Biology subjects

Goodarzi, H.

Publications and source records attributed to Goodarzi, H..

At least 19 recordsLinked to original sources

bioRxiv2026

A high-resolution protein-RNA binding atlas for structure-aware interaction prediction and post-transcriptional regulatory modeling

RNA-binding proteins (RBPs) control post-transcriptional gene expression by regulating RNA processing, localization, stability, and translation. However, our understanding of RNA regulation remains limited by the resolution, scale, and uniformity of available RBP binding maps. Here we introduce eclair CLIP (ecCLIP), a high-throughput CLIP-seq platform that combines endogenous protein tagging for uniform affinity capture with ligation-free library preparation to generate high-resolution RNA-protein binding profiles. We applied ecCLIP to 67 endogenously tagged RBPs in HEK293T cells, generating a high-quality transcriptome-scale atlas that recovered known binding preferences and captured post-transcriptional regulons linked to RNA half-life. To test the utility of the atlas, we incorporated ecCLIP into Borzuki, a Borzoi-derived sequence-to-function model trained on complementary RNA degradation, expression, CLIP, and Perturb-seq data. Inclusion of ecCLIP data improved RNA half-life prediction and assessment of 3' UTR variant effects on RNA stability, demonstrating that our atlas helps capture RNA dynamics. To directly model the cis-regulatory grammar underlying these RBP-RNA interactions, we developed Pythia, a structure-aware neural network architecture that incorporates context-free grammar-inspired fixed convolutions to explicitly encode RNA base-pairing features. Pythia predicted RBP binding from sequence and structural context and decomposed predictions into sequence- and structure-dependent components. We tested this decomposition for five RBPs using massively parallel reporter CLIP (MPRC) binding assays in cells, using synthetic RNA libraries containing mutations designed to preserve or disrupt predicted RNA structures across nominated binding sites. We used the same MPRC strategy to confirm allele-specific binding effects for variants prioritized by Pythia. Together, ecCLIP and Pythia provide a resource for structure-aware decoding of RNA-protein interaction grammar and interpretation of post-transcriptional regulatory variation.

genomics↗

bioRxiv2026

The adaptive architecture of tRNA dependencies across physiological tumor environments

The genetic code is decoded by a highly redundant transfer RNA (tRNA) repertoire, yet whether this apparent redundancy serves a functional role beyond ensuring robust translation remains unclear. Here, we establish an atlas of tRNA dependencies through large-scale CRISPRi perturbation mapping across matched in vivo tumor growth and in vitro cell culture for 16 cancer cell lines spanning seven tissue types. During in vivo tumor growth, tRNA dependencies exhibited tissue organization at isodecoder resolution and were associated with codon demand at the isoacceptor level. Strikingly, these relationships were reorganized when cells were cultured in vitro, establishing environmental sensitivity of tRNA dependency--in contrast to protein components of the decoding machinery such as aminoacyl-tRNA synthetases (aaRSs), whose dependencies were largely preserved across environments. Despite pronounced context specificity, environmental alterations in tRNA dependency were dominated by a shared response across cancer cell lines, organized across isoacceptor and isodecoder levels. By disentangling stable cellular identity from environment-associated dependency states, we resolved a shared axis of isoacceptor dependency remodeling and identified specific tRNA isoacceptors associated with promoting or restraining adaptation to the tumor environment. We further connect this functional architecture to differential translation of codon-usage-biased programs associated with proliferation and tumor-microenvironmental stress. Together, our study reveals that tRNA redundancy does not imply functional equivalence, but instead forms a structured and environmentally plastic control layer linking coding information to translational output.

genomics↗

bioRxiv2026

A map of RBP-miRNA regulatory connections for deciphering the mechanisms of transcriptome remodeling in cancer

Micro-RNAs (miRNAs) are key post-transcriptional regulators of gene expression that influence cancer progression and outcome. Whereas transcriptional dysregulation of miRNA abundance underlies abnormal functionality of some miRNAs in cancer, emerging data points to RNA-binding proteins (RBPs) as prominent post-transcriptional regulators of miRNA activity. Here, by modelling the regulon activities of well-characterized miRNA families in RNA-seq data of a comprehensive collection of RBP inhibition experiments, human cell lines and tissues, we report a map of regulatory connections between 34 RBPs and 28 miRNAs. Furthermore, through a global analysis of the abundance as well as the activity of miRNAs in thousands of tumor and control samples from 17 human cancers, we show that RBP-driven regulation of miRNA activity is common in tumors and contributes to the heterogeneity in miRNA activity profiles. We uncover a set of RBPs that regulate the activity of tumor-suppressor and oncogenic miRNAs, including miR-29-3p, miR-17-5p and miR-19-3p. Of particular interest, we describe a pan-cancer negative regulatory role for DDX55 on the activity of the miR-29-3p family, a critical miRNA family with tumor suppressor functions that is involved in most human cancers. Our functional studies confirm that DDX55 inhibition restores the gene-silencing activity of miR-29-3p and reduces cancer cell viability. Overall, our study provides the first pan-cancer investigation of functional regulatory connections between RBPs and miRNAs, and highlights RBPs whose interventional modulation may serve as a new approach to reprogram abnormal miRNA activities in cancer.

genomics↗

bioRxiv2026

PIE: Generalizing perturbation effects across unseen perturbations, contexts and datasets

Predicting cellular response to perturbations is key to understanding biological mechanisms and selecting therapeutic targets. However, generalizing across cellular contexts, perturbations, and experimental datasets remains challenging due to incomplete representations of the system being perturbed and the difficulty of measuring true perturbation effects within that system. We introduce PIE, a model that addresses these limitations by reformulating the learning task around population level perturbation effects and incorporating auxiliary inputs that better characterize the biological system. PIE directly predicts differentially expressed genes and changes in gene expression for each (context, perturbation) pair, using external biological knowledge, baseline gene expression, and observed perturbation responses. Its architecture accommodates input sources with variable feature sets and predicts effects on genes seen and unseen during training. On the Replogle Nadig dataset, PIE achieves state of the art performance on most metrics across all tested generalization settings, including the most challenging setting where both the context and perturbation are unseen during training. For predicting differentially expressed genes, PIE achieves 1.2 to 3.2 times the AUPRC of the strongest baseline in each setting: across unseen contexts, perturbations, experimental datasets, and jointly unseen contexts and perturbations. Overall, PIE provides a framework for generalizing perturbation effects across diverse experimental conditions.

bioinformatics↗

bioRxiv2026

Extracellular cGAMP transmission exacerbates neuroinflammation in aging and multiple sclerosis

Neuroinflammation drives neurodegeneration and cognitive decline in aging, but the signals that spread this inflammation across the CNS remain unclear. Here, we show in mice that advanced age and experimental autoimmune encephalomyelitis (EAE) both result in elevated cytosolic mtDNA in microglia, which stimulates the production and export of cGAMP from these cells. Meanwhile, ENPP1, the dominant cGAMP hydrolase in the brain and spinal cord, is downregulated in both EAE mice and MS patients, releasing the brake on extracellular cGAMP accumulation. In Enpp1H362A mice, which selectively lack cGAMP hydrolysis activity, unrestrained extracellular cGAMP worsens EAE-associated paralysis and drives premature aging phenotypes including impaired motor coordination (rotarod, pole test) and reduced exploration (open field test) in a manner dependent on STING. Single-cell RNA sequencing pinpoints microglia, astrocytes, and inhibitory neurons as the principal cGAMP responders that activate STING-dependent interferon-stimulated gene induction in response to elevated extracellular cGAMP. Mechanistically, microglia import cGAMP through the LRRC8A:C channel, while astrocytes and neurons import cGAMP via the glutamate antiporter system xc-, reinforced by SLC1A2-mediated glutamate uptake in astrocytes and SLC38A1/2-fueled glutamine-to-glutamate synthesis in neurons. In isolated primary astrocytes and neurons, high intracellular glutamate and system xc--driven redox demands favor cGAMP uptake, raising the possibility that neurotransmitter cycling and redox burden shape cellular vulnerability to cGAMP immunotransmission in vivo. Together, these findings position ENPP1 as a protective brake on neuroinflammation in aging and MS, nominating cGAS inhibition and CNS ENPP1 enhancement as therapeutic strategies.

cell biology↗

bioRxiv2026

A massively parallel synthetic gene atlas for learning compact cis-regulatory grammar across cellular contexts

Virtual-cell models increasingly learn from large perturbation atlases, but their view of cis regulation remains limited to endogenous genes embedded in broad native regulatory contexts. Here we introduce Therna Biosciences' Chronos platform and its first public dataset release, comprising two complementary modules of a massively parallel synthetic gene atlas: Penta-47x27K for 5' UTRs/internal-promoter elements and Tria-47x28K for 3' UTR stability elements. Together, these datasets measure ~60,000 compact cis-regulatory elements across approximately 50 cell lines in one pooled experiment. The choice of cell lines here, provided to us courtesy of Tahoe Tx, was rooted in our aim to make this dataset maximally useful for the virtual-cell modeling community. At Therna, we routinely apply Chronos -- part of our RNA-Logix(TM) platform -- beyond cancer cell-line pools, to systems such as primary cells and organoids, via LNP-formulated mRNA delivery. Chronos captures both transcriptional and post-transcriptional gene expression control. Episomal DNA delivery measures DNA-normalized mRNA output, while direct RNA delivery with longitudinal sampling reliably quantifies RNA decay. The RNA-delivery arm uses chemically modified synthetic mRNA incorporating N1-methylpseudouridine, a modification widely used in mRNA therapeutics. To resolve high-complexity libraries of up to 30,000 elements, Chronos pushes the sensitivity limits of single-cell RNA sequencing to quantify individual RNA molecules within single cells. By massively expanding gene regulatory networks with synthetic genes whose variable regulatory code is short and defined, Chronos provides an auxiliary cis-regulatory lens for virtual-cell modeling and enables context-specific cis-trans regulatory interactions to be learned directly.

systems biology↗

bioRxiv2026

APOE4 disrupts the central dogma by arresting neuronal proteome dynamics

Apolipoprotein E4 (APOE4) is the strongest genetic risk factor for late-onset Alzheimers disease and promotes neuronal dysfunction through incompletely understood mechanisms. Here, we integrated transcriptomic, translatomic, and proteomic profiling of isogenic APOE3 and APOE4 human iPSC-derived neurons and found that APOE4 fundamentally impairs neuronal proteome renewal. Although transcriptional changes were modest, APOE4 disrupted ribosome occupancy, altered translational dynamics, and uncoupled protein abundance from transcript levels. Proteome-wide turnover measurements revealed a global extension of protein half-lives and widespread accumulation of long-lived proteins. Functional proteomic analyses demonstrated concurrent lysosomal and proteasomal impairments associated with reduced proteasome activity and increased association of APOE with neuronal proteasomes. Longitudinal proteomics further showed that protein accumulation emerges during neuronal maturation and precedes a senescence-like cellular stress state. Together, these findings identify impaired proteome renewal as a central mechanism underlying neuronal vulnerability to APOE4 and establish defective proteostasis as an early pathogenic event in Alzheimers disease.

neuroscience↗

bioRxiv2026

Semantic fragment representations for coordinate-free analysis of genomics data

Many genomic assays begin with individual DNA fragments, but standard analysis quickly collapses those molecules into counts over genomic intervals. Rich information carried by each fragment, including its sequence, fragment body, cleavage boundaries, and local flanking context, is lost in this process. This loss is especially apparent in mixed-source and heterogeneous samples, where individual fragments originate from disparate cell types and can retain information about their cell of origin. To address this, we present LEAF-1, a fragment-level foundation model pre-trained on approximately 58 billion fragments spanning bulk ATAC-seq, single-cell ATAC-seq, and cell-free DNA profiles, representing each DNA molecule as a point in a learned semantic space defined by sequence context, assay modality, and explicit cleavage-boundary tokens. In sparse scATAC-seq datasets, mean-pooled LEAF-1 embeddings readily classify human cell types from as few as [~]1,000 fragments per cell, with high-scoring fragments linked to cell-type-associated transcription-factor programs. Similarly, in cell-free DNA profiling, LEAF-1 outperformed state-of-the-art coordinate-binning strategies and general-purpose DNA language model baselines across cancer detection tasks. Applying attention-based multiple-instance learning to LEAF-1 embeddings further improved cancer detection, reaching an area under the receiver operating characteristic (ROC) curve (AUC) of 0.95. This pan-cancer model generalizes beyond cancer types it is trained on, as we show by profiling plasma samples from clear cell renal cell carcinoma patients and healthy volunteers and applying the frozen classifier without retraining, achieving an AUC of 0.83. These results show that semantic learning over individual DNA fragments preserves biochemical, cell-associated, and disease-associated signals that are otherwise lost during coordinate-based aggregation.

genomics↗

bioRxiv2026

Back to basics: Observed statistics are sufficient to predict drug responses

Predicting how cells, tissues, and patients will respond to a drug, cytokine, or genetic perturbation is central to biological and clinical reasoning. The practical goal is to estimate the analysis-ready readouts that support this reasoning: which cellular responses are context-dependent, which perturbations reveal shared or divergent mechanisms, and which observations should become the basis for the next experiment. Here we introduce Rhaister, a perturbation-response predictor that operates directly on screen-level summary statistics. By measuring just a few perturbations in a new biological context, Rhaister predicts the unmeasured perturbations by learning how response patterns vary across reference contexts. This formulation applies to both fine-grained molecular readouts, such as transcriptional responses from Tahoe-100M or other large perturbation screen, or on phenotypic endpoints. To train and apply Rhaister on pheno-typic endpoints we created Emerald Bay, a purpose-built Tahoe dataset that unifies multi-day cancer drug perturbation, pooled Mosaic tumor contexts, and paired transcriptomic response measurements*. Across these settings, Rhaister matches or exceeds substantially more expensive virtual-cell models, often achieving the highest values possible in evaluation metrics, while training in seconds and running predictions in milliseconds. On Emerald Bay, Rhaister predicts context-specific drug phenotypes from sensitivity measurements alone and improves further when including transcriptomic information. We also introduce Rhaister-O, predicting drug responses in new contexts from baseline expression alone and, to our knowledge, provides the first zero-shot model for this task. Rhaister establishes summary-statistic perturbation modeling as a fast, interpretable framework for predicting biological response across new contexts.

genomics↗

bioRxiv2026

Recursive mutational robustness in cancer through intra- and inter-genic compensation

Mutational robustness is a fundamental property of biological systems allowing phenotypic stability against genotypic perturbation. At the gene level, this is primarily attributed to inter-genic functional redundancy driven by paralogs. Here, we show that intra-genic functional redundancy, driven by alternatively spliced isoforms, is another prevalent source of mutational robustness. Integrating isoform-resolved pan-cancer genomics and transcriptomics data across tumors and cancer cell lines, we show that alternative isoforms often bypass deleterious somatic mutations. Mutation-skipping isoforms exhibit compensatory upregulation indicative of robustness against deleterious effects, especially in highly expressed genes. We found that intra-genic robustness is strongly associated with perturbation tolerance, is highly context-specific and, notably, can be complemented and augmented recursively by inter-genic robustness. Non-tumour-suppressor genes generally show higher robustness than tumour suppressor genes, and intra-genic compensation of their mutations is associated with worse disease outcome in most cancer types. Furthermore, intra-genic compensation by mutation-skipping isoforms associates with differential isoform usage among proximal interactors, suggesting restorative protein interaction rewiring at isoform-level. Mechanistically, self-transcriptional adaptation (self-TA) following nonsense-mediated decay of perturbed isoforms provides an explanation for the observed up-regulation of mutation-skipping isoforms. Overall, the proposed recursive mutational robustness framework opens new avenues for deeper mechanistic investigation into cancer evolution and identification of selective vulnerabilities in cancer cells. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=155 SRC="FIGDIR/small/727768v1_ufig1.gif" ALT="Figure 1"> View larger version (46K): org.highwire.dtl.DTLVardef@1b0f79aorg.highwire.dtl.DTLVardef@14fa261org.highwire.dtl.DTLVardef@12da16aorg.highwire.dtl.DTLVardef@c03dee_HPS_FORMAT_FIGEXP M_FIG C_FIG

genomics↗

bioRxiv2026

Systematic identification of chromatin organizers as tuners of intratumoral heterogeneity

Intratumoral transcriptional heterogeneity (ITH) enables tumor cells to explore diverse phenotypic states, fueling therapeutic resistance and metastatic progression. Yet systematic approaches for identifying non-genetic regulators of this phenotype have been lacking. Here, we develop a multi-scale screening framework that integrates bulk patient transcriptomics, functional in vivo genetic screening, and single-cell transcriptomic and epigenomic profiling to nominate chromatin organizers that modulate ITH in breast cancer. From an initial set of 41 candidates nominated in silico, we identify RNF8 and MIS18A as factors whose perturbation produces coordinated shifts in transcriptomic variability and chromatin accessibility dispersion at the single-cell level, without altering mean expression or genomic copy number. MIS18A acts through a centromere-independent, DNA-methylation-associated pathway. Modulating RNF8 or MIS18A expression tunes chemotherapeutic sensitivity and metastatic potential in vivo: knockdown sensitizes cells and reduces metastatic burden, whereas overexpression confers resistance and promotes progression. Higher expression of both genes predicts shorter patient survival. These findings identify RNF8 and MIS18A as chromatin-level stochastic tuners of transcriptional variability in cancer and provide a generalizable framework for discovering regulators of non-genetic tumor heterogeneity.

cancer biology↗

bioRxiv2026

BioReason-Pro: Advancing Protein Function Prediction with Multimodal Biological Reasoning

Protein function annotation is fundamental to understanding biological mechanisms, designing therapeutics, and advancing biomedical research. Current computational methods either rely on shallow sequence similarity or treat function prediction as isolated classification tasks, failing to capture the integrative reasoning across sequence, structure, domains, and interactions that expert biologists perform to infer function. We introduce BioReason-Pro, the first multimodal reasoning large language model (LLM) for protein function prediction that integrates protein embeddings with biological context to generate structured reasoning traces. A key input into BioReason-Pro is the set of GO term predictions made by GO-GPT, our autoregressive transformer that captures hierarchical and cross-aspect dependencies of GO terms. BioReason-Pro is trained via supervised fine-tuning on synthetic reasoning traces generated by GPT-5 for over 130K proteins and further optimized through reinforcement learning. It achieves 73.6% Fmax on GO term prediction and an LLM judge score of 8/10 on functional summaries, substantially outperforming previous methods. Evaluations with human protein experts show that BioReason-Pro annotations are preferred over ground truth UniProt annotations in 79% of cases. Remarkably, BioReason-Pro predicted a novel interaction partner for the renal cancer biomarker RCDG1, which we confirmed in the lab by co-immunoprecipitation. In other binding-partner predictions, its per-residue attention localized to the exact contact residues resolved in cryo-EM structures. Together, GO-GPT and BioReason-Pro establish a framework for protein function prediction that combines precise ontology modeling with interpretable biological reasoning.

molecular biology↗

bioRxiv2026

Systemic hypoxia suppresses solid tumor growth

Local hypoxia is a hallmark of solid tumors and a negative prognostic factor in the progression and treatment of cancer. Here, we showed that systemic hypoxia, in contrast to localized tumor hypoxia, decreases tumor growth in vivo across multiple cancer types and preclinical models. The reduced tumor growth in systemic hypoxia was not explained by hypoglycemia, hypoinsulinemia, or HIF activation. Instead, metabolite profiling in tumors and tumor interstitial fluid revealed extensive perturbations in purine-related metabolites. Stable isotope tracing demonstrated that systemic hypoxia caused tumors to suppress de novo purine synthesis. Furthermore, tumors did not develop resistance to systemic hypoxia therapy, and when used in combination with chemotherapy or immunotherapy, systemic hypoxia dramatically suppressed tumor growth. Finally, we showed that systemic hypoxia can be achieved pharmacologically with the small molecule HypoxyStat. These findings challenge the long-held paradigm of hypoxia as a negative prognostic factor in cancer progression, and they suggest a potential therapeutic role for systemic hypoxia in suppressing solid tumor growth.

cancer biology↗

bioRxiv2026

Scalable probe-based single-cell transcriptional profiling for virtual cell perturbation mapping and synthetic biology phenotyping

Large-scale single-cell transcriptional phenotyping of genetic perturbations (perturb-seq) links genes to phenotypes and should enable virtual cell predictive modeling and cellular engineering. However, current perturb-seq single-cell methods are costly, information sparse and require barcodes for many applications. We developed ProPer-seq, a perturb-seq method that uses multiplexed custom DNA probe panels to measure and phenotype synthetic biology perturbations at single-cell resolution without barcodes, including multidomain proteins and sgRNAs. ProPer-seq faithfully reproduces gold-standard perturb-seq phenotypes while achieving 4-fold cost reduction and 50% increased gene detection per cell. As a scalable fixed-cell profiling method, ProPer-seq enables atlas-scale profiling for virtual-cell initiatives and demonstrates data quality suitable for training and validating predictive models. Lastly, ProPer-seqs targeted detection of modular transgenes enables library-on-library perturbation profiling of combinatorial synthetic protein design spaces. We applied this to 3,550 sgRNA x dCas9 effector combinations as well as 260 CAR x ORF combinations dynamically profiled in primary T cells, revealing principles of transcriptional control and cell state modulation by multidomain synthetic transgenes.

synthetic biology↗

bioRxiv2025

Orphan Non-Coding RNAs Drive Tumorigenesis and Enable Pre-Diagnostic Detection in HPV-Negative Head and Neck Cancer

HPV-negative head and neck squamous cell carcinoma (HNSCC) lacks reliable biomarkers for early detection and molecular stratification. We recently described orphan non-coding RNAs (oncRNAs) as a class of cancer-emergent small RNAs that arise from cryptic promoters largely silent in normal tissues. Here, we define the comprehensive landscape of oncRNAs in HNSCC. Using TCGA and CPTAC-3 data, we demonstrate that these transcripts are not only tumor-specific but provide digital fingerprints that accurately stratify tumors by anatomical site. Moving from association to function, we investigated whether these emergent transcripts actively drive tumorigenesis. We focused on oncRNA64585, a transcript highly enriched in oral cavity tumors and associated with tumor progression. Its depletion significantly impaired proliferation in vitro and suppressed tumor growth in syngeneic murine models. Mechanistically, transcriptomic and chromatin accessibility profiling revealed that oncRNA64585 regulates transcriptional programs governing cell cycle progression and apoptosis through interaction with the Polycomb Repressive Complex 2 (PRC2). Finally, we demonstrate the clinical utility of these signatures: oncRNA profiles in plasma predicted oral cavity cancer with >90% accuracy up to four years prior to clinical diagnosis. These findings establish oncRNAs as active drivers of HNSCC biology and potent tools for early detection and therapeutic targeting.

cancer biology↗

bioRxiv2025

Tahoe-x1: Scaling Perturbation-Trained Single-CellFoundation Models to 3 Billion Parameters

Foundation models have transformed natural language processing and computer vision, yet their potential in single-cell biology--particularly for complex diseases such as cancer-- remains underexplored. We present Tahoe-x1 (Tx1), a family of perturbation-trained single-cell foundation models with up to 3 billion parameters. Tx1 is pretrained on large-scale single-cell transcriptomic datasets, including the Tahoe-100M perturbation compendium, and fine-tuned for cancer-relevant tasks. Through architectural optimizations, data loader refinements, and efficient training strategies, Tx1 achieves 3-30x higher compute efficiency than prior implementations of cell-state models. Tx1 jointly learns representations of genes, cells, and compounds using a masked-expression generative objective that incorporates a drug token, enabling flexible adaptation to diverse downstream applications. We evaluate Tx1 across four key disease-relevant benchmarks: (1) prediction of overall and context-specific gene essentiality, (2) identification of genes contributing to the hallmarks of cancer, (3) cell-type classification, and (4) prediction of perturbation responses in held-out cellular contexts. Tx1 achieves state-of-the-art performance across all tasks. We release pretrained checkpoints, training code, and evaluation workflows to accelerate the development of perturbation-trained single-cell foundation models for applications in precision oncology and beyond. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=52 SRC="FIGDIR/small/683759v1_ufig1.gif" ALT="Figure 1"> View larger version (13K): org.highwire.dtl.DTLVardef@aa33d5org.highwire.dtl.DTLVardef@310a38org.highwire.dtl.DTLVardef@195f204org.highwire.dtl.DTLVardef@143f7e3_HPS_FORMAT_FIGEXP M_FIG C_FIG

systems biology↗

bioRxiv2025

An image-based CRISPR screen reveals splicing-mediated control of HP1α condensates

Heterochromatin Protein 1 (HP1) is a fundamental component of constitutive heterochromatin, forming subnuclear condensates whose regulation and function remain poorly understood. Here, we present an image-based CRISPR screen targeting nuclear factors that identifies splicing as a pivotal pathway regulating HP1 condensates. We discovered that unspliced intronic RNA modulates HP1 condensates by interacting co-transcriptionally with HP1. By modulating the intron content, RNA processing restricts HP1-RNA interactions at chromatin, thus enabling heterochromatin organization. Disruption of HP1 condensates due to enhanced interactions with unspliced RNA leads to loss of heterochromatin and the activation of stress response protective genes. We propose that RNA is a central component of heterochromatin that modulates HP1 condensates, and that RNA processing enzymes act as a surveillance mechanism for condensates by dynamically regulating the network of multi-valent interactions between RNA and chromatin factors. This model underscores the crosstalk between chromatin organization, transcription, and RNA processing, potentially governing broader nuclear functions.

genetics↗

bioRxiv2025

Predicting cellular responses to perturbation across diverse contexts with STATE

Cellular responses to perturbations are a cornerstone for understanding biological mechanisms and selecting drug targets. While machine learning models offer tremendous potential for predicting perturbation effects, they currently struggle to generalize to unobserved cellular contexts. Here, we introduce SO_SCPLOWTATEC_SCPLOW, a transformer model that predicts perturbation effects while accounting for cellular heterogeneity within and across experiments. SO_SCPLOWTATEC_SCPLOW predicts perturbation effects across sets of cells and is trained using gene expression data from over 100 million perturbed cells. SO_SCPLOWTATEC_SCPLOW improved discrimination of effects on large datasets by more than 30% and identified differentially expressed genes across genetic, signaling and chemical perturbations with significantly improved accuracy. Using its cell embedding trained on observational data from 167 million cells, SO_SCPLOWTATEC_SCPLOW identified strong perturbations in novel cellular contexts where no perturbations were observed during training. We further introduce Cell-Eval, a comprehensive evaluation framework that highlights SO_SCPLOWTATEC_SCPLOWs ability to detect cell type-specific perturbation responses, such as cell survival. Overall, the performance and flexibility of SO_SCPLOWTATEC_SCPLOW sets the stage for scaling the development of virtual cell models.

systems biology↗