Biology subjects

Patel, A.

Publications and source records attributed to Patel, A..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

Inhibition of miR-1307 Reverses Resistance to Cisplatin in Drug-Resistant Oral Squamous Cell Carcinoma

BackgroundChemo-resistance remains a major clinical challenge in Oral Squamous Cell Carcinoma (OSCC), attributed to the intrinsically resistant cells. Although tumour-derived extracellular vesicles (EVs) have been implicated in cell-cell communication, their role in propagating chemo-resistance remains poorly defined. This study aims to identify salivary EV-associated miRNAs capable of predicting chemoresistance and to delineate the role of miR-1307-5p in modulating CSC-driven therapeutic refractoriness. MethodsSalivary EV-derived expression profile of miR-1307-5p was assessed by qPCR in chemo resistant OSCC patients and further validated in TCGA small RNA sequencing datasets. Expression was validated by qPCR and correlated with clinicopathological outcomes. Functional assays including cell-cycle analysis, apoptosis, migration/invasion, 3D spheroids, angiogenesis, and CAM assays were performed in miR-1307-5p inhibited CD44 CSC subpopulation compared to its vehicular control. Transcriptomic profiling cross-referencing with TCGA was conducted to identify potential novel targets of miR-1307-5p. Chemo-sensitisation was assessed by treating the knockdown chemo resistant cells with low dose cisplatin and validating it using in-vitro functional assays and orthotopic xenograft model. ResultsmiR-1307-5p was significantly elevated in salivary EVs of chemo resistant OSCC patients and correlated with poor overall survival (p = 0.03). The miRNA was markedly enriched in endogenously resistant CD44 CSCs. Silencing of miR-1307-5p induced G2/M arrest, triggered apoptosis, impaired invasion, and reduced angiogenesis both in-vitro and in ex-vivo assays. Transcriptomic profiling, TCGA validation, and integrative pathway analysis identified key oncogenic hubs which converge on PI3K-AKT, MAPK/ERK, and YAP signalling pathways governing EMT. Inhibition of miR-1307-5p restored cisplatin sensitivity in resistant CSCs, with low-dose cisplatin producing substantial tumour suppression in-vitro and in-vivo. Reduced CD44 expression in xenograft models confirmed CSC reprogramming. EVs from anti-miR-treated cells confer chemo sensitisation upon uptake by resistant CSCs. Xenograft models substantiated that EVs can initiate tumour formation and that EV-mediated delivery of anti-miR-1307-5p drives significant tumour regression. ConclusionThis study identifies salivary EV-derived miR-1307-5p as a clinically relevant biomarker of chemoresistance in OSCC and reveals its mechanistic role in sustaining CSC-driven therapeutic failure. Targeting miR-1307-5p offers a promising avenue for restoring cisplatin sensitivity and developing exosome-based therapeutic strategies. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=150 SRC="FIGDIR/small/709730v1_ufig1.gif" ALT="Figure 1"> View larger version (38K): org.highwire.dtl.DTLVardef@19f88e0org.highwire.dtl.DTLVardef@d36b95org.highwire.dtl.DTLVardef@3c2579org.highwire.dtl.DTLVardef@c04ef5_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2026

Precise measurement of rodent drinking using CLiQR (Capacitive Lick Quantification in Rodents)

Accurate quantification of rodent licking behavior is essential for studies of fluid intake, including investigations of alcohol use disorder and obesity. Lickometry systems vary widely in sensing modality, cost, scalability, and data resolution. Also, many available systems require specialized housing or store only binary lick/no lick data based on thresholding. Here, we present CLiQR (Capacitive Lick Quantification in Rodents), an open-source capacitive lickometry system designed for high-throughput recording of licking behavior in home-cage environments while preserving the full capacitance time series. The system uses MPR121 capacitive sensors connected to custom metal-tipped serological pipette sippers and a centralized desktop computer to record data from up to 24 mice concurrently, with capacity for two-bottle choice experiments. Validation experiments demonstrated that the capacitive signals reliably distinguish licking from non-licking interactions. Total lick counts showed a positive correlation with measured fluid consumption (R2 = 0.849, p < 0.0001), confirming that detected events provide a meaningful proxy for intake. All information necessary to reproduce the system is shared openly in this manuscript and online. By combining scalability, full-trace data acquisition, and low cost, CLiQR provides a flexible and extensible platform for high-throughput behavioral neuroscience experiments and enables retrospective improvement of lick-detection algorithms.

animal behavior and cognition↗

bioRxiv2026

Individualized and stereotypical seizure semiology in a porcine model of post-traumatic epilepsy.

Large-animal models of traumatic brain injury may yield translatable data on epileptogenesis, given their similarities in anatomy, brain size, and immune systems to humans. Adult male and female swine received bilateral cortical impact (N=16) or sham surgery (N=6) and were screened for convulsions via video-EEG for up to one year. Post-traumatic epilepsy (PTE) was defined as 2 seizures after 1-week post-injury. Nine out of sixteen pigs (56%) receiving bilateral cortical impact developed PTE, with an average 4.6 months ({+/-}3.4,SD) latent period. Seizures (N = 199) began focally, sometimes with motor onset including automatisms, before becoming generalized, with tonic-clonic or tonic convulsions. We defined a library differentiating peri-ictal behaviors (N = 31) from rhythmic/odd behaviors typical in healthy pigs (N = 12). Seizures had an average of 7.3 behaviors per seizure (max 26) lasting an average of 1.8 minutes (max 7.9). For seizures comprised of multiple convulsive episodes, the first convulsion had a greater number of peri-ictal behaviors than subsequent convulsions (P < 0.001). The array of peri-ictal behaviors displayed was pig-specific, with many behaviors consistently observed across seizures. The seizure frequency detected was 0.38/day. This large-brain model of PTE exhibits a variable period of epileptogenesis, a substantial rate of PTE, and an expansive repertoire of ictal behaviors. This first description of semiology in this species will serve as a guide for other porcine epilepsy models. Biofidelic models of PTE are expected to increase our understanding of the pathophysiology, enabling the identification and testing of therapeutics that translate into human patients. HighlightsO_LIThe average time from bilateral cortical impact to post-traumatic epilepsy in swine is 6 months, and is highly variable, ranging from 2 to 47 weeks post-TBI. C_LIO_LISwine with post-traumatic epilepsy display an array of specific behaviors around convulsions, distinct from pigs without post-traumatic epilepsy. C_LIO_LIThough the duration of convulsion was typically a few seconds, the entire seizure, with the associated peri-ictal behaviors, lasts up to 7.9 minutes. C_LIO_LIThe complexity of behaviors around convulsions tended to increase from early traumatic seizures to post-traumatic seizures. C_LIO_LIPeri-ictal behaviors observed around convulsions in an individual were often displayed prior to the first convulsion. C_LI Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=115 SRC="FIGDIR/small/708000v2_ufig1.gif" ALT="Figure 1"> View larger version (55K): org.highwire.dtl.DTLVardef@9523aborg.highwire.dtl.DTLVardef@158cfa6org.highwire.dtl.DTLVardef@1a8b499org.highwire.dtl.DTLVardef@e8c86f_HPS_FORMAT_FIGEXP M_FIG C_FIG

neuroscience↗

bioRxiv2026

Short-Context Regulatory DNA Language Models with Motif-Discovery Regularization

DNA language models (DNALMs) aim to learn representations of genomic sequence for variant interpretation, regulatory prediction, and sequence design. Most DNALMs are trained on whole genomes and long contexts, but regulatory DNA poses a distinct challenge: functional elements are sparse, context dependent, and encoded by short transcription factor motif syntax embedded in extensive background sequence. We introduce ARSENAL, a short-context masked DNA language model pretrained on ENCODE candidate cis-regulatory elements. ARSENAL recovers diverse transcription factor motifs de novo and improves zero-shot regulatory variant effect prediction relative to other DNALM foundation models. ARSENAL embeddings also improve supervised regulatory sequence models at predicting chromatin accessibility and regulatory variant scoring. Finally, ARSENAL serves as an efficient generative prior, enabling multi-objective regulatory sequence design with supervised oracles. ARSENAL shows that targeted self-supervised pretraining on regulatory regions can learn biologically meaningful and transferable regulatory representations without genome-scale training, long contexts or task-specific labels. Code is available at https://github.com/kundajelab/regulatory_lm. Models and data are shared at https://sageb.io/ydjhqM

genomics↗

bioRxiv2025

High dynamic range shortwave infrared (SWIR) imaging of mice with an InGaAs camera

SignificanceWhile shortwave infrared (SWIR) imaging provides superior tissue penetration and reduced autofluorescence for preclinical applications, quantitative fluorescence analysis is hindered by the limited dynamic range of InGaAs cameras, forcing a focus on either bright or dim anatomical features. AimWe develop a high dynamic range (HDR) imaging method specifically adapted for the high-noise characteristics of InGaAs detectors to enable quantitative fluorescence imaging across wide intensity ranges. We demonstrate that one-time camera calibration based on a series of images encompassing the range of radiance intensities enables all subsequent image processing. ApproachWe modified classical HDR algorithms with exposure-time-dependent dark current subtraction, preprocessing to exclude saturated and noisy pixels before camera response function recovery, and dynamic weighting range adjustment to account for shrinking intensity ranges at longer exposures. HDR image processing effects on preclinical imaging outcomes were analyzed using indocyanine green and SWIR-emitting PbS/CdS quantum dots in mouse models. ResultsHDR imaging achieved a 22 dB improvement in dynamic range over single exposures, enabling simultaneous quantification across more than three orders of magnitude of fluorophore concentration. In vivo studies showed improvements in contrast-to-noise ratios across all anatomical features, with improvements in vascular contrast while maintaining quantitative accuracy. After one-time camera calibrations, this approach enables rapid processing of subsequent datasets. ConclusionsThis software-based HDR SWIR imaging approach eliminates exposure parameter optimization and enables comprehensive biodistribution analysis across all anatomical structures from a single acquisition sequence, significantly streamlining preclinical imaging workflows while preserving quantitative accuracy.

bioengineering↗

bioRxiv2025

A metabolite extracted from E. coli suppresses tau aggregation

Tau aggregation is a key pathological feature of neurodegenerative diseases termed tauopathies. Identifying the various cellular factors that function to prevent tau aggregation in cells can generate key insights into how to mitigate diseases associated with protein misfolding. During an investigation into developing purification methods for the protein tau, we observed that isolates of E. coli lysate prevented human tau aggregation in vitro. Fractionation of the lysate was used to further isolate a small molecular weight (MW) inhibitory fraction containing multiple components, as determined by mass spectrometry and NMR. A putative inhibitory component, methylphosphonic acid (MePn), decreased tau amyloid formation when supplemented to in vitro aggregation assays. MePn also blocked the aggregation of expressed tau in live E. coli when supplemented to the culture media. Our findings can be directly applied to optimizing purification of recombinant tau protein and more broadly, highlight the potential of cellular metabolites to directly modulate tau amyloid formation.

biochemistry↗

bioRxiv2025

Molecular and cellular dynamics of measurable residual disease progression in myelodysplastic syndromes

Cancer relapse after treatment invariably proceeds from the persistence and progression of measurable residual disease (MRD), an ultrasmall malignant population. Despite the poor prognostic impact established by increasingly sensitive MRD detection methods, the molecular pathways defining MRD remain unknown. To identify the unique features of MRD and the molecular forces shaping its progression, we performed single-cell multi-omic profiling on DNA, RNA and protein layers for a longitudinal patient cohort with relapsed myelodysplastic syndromes (MDS) after stem cell transplantation (SCT), the only curative modality for MDS. We provide a comprehensive molecular portrait of MRD cells with novel markers, shared across genetically heterogeneous patients. MDS relapse after SCT manifested universally with marked phenotypic evolution. Genotype and phenotype analyses revealed MRD progression as a dynamic, evolutionary process rather than a static expansionary one, driven by both subclonal sweeping and cell state transitions. Malignant cells adapted to infiltrating T cells by rewiring IFN-{gamma} responses to activate a key immunoevasive pathway. Our study demonstrates the power of longitudinal, single-cell multi-omic analysis for identifying, tracking, and understanding MRD cells, opening new avenues to target MRD persistence and progression.

cancer biology↗

bioRxiv2025

Characterization of quiescent subpopulations and proliferative compartments in glioblastoma

Glioblastoma (GBM) quiescent (Q) cell populations are hypothesized to contain cancer stem-like cells (CSC) that drive tumor growth, cellular heterogeneity, and recurrence. However, GBM tumors do not neatly resolve into developmental hierarchies and Q stem-like activities are difficult to assess. Here, we evaluated tumor Q subpopulations in patient-derived GBM xenograft tumors using live cell reporters, DNA label retention assays, and single cell genomics. Compared to adult neural stems cells (NSCs), GBM Q populations contain hybrid transcriptional states composed of networks found in both dormant and activated adult NSCs, resulting in constitutive expression of key Q egress transcription factors and their targets (e.g., AP-1 and CCND1/2). As a result, even the longest Q-residing cells ([~]12 days) in xenograft tumors continuously cycle and fail to enter dormant Q states. We provide evidence and hypothesize that transient Q states in primary tumors arise as part of distinct proliferative compartments rather than deterministic developmental hierarchies driven by CSC activity. We further speculate that increases in basal translation rates drive Q instability in GBM tumors.

cancer biology↗

bioRxiv2025

Conditioned orienting predicts preference for, and affective response to, amphetamine after multiple exposures in male rats.

Conditioned orienting (OR) is a form of cue-directed behavior that is thought to reflect a behavioral phenotype of increased incentive-motivational processing of conditioned stimuli. Previous research has shown that OR also represents a cognitive phenotype that includes impaired attentional function and response inhibition, increased novelty seeking and, in females, resistance to extinction of conditioned place preference (CPP) for the stimulant drug amphetamine (AMP). The following experiments aim to increase our understanding of the OR cognitive phenotype in drug reward processing in male rats. In the first experiment, preference for AMP in males with or without an OR phenotype is assessed using CPP. In the second experiment, our understanding of OR as a predictor of AMP reward response is expanded by measuring potential behavioral mechanisms including affective and motor response to multiple AMP administrations. The results of the two experiments suggest that while the OR phenotype is not a predictor of resistance to extinction of AMP preference in males, OR as a discrete response variable predicts AMP preference after repeated exposures (i.e., AMP challenge). Moreover, we observed that rats with an OR phenotype had a higher positive affective response, measured via ultrasonic vocalizations (USVs) in the 50 kHz range, after multiple AMP administrations. OR scores predicted USVs but did not alter the gross motor response to AMP. Taken together, these results suggest that orienting behavior is a predictor of AMP preference in males, particularly after multiple exposures, and that one behavioral mechanism that could explain this is that male Orienters have a strong positive affective response to AMP.

animal behavior and cognition↗

bioRxiv2025

Cytoneme feedback ensures signaling specificity when multiple ligands converge on a common receptor

Cell-cell communication via conserved signaling pathways often involves multiple paracrine ligands that act through shared receptors and/or downstream cascades yet elicit distinct outcomes, raising the question of how specificity is achieved. In the Drosophila adult-muscle-precursor (AMP) niche, the FGF ligands Pyramus and Thisbe both signal through the Heartless (Htl) receptor in isogenic AMPs yet drive divergent responses. In-vivo imaging of endogenous Pyramus:mCherryendo and Thisbe:sfGFPendo knock-ins revealed that the ligands segregate into distinct, receptor-bound territories among isogenic AMPs. This segregation arises as AMPs acquire ligands via polarized, Htl-containing cytonemes that directly contact each ligand source. Cytoneme polarity and target-specificity are position-dependent, enabling ligand distribution to scale with tissue size and AMP organization. A positive feedback loop through the Htl-Pointed axis reinforces cytoneme polarity, amplifying ligand-specific segregation and signaling. These findings establish cell-intrinsic feedback as a general principle for sustaining cytoneme polarity and specificity, and reveal how cytonemes confer signaling specificity to pathways activated by multiple ligands within the same progenitor population.

developmental biology↗

bioRxiv2025

Influence of ULK1 activity on the memory of ATG101 auto-activation dynamics

The initiation of autophagy is marked by the prompt convergence of initiation proteins to form super molecular complexes in spatially-defined autophagic hubs. The recruitment and activation of the ULK-kinase complex is a pivotal event in starting the initiation cascade. ULK complex-component ATG101 is required for initiating autophagy in mammalian cells, but its function is unknown. Its HORMA domain interacts with the HORMA domain of ATG13 and transmembrane protein ATG9A to collectively form the essential autophagy initiation complex ATG9A-ATG13-ATG101. ATG101 has structurally malleable elements, reminiscent of topological conversions used in related metamorphic HORMA domain proteins to control the rate of complex formation. To elucidate the function of ATG101, we monitored the interaction kinetics of ATG101 with ATG13 and ATG9A, and observed that the interaction of ATG101 with ATG9A and ATG13 is exceptionally slow. Dramatic acceleration of complex formation is observed upon a change in the fold of ATG101 induced by its transient homo-dimerization, which in turn is initiated by phosphorylation by the ULK1 kinase. In an auto-catalytic mechanism, ATG101 dimers create a positive feedback to propagate activation to further ATG101 molecules in the absence of ULK1 activity. Despite the competitive nature of the interaction interfaces, homo-dimerization of ATG101 surprisingly accelerates its association to ATG13. Memory of ATG101 activation persists for many hours after dephosphorylation and continues to accelerate the assembly of the ATG9A-ATG13-ATG101 complex. Overall, this work proposes an unusual regulatory mechanism where UKL1 initiates an ATG101 auto activation cascade, whose memory creates a responsive positive feedback that dictates the assembly rate of a key complex in autophagy initiation.

biochemistry↗

bioRxiv2025

Multimodal single-cell analyses reveal distinct fusion-regulated transcriptional programs in Ewing sarcoma.

Ewing sarcoma (EwS) is a fusion-driven malignancy, peaking in adolescence. Although EwS tumors are driven uniquely by EWS::FLI1 and related fusions, patient outcomes vary greatly. If and how tumor plasticity of EWS::FLI1-regulated transcriptional signatures contribute to disease progression is not known. To address this, we utilized a single-cell co-assay of RNA and chromatin accessibility (ATAC) sequencing to identify gene regulatory networks in EwS. By comprehensively characterizing regulatory elements across cell lines, we identified multiple unique modules of gene regulation. Differential usage and prevalence of these modules was evident across cell lines, associated with distinct epigenetic and transcriptomic signatures, and in specific cases, modifiable by exogenous TGF-{beta}. When we examined primary EwS patient tumors, we observed these same regulatory modules were variably enriched both across and within tumors, highlighting the existence of intratumoral heterogeneity in gene regulatory networks. Our findings demonstrate that multiple, co-existing transcriptional programs shape the phenotypic diversity of EwS and suggest that the balance between these networks may have important implications for clinical outcomes and targeted therapy development. SummaryMultimodal transcriptional analysis reveal how Ewing sarcoma tumors use distinct gene programs, including one linked to TGF-{beta}, to drive cancer behavior and progression.

genomics↗

bioRxiv2025

IL-12 Secreting CAR-T Cells Reprogram the Tumor Microenvironment and Improve Efficacy Against Heterogeneous Models of Glioblastoma

BackgroundGlioblastoma (GBM) remains uniformly lethal due to pronounced intratumoral heterogeneity and a highly immunosuppressive microenvironment that limits the efficacy of targeted therapies. MethodsWe engineered chimeric antigen receptor (CAR) T cells targeting EGFRvIII and armored them with a single-chain interleukin-12 (scIL12) payload. These cells were tested in syngeneic, orthotopic GBM mouse models exhibiting heterogeneous EGFRvIII expression. CAR T cells were delivered intracranially without lymphodepletion. ResultsIntracranial administration of of scIL12-secreting CAR-T cells eradicated tumors without requiring lymphodepletion, achieving 50% long-term survival. Survival benefits depended entirely on endogenous CD8 T cells, as efficacy was abolished in CD8-deficient hosts and unaffected by NK cell depletion. Notably, therapeutic efficacy was abrogated by lymphodepletion, underscoring the necessity of an intact endogenous immune response. Mechanistically, scIL12 enhanced CAR-T cell persistence and reprogrammed tumor-associated microglia, indicating potential epitope spreading through polyclonal endogenous CD8+ T-cell responses, which facilitate the elimination of EGFRvIII-negative tumor cells. ConclusionsThis study demonstrates the pleiotropic benefits of IL-12 armored CAR-T cells with improved targeting of antigen-positive tumor cells and simultaneous remodeling of the microenvironment to engage adaptive immunity against antigen-negative clones. This strategy offers a potential clinically actionable approach to improve outcomes in GBM by circumventing the need for toxic lymphodepletion and addressing tumor heterogeneity.

immunology↗

bioRxiv2025

The CYP71A, NIT, AMI, and IAMH gene families are dispensable for indole-3-acetaldoxime-mediated auxin biosynthesis in Arabidopsis

Indole-3-acetic acid (IAA) is a crucial auxin governing plant development and environmental responses. While the indole-3-pyruvic acid (IPyA) pathway is the predominant IAA biosynthesis route, other pathways, like the indole-3-acetaldoxime (IAOx) pathway, have been proposed. The IAOx pathway has garnered attention due to its supposed activation in auxin-overproducing mutants (e.g., sur1, sur2, ugt74b1) and the auxin-like responses triggered by exogenous application of its proposed intermediates: IAOx, indole-3-acetonitrile (IAN), and indole-3-acetamide (IAM). However, despite supporting evidence for individual steps, conclusive physiological relevance of the IAOx pathway remains unproven. Using a comprehensive genetic approach combined with metabolic and phenotypic profiling, we demonstrate that mutating gene families proposed to function in the IAOx pathway does not result in prominent auxin-deficient phenotypes, nor are these genes required for high-auxin production in the sur2 mutant. Our findings also challenge the previously postulated linear IAOx pathway. While exogenously provided IAOx, IAN, and IAM can be converted to IAA in vivo, they do not act as precursors for each other. Finally, our findings question the physiological relevance of IAM and IAN as IAA precursors in plants and suggest the existence of a yet uncharacterized auxin biosynthetic route, likely involving IAOx as an intermediate, for the production of IAA in the sur2 mutant. Future identification of the metabolic steps and the corresponding genes in this new pathway may uncover the previously unknown way of synthesizing IAA in plants.

genetics↗

bioRxiv2025

Deep Phenomics of Pattern Recognition Receptor Agonist Specific Activation of Human Blood

Human whole blood (WB) immunophenotyping may represent the in vivo immunological state with better fidelity than artificially isolated peripheral blood mononuclear cells (PBMCs). We used a deep phenomics modeling approach to elucidate the quantitative differences in major immune cell lineages in WB and PBMC compartments in a steady-state in vitro setting. We studied functional innate immune responses induced by pattern recognition receptor agonist adjuvants (PRRa). Using an optimized 49-parameter CyTOF panel and implementing machine learning (ML) algorithms, we mapped PRRa-mediated CD69, CD40, CD80, CD86 and CCR7 activation at the nodal innate immune subsets level. We also portrayed cellular origin of innate functional chemokine CCL4 and intracellular cytokine IFN{gamma} production. We mapped neutrophils as the primary source of TLR7/8 agonist (TLR7/8a) and STING agonist mediated CCL4 responses in WB. Notably, in the PBMC fraction, where neutrophils are limited, natural killer (NK) cells became the major source of innate CCL4 production. TLR7/8a-mediated IFN{gamma} induction by early NK cells was mapped in PBMCs, which was limited in WB. Considering such distinctions, we hypothesized that deep phenomics employing a clinical sample that has not been manipulated, i.e., WB, may be additive in translating in vitro innate fingerprinting into in vivo biology.

immunology↗

bioRxiv2025

Peptide nucleic acid (PNA) and DNA hybrid three-way junctions and mesojunctions

Three-way junctions are simple and fundamental structural motifs that impart the typical branching property in most DNA nanostructures. While conventional three-way DNA junctions are well-known, mesojunctions are relatively unexplored. Here, we report the synthesis of peptide nucleic acid (PNA)/DNA hybrid three-way conventional and mesojunctions, containing a 14 bp duplex DNA arm and two PNA/DNA hybrid arms. We show that the PNA/DNA mesojunction can be assembled in magnesium-free buffers containing low concentrations of calcium or sodium salts. PNA/DNA hybrid junctions and mesojunctions showed higher thermal stability compared to the DNA versions. Further, PNA/DNA hybrid junctions assembled in sodium exhibited higher nuclease resistance against DNase I. Our results pave the way for PNA/DNA hybrid three-way junctions to be integrated into complex DNA nanostructure designs to improve their structural and enzymatic stability.

synthetic biology↗

bioRxiv2025

Scaling Large Language Models for Next-Generation Single-Cell Analysis

AO_SCPLOWBSTRACTC_SCPLOWSingle-cell RNA sequencing has transformed our understanding of cellular diversity, yet current single-cell foundation models (scFMs) remain limited in their scalability, flexibility across diverse tasks, and ability to natively integrate textual information. In this work, we build upon the Cell2Sentence (C2S) framework, which represents scRNA-seq profiles as textual "cell sentences," to train Large Language Models (LLMs) on a corpus comprising over one billion tokens of transcriptomic data, biological text, and metadata. Scaling the model to 27 billion parameters yields consistent improvements in predictive and generative capabilities and supports advanced downstream tasks that require synthesis of information across multi-cellular contexts. Targeted fine-tuning with modern reinforcement learning techniques produces strong performance in perturbation response prediction, natural language interpretation, and complex biological reasoning. This predictive strength enabled a dual-context virtual screen that nominated the kinase inhibitor silmitasertib (CX-4945) as a candidate for context-selective upregulation of antigen presentation. Experimental assessment in human cell models unseen during training supported this prediction, demonstrating that C2S-Scale can effectively guide the discovery of context-conditioned biology. C2S-Scale unifies transcriptomic and textual data at unprecedented scales, surpassing both specialized single-cell models and general-purpose LLMs to provide a platform for next-generation single-cell analysis and the development of "virtual cells."

bioinformatics↗

bioRxiv2025

Consequences of resource constraint on stochastic gene regulation

In this contribution, we systematically investigate how intracellular constraints on resources impact stochastic gene expression and regulation. We first consider a model of a single gene with discrete integer-valued mRNA and protein copy numbers that evolved stochastically based on probability occurrences of biochemical reactions. The resource constraints are imposed by considering a finite number of ribosomes binding to mRNAs to form a translation complex, and the complex dissociates to give back a free ribosome and a protein molecule. Analytical analysis reveals that ribosomal constraints reduce the magnitude of stochastic fluctuations in protein copy numbers, and also lead to lower statistical single-cell concordance between mRNA and protein levels of the same gene. We also identify parameter regimes where copy-number fluctuations become sub-Poisson - less variation than expected from a Poisson distribution. Considering fast ribosomal binding/unbinding to mRNAs, we also develop a reduced stochastic model that faithfully captures the statistical fluctuation of the system. Finally, we extend the model to consider an incoherent feedforward loop, and our results show noise minimization by resource constraints in gene regulation.

systems biology↗