Biology subjects

Song, L.

Publications and source records attributed to Song, L..

At least 73 records · Page 4Linked to original sources

bioRxiv2025

Saccades orchestrate intraocular glucose to shape visual responses in birds

Birds exhibit remarkable vision despite lacking the typical network of blood vessels in their eyes. The characteristic poses a long-standing question about how avian retinas fulfill high energy demands necessary for sight. Here we show that natural, rhythmic eye movements, known as oscillatory saccades, orchestrate intraocular metabolic dynamics and attention-guided visual processing in pigeons. In a series of integrated experiments, we monitored eye movements along with glucose levels in the eye and neuronal activity in key brain regions receiving direct input from the retina. We found that these saccades generate fluctuations in intraocular glucose concentrations, closely linked to changes in neuronal visual responses over timescales of seconds to minutes. Moreover, pharmacological manipulations that altered glucose availability and eliminated these oscillatory saccades resulted in corresponding shifts in neuronal responses, demonstrating a causal linkage between oscillatory saccades, metabolic regulation, and visual processing. These findings reveal a mechanism by which birds actively evoked saccades during attention-driven information gathering, propelling retinal metabolism and facilitating their vision in the absence of retinal vasculature. This study underscores the interplay among eye movements, metabolic regulation, and high-level visual performance, suggesting broader implications for how eye movements contribute to retinal health, attention, and visual function across species.

neuroscience↗

bioRxiv2025

Insights into the Biocontrol Mechanisms of a Paenibacilluspeoriae Strain on Maize Seedling Blight through Multi-Omics Analysis

Maize seedling blight, caused by the phytopathogenic fungus Fusarium verticillioides, is a common and rapidly spreading disease that negatively impacts grain quality and productivity. Control of this pathogen is complicated by its complex infection process and the tendency for resistance to conventional chemical pesticides. The use of biological control agents has been recognized as an environmentally friendly and sustainable solution for the control of plant diseases. In our study, we isolated and identified Paenibacillus peoriae 3-B4 from maize leaves, which exhibited an inhibition rate of 59.92% against F. verticillioides in greenhouse experiments. Genome sequencing of P. peoriae 3-B4 revealed a chromosome of 5,912,131 bp, featuring a GC content of 45.51% on average and 5383 annotated coding sequences. Eight gene clusters associated with secondary metabolites with antifungal activity and thirteen genes associated with induced systemic resistance and pattern-triggered immunity were identified. Transcriptomic analysis identified 8,997 differentially expressed maize genes, with key defense genes (e.g., NPR1, bZIP, MYB, LRR, WRKY) enriched in MAPK signaling and plant-pathogen interactions. 16S rRNA analysis showed significant shifts in microbial communities, particularly with an increase in the abundance of beneficial genera like Paenibacillus, Delftia, and Corynebacterium. The combined analysis of differentially expressed genes and microbial communities indicated that they synergistically enhance pathogen resistance in maize. Our findings outline the potential mechanisms by which P. peoriae 3-B4 inhibits F. verticillioides infection and open possibilities of exploiting biological control strategies to control maize seedling blight.

microbiology↗

bioRxiv2025

The Arabidopsis TIRome informs the design of artificial TIR (Toll/interleukin-1 receptor) domain proteins

The TIR (Toll/interleukin-1 receptor) domain is an ancient protein module that functions in immune and cell death responses across the Tree of Life. TIR domains encoded by plants and prokaryotes function as enzymes to produce diverse small molecule immune signals. Plant genomes can encode hundreds of TIR-domain containing proteins - many of which confer important agricultural disease resistance as TIR-NLR (nucleotide-binding, leucine-rich repeat) immune receptors. Despite their importance, how natural variation influences TIR enzymatic output and immunity-associated cell death is largely unexplored. We assayed a complete collection of the TIR domains of Arabidopsis thaliana Col-0 (the "AtTIRome") to explore variation in TIR metabolite production and cell death signaling. Roughly half of the AtTIRome triggered cell death in transient assays. Artificial TIR proteins designed based on consensus sequences of the AtTIRomes cell death phenotypic classes revealed polymorphisms controlling variation in TIR cell death elicitation and metabolite production. Structure-function analyses of artificial TIRs revealed that natural variation in the "BB-loop", a flexible region overlying the catalytic pocket, determines differences in function across Arabidopsis TIR-containing proteins. We further demonstrate that artificial TIRs are functional on an NLR chassis and that BB-loop variation can tune the activity of a natural TIR-NLR protein. These findings shed light on the diversity of TIR outputs and reveal methods to design and engineer TIR-based immune receptors. Significance StatementTIR (Toll/interleukin-1 receptors) domain proteins perform immune signaling across the Tree of Life. TIR domains are enzymes that can process NAD+ to generate diverse small molecule signals. In order to better understand TIR-based signaling, we leveraged Arabidopsis natural variation across [~]150 TIR proteins to inform the design of artificial TIRs, define features that control output, and tune the output of a natural TIR immune receptor. The engineering of plant immune pathways will enable the optimization of disease resistance to safeguard yields.

plant biology↗

bioRxiv2025

Divergent stem cell mechanisms governing the primary body axis and appendage regeneration in the axolotl

Exploring the fundamental mechanisms of organ regeneration is crucial for advancing regenerative medicine. The axolotl tail represents a unique opportunity to study regeneration of the primary axis including segmented muscle, vertebrae and skin. During tail development, muscle stem cells (MuSCs) displayed expected specificity to the muscle lineage. Tail amputation, however, induced expansion of MuSC potential yielding clonal contribution to muscle, connective tissue including cartilage, pericytes, and fibroblasts. This expanded potential was not observed during limb regeneration, and cross-transplantation showed these differences in potential are intrinsic. ScRNA-Seq profiling revealed that tail MuSCs revert to an embryonic mesoderm-like state. Through genetic manipulation involving the over-expression of constitutively active TGF-{beta} receptors or Smad7 (antagonist of TGF-{beta} signaling) in MuSCs, we demonstrated that the levels of TGF-{beta} signal determine the fate outcome of MuSCs to connective tissue lineage or muscle respectively. Our findings illustrate a fundamental difference between regeneration of primary axis versus limb and offers a novel stem cell source for regeneration of axial skeletal tissues.

developmental biology↗

bioRxiv2025

Unlocking the biosynthetic potential of Paenibacilli through a genus-wide exploration of gene clusters for secondary metabolite production

The genus Paenibacillus is a prolific producer of secondary metabolites with diverse ecological and industrial applications. However, a comprehensive overview of the biosynthetic gene cluster (BGC) diversity and distribution throughout the genus has been limited. Here, we performed large-scale genome mining on 284 high-quality genomes and generated a non-redundant dataset of 126 representative genomes to explore the biosynthetic potential of this genus. A total of 3,273 BGCs were identified from the 284 genomes that clustered into 1,013 gene cluster families (GCFs), with 98.7% classified as unknown, indicating vast potential for novel secondary metabolite discovery in the Paenibacillus genus. Comparative analysis revealed significant phylogenetic and clade-specific distribution patterns of GCFs, with certain clades enriched in unique biosynthetic pathways while others exhibited low similarity to known BGCs, suggesting evolutionary adaptation to diverse ecological niches. This study uncovers the rich and largely untapped biosynthetic potential of the genus Paenibacillus, providing a foundation for future exploration of its natural products and their applications in biotechnology and medicine. IMPORTANCEBacterial secondary metabolites have been instrumental in the development of antibiotics, antifungals, and other bioactive compounds. The genus Paenibacillus is an underexplored source of such metabolites, with significant potential for novel discoveries. By integrating genome mining and phylogenetic analysis, this study systematically characterizes the diversity, distribution, and novelty of biosynthetic gene cluster across the genus. The identification of clade-specific biosynthetic patterns and numerous unknown gene cluster families highlights Paenibacillus as a promising target for uncovering novel compounds with ecological and therapeutic relevance. These findings not only expand our understanding of bacterial secondary metabolite biosynthesis but also offer new opportunities for the development of sustainable biotechnological applications.

microbiology↗

bioRxiv2025

Microbial general model: Leveraging large language model for contextualized microbiome analysis

Microbial communities significantly impact medicine, biotechnology, and agriculture. Advanced sequencing technologies have generated extensive microbiome data, enabling the discovery of substantial evolutionary and ecological patterns. However, traditional supervised learning methods struggle to capture universal patterns in microbial community data, largely due to the large data heterogeneity and profound batch effects among samples, rendering it difficult to classify samples as well as detect biomarkers from millions of samples, not to say the intricate but important dynamic patterns from a variety of contextualized sceneries. In this study, we introduce the Microbial General Model (MGM), the first microbiome community foundation model pre-trained on a dataset of 263,302 microbiome samples using language modeling techniques. MGM demonstrated significant improvements in microbial community classification compared to traditional machine learning methods. Additionally, MGM has enabled contextualized classification, effectively overcomes cross-regional limitations, showing enhanced performance on intercontinental datasets through transfer learning. Furthermore, fine-tuning MGM on a longitudinal infant dataset revealed distinct keystone genera during development, with Bacteroides and Bifidobacterium exhibiting higher attention weights in vaginal deliveries, and Haemophilus in cesarean deliveries. Finally, through in silico modeling, the model also uncovered novel microbial dynamic patterns in a Crohns disease cohort following antibiotic treatment. In conclusion, by leveraging self-attention and autoregressive pre-training, MGM serves as a versatile model for various downstream microbiome tasks and holds significant potential for achieving contextualized aims. Key pointsO_LIThe Microbial General Model (MGM) is a foundation model with millions of parameters pre-trained on sub-million microbial community data. C_LIO_LIMGM outperforms traditional methods in various microbiome classification and prediction tasks, such as microbial community classification. C_LIO_LIMGM effectively captures the spatial and temporal dynamics of microbial communities. C_LIO_LIMGM could detect the effects of perturbation on microbial community through in silico experiments. C_LI

bioinformatics↗

bioRxiv2024

HMG-DTI: a drug-target interaction prediction framework combining Hierarchical Molecular Graphs and improved Convolutional Block Attention Module

Drug-target interactions (DTIs) are fundamental to understanding the therapeutic mechanisms of drugs, yet accurately predicting these interactions remains a significant challenge in drug discovery. Current computational approaches often fail to capture essential molecular motifs and spatial information of proteins, limiting their effectiveness, particularly when encountering proteins or compounds absent in the training datasets. To address these limitations, we propose mHMG-DTI, a novel framework that leverages an improved Convolutional Block Attention Module (iCBAM) for enhanced protein feature extraction and modified Hierarchical Molecular Graphs (mHMGs) for comprehensive molecular encoding. This hierarchical approach not only captures detailed local structures and broader connectivity patterns but also incorporates guiding knowledge to improve feature representation. Across a total of 16 experimental evaluations on four benchmark datasets spanning both classification and regression tasks, mHMG-DTI surpasses existing baseline models in 11 cases. These results highlight the potential of mHMG-DTI to enhance DTI prediction accuracy, thereby accelerating the drug discovery process and providing valuable insights into drug resistance and side effect mechanisms. Contactzeruiyang2-c@my.cityu.edu.hk

bioinformatics↗

bioRxiv2024

StayRose: a photostable StayGold derivative red-shifted by genetic code expansion

Photobleaching of fluorescent proteins often limits the acquisition of high-quality images in microscopy. StayGold, a novel dimeric green fluorescent protein recently monomerised through sequence engineering, addresses this challenge with its high photostability. There is now focus on producing different colour StayGold derivatives to facilitate concurrent tagging of multiple targets. The unnatural amino acid 3-aminotyrosine has previously been shown to red-shift super folder GFP upon incorporation into its chromophore via genetic code expansion. Here we apply the same strategy to red-shift StayGold through substitution of Tyrosine-58 with 3-aminotyrosine. The resultant red fluorescent protein, StayRose, shows 530 nm excitation and 588 nm emission peaks, shifting from the 497 nm and 504 nm excitation and emission peaks of StayGold. StayRose also retains the favourable photostability of StayGold and can be similarly monomerised using mutations at the dimer interface. A high-resolution crystal structure of StayRose confirms the modified structure of the amino-chromophore within an unperturbed 3D fold. Although reliant on genetic code expansion, StayRose provides an important step towards developing red-shifted StayGold derivatives.

bioengineering↗

bioRxiv2024

Mechanistic Insights into Dengue Virus Inhibition by a Clinical Trial Compound NITD-688

Dengue, caused by the dengue virus (DENV), presents a significant public health challenge with limited effective treatments. NITD-688 is a potent pan-serotype DENV inhibitor currently in Phase II clinical trials. However, its mechanism of action is not fully understood. Here, we present the molecular details of how NITD-688 inhibits DENV. NITD-688 binds directly to the nonstructural protein 4B (NS4B) with nanomolar affinities across all four DENV serotypes and specifically disrupts the interaction between NS4B and nonstructural protein 3 (NS3) without significantly changing the interactions between NS4B and other viral or host proteins. NS4B mutations that confer resistance to NITD-688 reduce both NITD-688 binding to NS4B and disruption of the NS4B/NS3 interaction. Specifically, NITD-688 blocks the interaction of NS3 with a cytosolic loop within NS4B. This inhibits the formation of new NS4B/NS3 complexes and disrupts pre-existing complexes in vitro and DENV-infected cells, ultimately inhibiting viral replication. Consistent with this mechanism, NITD-688 retains greater potency in cellular assays with delayed treatment compared to JNJ-1802, another NS4B inhibitor that has been studied in Phase II clinical trials. Together, these findings provide critical insights into the mechanism of action of NITD-688, facilitating the development of novel flavivirus NS4B inhibitors and informing future clinical interventions against DENV.

microbiology↗

bioRxiv2024

Accurate and General DNA Representations Emerge from Genome Foundation Models at Scale

Language models applied to protein sequences have become a panacea, enabling therapeutics development, materials engineering, and core biology research. Despite the successes of protein language models, genome language models remain nascent. Recent studies suggest the bottleneck is data volume or modeling context size, since long-range interactions are widely acknowledged but sparsely annotated. However, it may be the case that even short DNA sequences are modeled poorly by existing approaches, and current models are unable to represent the wide array of functions encoded by DNA. To study this, we develop AIDO.DNA, a pretrained module for DNA representation in an AI-driven Digital Organism [1]. AIDO.DNA is a seven billion parameter encoder-only transformer trained on 10.6 billion nucleotides from a dataset of 796 species. By scaling model size while maintaining a short context length of 4k nucleotides, AIDO.DNA shows substantial improvements across a breadth of supervised, generative, and zero-shot tasks relevant to functional genomics, synthetic biology, and drug development. Notably, AIDO.DNA outperforms prior encoder-only architectures without new data, suggesting that new scaling laws are needed to achieve computeoptimal DNA language models. Models and code are available through Model-Generator in https://github.com/genbio-ai/AIDO and on Hugging Face at https://huggingface.co/genbio-ai.

bioinformatics↗

bioRxiv2024

Balancing Locality and Reconstruction in Protein Structure Tokenizer

The structure of a protein is crucial to its biological function. With the expansion of available protein structures, such as those in the AlphaFold Protein Structure Database (AFDB), there is an increasing need for efficient methods to index, search, and generate these structures. Additionally, there is a growing interest in integrating structural information with models from other modalities, such as protein sequence language models. We present a novel VQ-VAE-based protein structure tokenizer, AIDO.StructureTokenizer (AIDO.St), which is a pretrained module for protein structures in an AI-driven Digital Organism [1]. AIDO.StructureTokenizer is a 300M parameter model consisting of an equivariant encoder to discretize input structures into tokens, and an invariant decoder to reconstruct the inputs from these tokens. In addition to evaluating structure reconstruction ability, we also compared our model to Foldseek, ProToken, and ESM3 in terms of protein structure retrieval ability. Through our experiments, we discovered an intriguing trade-off between the encoders locality and retrieval ability and the decoders reconstruction ability. Our results also demonstrate that a better balance between retrieval and reconstruction enables a better alignment between the structure tokens and a protein sequence language model, resulting in better structure prediction accuracy. Models and code are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

Retrieval Augmented Protein Language Models for Protein Structure Prediction

The advent of advanced artificial intelligence technology has significantly accelerated progress in protein structure prediction, with AlphaFold2 setting a new benchmark for prediction accuracy by leveraging the Evoformer module to automatically extract co-evolutionary information from multiple sequence alignments (MSA). To address AlphaFold2s dependence on MSA depth and quality, we propose two novel models: AIDO.RAGPLM and AIDO.RAGFold, pretrained modules for Retrieval-AuGmented protein language model and structure prediction in an AI-driven Digital Organism (Song et al., 2024). AIDO.RAGPLM integrates pre-trained protein language models with retrieved MSA, surpassing single-sequence protein language models in perplexity, contact prediction, and fitness prediction. When sufficient MSA is available, AIDO.RAGFold achieves TM-scores comparable to AlphaFold2 while operating up to eight times faster, and significantly outperforms AlphaFold2 when MSA is insufficient ({Delta}TM-score=0.379, 0.116 and 0.059 for 0, 5 and 10 MSA sequences as input). Additionally, we developed an MSA retriever using hierarchical ID generation that is 45 to 90 times faster than traditional methods, expanding the MSA training set for AIDO.RAGPLM by 32%. Our findings suggest that AIDO.RAGPLM provides an efficient and accurate solution for protein structure prediction, particularly in scenarios with limited MSA data. The AIDO.RAGPLM model has been open-sourced and is available on https://huggingface.co/genbio-ai/AIDO.Protein-RAG-3B.

bioinformatics↗

bioRxiv2024

Scaling Dense Representations for Single Cell with Transcriptome-Scale Context

Developing a unified model of cellular systems is a canonical challenge in biology. Recently, a wealth of public single-cell RNA sequencing data as well as rapid scaling of self-supervised learning methods have provided new avenues to address this longstanding challenge. However, rapid parameter scaling has been essential to the success of large language models in text and images, while similar scaling has not been attempted with Transformer architectures for cellular modeling. To produce accurate, transferable, and biologically meaningful representations of cellular systems, we develop AIDO.Cell, a pretrained module for representing gene expression and cellular systems in an AI-driven Digital Organism [1]. AIDO.Cell contains a series of 3M, 10M, 100M, and 650M parameter encoder-only dense Transformer models pre-trained on 50 million human cells from diverse tissues using a read-depth-aware masked gene expression pretraining objective. Unlike previous models, AIDO.Cell is capable of handling the entire human transcriptome as input without truncation or sampling tricks, thus learning accurate and general representations of the human cells entire transcriptional context. This pretraining with a longer context was enabled through FlashAttention-2, mixed precision, and large-scale distributed systems training. AIDO.Cell (100M) achieves state-of-the-art results in tasks such as zero-shot clustering, cell-type classification, and perturbation modeling. Our findings reveal interesting loss scaling behaviors as we increase AIDO.Cells parameters from 3M to 650M, providing insights for future directions in single-cell modeling. Models and code are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

Mixture of Experts Enable Efficient and Effective Protein Understanding and Design

Proteins play a fundamental role in life. Understanding the language of proteins offers significant potential for gaining mechanistic insights into biological systems and introduces new avenues for treating diseases, enhancing agriculture, and safeguarding the environment. While large protein language models (PLMs) like ESM2-15B and xTrimoPGLM-100B have achieved remarkable performance in diverse protein understanding and design tasks, these models, being dense transformer models, pose challenges due to their computational inefficiency during training and deployment. In this work, we introduce AIDO.Protein, a pretrained module for protein representation in an AI-driven Digital Organism [1]. AIDO.Protein is also the first mixture-of-experts (MoE) model in the protein domain, with model size scales to 16 billion parameters. Leveraging a sparse MoE architecture with 8 experts within each transformer block and selectively activating 2 experts for each input token, our model is significantly more efficient in training and inference. Through pre-training on 1.2 trillion amino acids collected from UniRef90 and ColabfoldDB, our model achieves state-of-the-art results across most tasks in the xTrimoPGLM benchmark. Furthermore, on over 280 ProteinGym Deep Mutational Scanning (DMS) assays, our model achieves nearly 99% of the overall performance of the best MSA-based model and significantly outperforms the previously reported state-of-the-art models that do not utilize MSA. We also adapted this model for structure-conditioned protein sequence generation tasks and achieved new SOTA in this domain. These results indicate that AIDO.Protein can serve as a strong foundation model for protein understanding and design. Models and codes are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

AtSDR4L and paralog promote seed-to-seedling transition in multiple seed compartments

Development is a series of decision-making events. Success of seed plants at individual and population levels strongly depends on the timing of germination and the rapidness of seedling establishment. Arabidopsis thaliana SEED DORMANCY 4-LIKE (AtSDR4L) and its paralog Dynamic Influencer of Gene Expression 2 (DIG2) are transcriptional co-repressors that promote seed-to-seedling phase transition. Their regulatory roles in promoting germination at the temporal and tissue-specific scales remain elusive. We show that strong germination arrest of Atsdr4l dig2 is alleviated by ABA antagonists. Isolated mutant embryos develop faster than intact seeds, but still exhibit delayed growth. Atsdr4l dig2 seeds show extensive changes in gene expression in both the seed coat and the embryo, with a subset of the genes differentially expressed tissue-specifically. MIKC-type MADS-box genes are the top-enriched transcription factor family among up-regulated genes in both seed compartments of Atsdr4l dig2, and AGAMOUS-LIKE44 (AGL44) is a dirct target of both AtSDR4L and DIG2. Many hormonal genes and hypoxia-responsive genes are misregulated in the double mutant seeds, accompanied by an over-accumulation of abscisic acid, auxin, their derivatives, as well as the immediate precursor of ethylene. Together, these results provide new mechanistic insights into how AtSDR4L and DIG2 work in concert to coordinate multiple pathways and prepare seeds for germination.

plant biology↗

bioRxiv2024

A Large-Scale Foundation Model for RNA Function and Structure Prediction

Originally marginalized as an intermediate in the information flow from DNA to protein, RNA has become the star of modern biology, holding the key to precision therapeutics, genetic engineering, evolutionary origins, and our understanding of fundamental cellular processes. Yet RNA is as mysterious as it is prolific, serving as an information store, a messenger, and a catalyst, spanning many underchar-acterized functional and structural classes. Deciphering the language of RNA is important not only for a mechanistic understanding of its biological functions but also for accelerating drug design. Toward this goal, we introduce AIDO.RNA, a pre-trained module for RNA in an AI-driven Digital Organism [1]. AIDO.RNA contains a scale of 1.6 billion parameters, trained on 42 million non-coding RNA (ncRNA) sequences at single-nucleotide resolution, and it achieves state-of-the-art performance on a comprehensive set of tasks, including structure prediction, genetic regulation, molecular function across species, and RNA sequence design. AIDO.RNA after domain adaptation learns to model essential parts of protein translation that protein language models, which have received widespread attention in recent years, do not. More broadly, AIDO.RNA hints at the generality of biological sequence modeling and the ability to leverage the central dogma to improve many biomolecular representations. Models and code are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

A T6SS DNase effector induces nuclear DNA Damage to trigger apoptosis via activation of the cGAS-STING-TNF axis

In the relentless battle for survival, bacteria have evolved sophisticated weaponry like the Type VI secretion system (T6SS) to not only outcompete rivals but also manipulate host cell processes to their advantage. Apoptosis, a controlled form of programmed cell death without triggering inflammatory responses, is often exploited by pathogens. While many bacterial pathogens are known to induce apoptosis, the role of T6SS effectors in this process remains unexplored. In this study, we discovered that Yersinia pseudotuberculosis (Yptb) utilizes its T6SS to secrete a DNase effector, TkeA, which induces apoptosis in host cells. Mechanistically, the translocation of TkeA into host cells causes nuclear DNA damage, leading to the release of fragmented DNA into the cytoplasm. This, in turn, activates the DNA-sensing cyclic GMP-AMP synthase (cGAS)/stimulator of interferon genes (STING) pathway. The activation of the cGAS-STING pathway by TkeA subsequently triggers apoptosis in host cells via extrinsic pathways, with tumor necrosis factor (TNF) signaling playing a critical role. Additionally, TkeA also mediates interbacterial competition, promoting the colonization of Yptb in mices gut. Our findings reveal that the T6SS-secreted DNase effector TkeA acts as a trans-kingdom effector, enhancing interbacterial competition and inducing apoptosis in host cells by activating the cGAS-STING-TNF axis. This discovery adds a new dimension to the mechanisms employed by bacterial pathogens to manipulate host cell death pathways by using their secretion systems. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/613802v1_ufig1.gif" ALT="Figure 1"> View larger version (62K): org.highwire.dtl.DTLVardef@1bfd70org.highwire.dtl.DTLVardef@19bfc89org.highwire.dtl.DTLVardef@1e0bf9dorg.highwire.dtl.DTLVardef@4412b9_HPS_FORMAT_FIGEXP M_FIG C_FIG HIGHLIGHTS* Yptb T6SS secretes a trans-kingdom DNase effector TkeA * TkeA degrades bacterial DNA to mediate interbacterial competition * TkeA causes nuclear DNA damage in host cells to induce apoptosis * TkeA induces apoptosis via the cGAS-STING-TNF axis

microbiology↗

bioRxiv2024

A benchmarking study of copy number variation inference methods using single-cell RNA-sequencing data

Single-cell RNA-sequencing (scRNA-seq) has emerged as a powerful tool for cancer research, enabling in-depth characterization of tumor heterogeneity at the single-cell level. Recently, scRNA-seq copy number variation (scCNV) inference methods have been developed, expanding the application of scRNA-seq to study genetic heterogeneity in cancer using transcriptomic data. However, the fidelity of these methods has not been investigated systematically. In this study, we benchmarked five commonly used scCNV inference methods, HoneyBADGER, CopyKAT, CaSpER, inferCNV, and sciCNV. We evaluated their performance across four different scRNA-seq platforms derived data from a multicenter study. We further evaluated the scCNV performance using scRNA-seq datasets derived from mixed samples consisting of five human lung adenocarcinoma cell lines and generated a clinical scRNA-seq dataset from a human small cell lung cancer patient to validate our findings. Our evaluation criteria included sensitivity and specificity of CNV detection, and subclone identification from mixed cancer samples. We found that the sensitivity and specificity of the five scCNV inference methods varied, depending on the selection of reference data, sequencing depths, and read lengths. Overall, CopyKAT and CaSpER exhibited superior performance to other methods, while inferCNV, sciCNV, and CopyKAT outperformed other methods in subclone identification accuracy. Remarkably, inferCNV achieved high accuracy in subclone identification when using data from a "single scRNA-seq protocol", however, when applying these methods to a dataset derived from multiple scRNA-seq platforms from the mixed samples, we found that batch effects significantly affected the performance of subclone identification for most methods, except for HoneyBADGER. Our benchmarking study revealed the strengths and weaknesses of each of the five scCNV inference methods and provided guidance for selecting the optimal CNV inference method using scRNA-seq data.

bioinformatics↗