and in the host, largely by suppressing expression, with some enhancing it. The transcription initiations of the miRNAs were analyzed. Our results provide some insight into the finely regulated process of baculovirus infection.

microbiology↗

bioRxiv2019

Accurate inference of the full base-pairing structure of RNA by deep mutational scanning and covariation-induced deviation of activity

Despite the transcription of noncoding RNAs in 75% of the human genome and their roles in many diseases include cancer, we know very little about them due to lack of structural clues. The centerpiece of the structural clues is the full RNA base-pairing structure of secondary and tertiary contacts that can be precisely obtained only from costly and inefficient 3D structure determination. Here, we performed deep mutational scanning of self-cleaving CPEB3 ribozyme by error-prone PCR and showed that a library of <5x104 single-to-triple mutants is sufficient to infer all 26 including nonhelical and noncanonical base pairs at the precision of a single false positive. The accurate inference, further confirmed by a twister ribozyme, is resulted from covariation analysis by utilizing both functional and nonfunctional variants for unsupervised learning, followed by restrained optimization. The result highlights the usefulness of deep mutational scanning for high-accuracy structural inference.

molecular biology↗