Biology subjects

Kundaje, A.

Publications and source records attributed to Kundaje, A..

At least 19 recordsLinked to original sources

bioRxiv2026

SMFnet reveals principles of combinatorial transcription factor assembly

Combinatorial assembly of transcription factors (TFs) opens chromatin at cis-regulatory elements (cREs) to establish cell-type-specific gene expression. Dissecting the role of individual TFs and their collective function has been limited by the diversity of TF binding-site arrangements across cREs. A single regulatory sequence can support multiple configurations of TF and nucleosome occupancy, raising the possibility that distinct chromatin states reflect different TF assemblies. Predictive models linking regulatory sequence to chromatin accessibility collapse these distinct molecular states into population averages, obscuring how TF assemblies shape accessible chromatin. Here, we introduce SMFNet, a generative diffusion model that predicts distributions of single-molecule chromatin accessibility profiles from DNA sequence, and PRISM, which identifies sequence features associated with individual molecular states. Together, these approaches enable us to systematically associate single molecule chromatin states at cREs with specific TF binding motifs. Experimental perturbation of these motifs redistributes the corresponding states as predicted by SMFNet, supporting their interpretation as TF-dependent configurations underlying chromatin opening. We then use controlled in silico motif screens to systematically dissect individual and combinatorial TF contributions to chromatin opening. Individual TFs show distinct chromatin-opening signatures, differing in accessible-region width and nucleosome-positioning precision even when associated with similar levels of chromatin accessibility. Combining motifs reveals partner- and spacing-dependent effects on chromatin accessibility, TF occupancy and nucleosome positioning. Genetic perturbation of TF binding sites using natural variation at endogenous cRE confirms that TF-specific properties shape chromatin accessibility in vivo. Together, our findings reveal sequence-encoded principles of combinatorial TF assembly during chromatin opening.

genetics↗

bioRxiv2026

A comprehensive map of the sequence basis of transcription initiation in humans

Transcription initiation is a key marker of active enhancers and promoters. Here we describe an atlas of efficient and interpretable deep learning models trained on over 200 transcription initiation profiling experiments representing all major human organs and many cancer types. We show that our models accurately predict initiation patterns from local DNA sequence and decode their sequence determinants across diverse cellular contexts.

bioinformatics↗

bioRxiv2026

Systematic mapping of enhancers, silencers, and topological elements with prime editing deletion screens

Noncoding regulatory elements in the human genome can control gene expression through distinct mechanisms, including acting as enhancers, silencers, or topological elements. However, we lack tools to identify all such classes systematically and quantify their effects on gene expression in an endogenous genomic context. Here we develop Swap-seq, a method that combines pooled twin prime editing with flow-sorting to delete hundreds of endogenous regulatory sequences and measure their quantitative effects on the expression of a nearby gene. We apply Swap-seq to dissect the regulatory elements that control PPIF expression in THP-1 monocytes. We quantify the effects of distal and intronic enhancers, characterize CTCF insulators, and discover a silencer that regulates distal genes through histone deacetylase (HDAC)-mediated repression of neighboring enhancers. Hundreds of other genomic elements share its chromatin signatures and transcription factor binding sites, suggesting that this silencer represents a broader class of repressive elements. We leverage Swap-seq data to benchmark sequence-to-function models and show that they often capture the local activity of regulatory elements but fail to link them to distal target genes. Swap-seq thus provides a generalizable tool to discover regulatory elements across mechanistic classes, characterize their effects on gene expression, and generate large-scale datasets for evaluating sequence-to-function models.

genomics↗

bioRxiv2026

Mapping and rewiring the MYBPC3 promoter for rescue of haploinsufficiency driven hypertrophic cardiomyopathy

Autosomal dominant loss-of-function variants in the gene MYBPC3 are, collectively, the most common genetic cause of hypertrophic cardiomyopathy (HCM) and are a prototype of haploinsufficient human disease. Typical for haploinsufficiency-associated genes, hundreds of unique loss-of-function pathogenic variants have been reported for MYBPC3 - therapeutic gene editing to correct each of these variants poses major regulatory and logistical hurdles. Upregulating wild-type allele expression could offer a generalizable therapeutic strategy, but the capacity to modulate native MYBPC3 transcription is unknown. Here, we present a variant-agnostic approach to rescue haploinsufficiency by mapping and rationally redesigning the MYBPC3 promoter. Using massively parallel reporter assays (MPRAs) in human induced pluripotent stem cell-derived cardiomyocytes, we performed saturation mutagenesis of the MYBPC3 promoter at single base-pair resolution. This defined a new class of clinically relevant noncoding loss-of-function variants while revealing essential cis-regulatory grammar anchored by key transcription factor binding sites (TFBSs). Furthermore, by systematically screening thousands of variant combinations, modular promoter elements, and heterologous TFBS insertions, we identified synergistic sequence edits that drive robust increases in MYBPC3 expression. Together, our findings improve the clinical interpretation of noncoding variants and establish a scalable blueprint for promoter editing to treat MYBPC3-associated HCM and other haploinsufficient diseases.

molecular biology↗

bioRxiv2026

A Cytokine Receptor Signaling Atlas Reveals How STAT Mosaics Fine-Tune T Cell Function

The extent to which JAK/STAT cytokine signaling is functionally redundant or selective remains debated. Here we engineered a double orthogonal IL-2/IL-2R{beta}/{gamma}c ternary system enabling programmable, interference-free activation of each of the 36 mammalian cytokine receptors, and their downstream six STATs, in T cells. At the membrane-proximal level, comprehensive phospho-signaling profiling revealed that while each receptor activates a dominant STAT, unique STAT activation fingerprints derived from combinatorial biases fine-tune nuanced T cell fates. At the membrane-distal level, single-cell transcriptomic atlas of all cytokine receptors confirmed that these STAT mosaics sensitively specify non-redundant transcriptional programs. STAT5-dominant receptors drove proliferative expansion at the expense of stemness; STAT3-driven programs instructed a continuum from stem cell memory to terminal effector states with preserved cytotoxic capacity and mediated superior curative antitumor responses; while other STATs specified highly restricted phenotypes. These findings decode a STAT signaling vocabulary that defines the intrinsic functional bandwidth of natural cytokines.

immunology↗

bioRxiv2026

Architectural chromatin interactions provide a framework for context-dependent gene regulation

Gene regulation depends on coordinated interactions between promoters and distal cis-regulatory elements, yet understanding how these regulatory elements communicate remains a fundamental challenge in mammalian genomics. Chromatin interaction assays provide one approach for identifying potential regulatory relationships, but interpreting the biological significance of individual interactions remains difficult; chromatin interactions comprise multiple biologically distinct classes that are only partially captured by any single assay. Here, we integrate Hi-C, RNAPII ChIA-PET, and CTCF ChIA-PET with the ENCODE Registry of candidate cis-regulatory elements (cCREs) and complementary functional genomic datasets to develop an integrative framework for classifying and interpreting promoter-centric chromatin interactions. Using this framework, we identify a distinct class of candidate architectural promoter-enhancer interactions that are characterized by increased recurrence across cellular contexts, broader promoter connectivity, and reduced dependence on linear genomic proximity. We further show that many regulatory elements anchoring these interactions transition between enhancer and CTCF-only states while maintaining stable chromatin interactions. These dual-state regulatory elements also acquire context-specific transcription factor inputs within evolutionarily conserved architectural scaffolds, suggesting that stable chromatin architecture can be repeatedly repurposed for new regulatory functions. Genes connected to these dual-state regulatory elements are enriched for developmental and signaling pathways and exhibit increased expression specificity across cell types, consistent with specialized roles in context-dependent gene regulation. Together, our findings provide a biologically informed framework for classifying and interpreting chromatin interactions and support a model in which conserved chromatin architecture provides a stable foundation upon which new regulatory programs evolve.

genomics↗

bioRxiv2026

Massively parallel characterization and predictive modelling of neuronal regulatory variation

Disease-associated variants reside frequently in noncoding cis-regulatory elements (CREs), yet their functional consequences remain poorly understood. We performed a large-scale lentiMPRA in human excitatory neurons, quantifying the impact of >46,000 naturally occurring variants across >27,000 candidate CREs near 524 disease-associated genes. These data improved regulatory variant effect predictions beyond state-of-the-art models. Significant allelic effects occurred at comparable rates across common, rare, and singleton variants, demonstrating that, within MPRA-measurable effects, population frequency carries limited information about per-variant regulatory impact. Variant effect detectability and magnitude were governed primarily by baseline activity of the enclosing regulatory element and local sequence context. Regulatory effects were distributed across numerous transcription factors rather than concentrated in master regulators, consistent with a combinatorial enhancer architecture. We establish a large-scale functional variant catalog and provide a complementary benchmark and resource for developing and evaluating models of noncoding regulatory variation.

genetics↗

bioRxiv2026

BenchRep-T: A Systematic Evaluation of T-Cell Repertoire-Based Disease Diagnostics

Adaptive immune receptor repertoire sequencing data has emerged as a promising potential modality for disease diagnosis, relying on computational methods to analyze T-cell receptor (TCR) sequences from an individuals blood sample. Published methods rely on different cohorts, data preprocessing pipelines, and evaluation metrics, making direct comparison across methods challenging. We present BenchRep-T, a unified benchmark that standardizes multiple publicly available TCR repertoire datasets and evaluates nine computational approaches, spanning statistical enrichment of shared sequences, feature-engineered ensembles, deep learning, and sequence clustering. BenchRep-T evaluates methods on four tasks: disease classification across conditions, performance scaling under restricted sequence-sampling depth, recovery of known antigen-specific driver sequences, and evaluation of sensitivity to demographic confounding. Under controlled evaluation, simple baselines prove competitive, with tree-based models trained on V- and J-gene usage and short sequence motifs approaching the classification performance of more complex methods. Our findings underscore the complexity of modeling TCR repertoire data, and show that no single method dominates across all tasks. BenchRep-T provides a framework for rigorous and reproducible evaluation of TCR repertoire classification methods to accelerate the development of immune repertoire-based diagnostics.

immunology↗

bioRxiv2026

Amplicon/Protein Bead Display enables quantitative in vitro biochemistry at scale

Scalable methods for producing and characterizing protein libraries are essential for generating standardized datasets needed to train models linking sequence to function. Here, we present Amplicon/Protein Bead Display (APB-Display), which expresses and purifies >100,000 protein variants in vitro in 1 day. APB-Display uses particle-templated emulsification to generate libraries of hydrogel beads that covalently display many copies of a given protein variant and its encoding DNA. By incubating beads with fluorescently-labeled ligands, sorting, and sequencing to generate titration curves (APB-TiteSeq), we simultaneously quantified expression levels and binding affinities (Kds) for >18,000 FLAG epitope variants interacting with M2 anti-FLAG antibody in 3 days, revealing chemistry-dependent epistasis between positions 4 and 5. Single-concentration binding measurements (APB-SortSeq) paired with neural network denoising further returned quantitative affinities for >88,000 variants. APB-Display requires only standard laboratory equipment and access to a FACS sorter, providing an accessible platform for quantitative in vitro biochemistry at scale.

biophysics↗

bioRxiv2026

Massively multiplex multimodal chemical screens at single-cell resolution

Recent applications of scRNA-seq for massively multiplexed chemical screens have enabled comprehensive profiling of drug responses at unprecedented scale and resolution. However, current assays remain limited to RNA readouts, lacking information on other phenotypic and mechanistic layers such as chromatin accessibility, protein abundance and post-translational modifications. Here, we introduce a scalable framework for multimodal chemical screens, combining parallel small-molecule perturbations with multimodal readouts. We extend existing experimental platforms into icCITE-plex and DOGMA-plex, enabling joint profiling of RNA, protein, and epigenomic responses to thousands of chemical perturbations in parallel. To systematically decode the regulatory circuitry underlying these responses, we develop MoCAVI, a contrastive analysis framework that disentangles the effect of small molecules from control variation in multimodal measurements, and PERCISTRA, a pipeline that infers causal links between chromatin accessibility and gene expression. Applied across ~410,000 primary T cells under ~2,800 conditions, our approach resolves compound-specific mechanisms, highlights off-target effects, and links chromatin accessibility changes to transcription factor networks in primary T cells. Our results establish a generalizable platform for profiling and analyzing cellular responses to chemical perturbations across multiple modalities.

genomics↗

bioRxiv2026

Bio-BLIP: A Multimodal Architecture for Transferable Reasoning in Genomic Variant Interpretation

Developing scientific hypotheses in biology requires integrating heterogeneous evidence across DNA sequence, gene context, protein function, and prior literature. Existing multimodal AI systems expose biological evidence to reasoning models through textification or by projecting biological embeddings into fine-tuned language models. However, these models are typically highly optimized the specific set of tasks for which they are fine-tuned. Here we present Bio-BLIP, a multimodal Q-former based architecture which leverages biological embeddings and a LLM to generalize to complex reasoning tasks without task-specific fine-tuning. The key to Bio-BLIP is a new neural network architecture that integrates four data modalities - DNA, genes, proteins, and text - through a master Qformer model, which integrates the modality-specific information into a fixed-length prefix for the LLM backbone. Bio-BLIP is pretrained on the task of human genetic variant annotation and achieves a 29.8% increase in generating accurate variant features over frontier LLMs. We evaluate Bio-BLIP zero-shot on downstream genomic tasks of variant prioritization and target gene prediction. Bio-BLIP outperforms two alignment-free genomic language models on regulatory variant prioritization for Mendelian disease. Across the target gene prediction task, Bio-BLIP improves accuracy over LLMs by leveraging learned genomic variant knowledge in difficult cases. Our model produces rich, transparent reasoning traces. In biological domains characterized by multiple scales of data and varied downstream tasks, Bio-BLIP offers a step toward natively multimodal, generalizable reasoning.

genomics↗

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↗

bioRxiv2026

Deep learning-guided design of cell type-specific AAV promoters

Precise cell type targeting is critical for both clinical and experimental applications of adeno-associated viral (AAV) vectors, yet engineering vectors with cell type-specific activity remains a challenge. Here, we compared three strategies leveraging single-cell chromatin accessibility data to design cell type-specific AAV promoters, including a deep learning-based method to generate de novo regulatory sequences. When applied to target retinal ganglion cells or horizontal cells in mouse retina, deep learning-guided design consistently outperformed rational approaches, yielding synthetic promoters with stronger and more specific expression in vivo. Synthetic AAV promoters supported diverse transgenes, enabling the recording and ablation of targeted cells. Promoter activity was also maintained in human retinal organoids, suggesting that deep learning-designed sequences may be suitable for translation. Our findings highlight the potential of deep learning to synthesize cell type-specific AAV promoters and establish a versatile platform for cell type targeting with broad implications for gene therapy and basic research.

synthetic biology↗

bioRxiv2025

Deep learning the dynamic regulatory sequence code of cardiac organoid differentiation

Defining temporal gene regulatory programs driving human organogenesis is essential for understanding congenital defects. We combined a time-resolved, single-cell multi-omic atlas of human induced pluripotent stem cell-derived cardiac organoids with deep learning models of chromatin accessibility, enabling systematic discovery of cis-regulatory syntax underlying heart development and disease. This framework identified cell-state-specific motif syntax, linked motif instances to candidate target genes, and resolved programs governing lineage divergence. Integrating cell-state-resolved molecular profiles with computationally predicted variant effects from congenital heart disease (CHD) cases enabled the prioritization of noncoding variants predicted to disrupt developmental transitions, supporting the paradigm that disease etiology derives from perturbations to regulatory networks governing cardiogenesis. Experimental validation demonstrated that an intronic ANGPTL2 variant altered differentiation outcomes, implicating ANGPTL2 in CHD. This study bridges developmental regulation with disease genetics, establishing a framework for discovering the genetic and molecular basis of congenital disorders.

genetics↗

bioRxiv2025

Sensitive, direct detection of non-coding off-target base editor unwinding and editing in primary cells

Base editors create precise nucleotide changes in DNA, but their off-target activity remains challenging to quantify. Here, we develop and deploy a direct, in cellulo sequencing assay that simultaneously measures both Cas9-mediated unwinding and deaminase editing of genomic DNA (beCasKAS). Our strategy nominates >460-fold more potential off-target sites than other methods by enriching for Cas9-dependent R-loops immediately preceding editing. Using beCasKAS in primary human T-cells, we observe that mRNA-encoded ABE8e and PAMless ABE8e-SpRY base editors have distinct off-target profiles that can be mitigated by optimizing mRNA dose. Finally, we combine beCasKAS with base-resolution deep learning models to risk-stratify off-target edits by their likelihood of epigenetic dysregulation. Collectively, beCasKAS offers a sensitive and facile tool to optimize the balance between base editor on- and off-target activity.

genomics↗

bioRxiv2025

Polyclonal origins of human premalignant colorectal lesions

Cancer is generally thought to be caused by expansion of a single mutant cell1. However, analyses of early colorectal cancer lesions suggest that tumors may instead originate from multiple, genetically distinct cell populations2,3. Detecting polyclonal tumor initiation is challenging in patients, as it requires profiling early-stage lesions before clonal sweeps obscure diversity. To investigate this, we analyzed normal colorectal mucosa, benign and dysplastic premalignant polyps, and malignant adenocarcinomas (123 samples) from six individuals with familial adenomatous polyposis (FAP). Individuals with FAP have a germline heterozygous APC mutation, predisposing them to colorectal cancer and numerous premalignant polyps by early adulthood4. Whole-genome and/or whole-exome sequencing revealed that many premalignant polyps--40% with benign histology and 28% with dysplasia--were composed of multiple genetic lineages that diverged early, consistent with polyclonal origins. This conclusion was reinforced by whole-genome sequencing of single crypts from multiple polyps in additional patients which showed limited sharing of mutations among crypts within the same lesion. In some cases, multiple distinct APC mutations co-existed in different lineages of a single polyp, consistent with polyclonality. These findings reshape our understanding of early neoplastic events, demonstrating that tumor initiation can arise from the convergence of diverse mutant clones. They also suggest that cell-intrinsic growth advantages alone may not fully explain tumor initiation, highlighting the importance of microenvironmental and tissue-level factors in early cancer evolution.

cancer biology↗

bioRxiv2025

Cancer cell phagocytosis induces an anti-inflammatory gene regulatory program in macrophages

Macrophages are capable of eliminating cancer cells by phagocytosis, particularly in the presence of monoclonal antibody (mAb) therapies targeting tumor antigens. Paradoxically, tumor-associated macrophages are typically associated with poor patient outcome, and can promote tumor growth by secretion of immunosuppressive cytokines and growth factors. The mechanisms by which these pro-tumor macrophage states arise are poorly understood, and it is unclear how mAb-induced cancer cell phagocytosis may contribute to these states. To understand how antibody-dependent cancer cell phagocytosis (ADCP) alters macrophage state and function, we profiled gene expression and chromatin accessibility changes over time after ADCP. We observed that after ADCP, macrophages upregulate an anti-inflammatory gene regulatory program, characterized by expression of pro-angiogenic and immunosuppressive chemokine genes, and increased activity by cellular, oxidative, and lysosomal stress transcription factors. This gene regulatory program was shared among phagocytic macrophages following either ADCP or apoptotic cancer cell phagocytosis, in addition to substrate-specific pathways. Conditioned media from macrophages promoted EMT in cancer cells, but this pro-EMT macrophage phenotype was attenuated following ADCP, but not following apoptotic cancer cell phagocytosis. The phagocytic gene signature we identified in vitro is also expressed by tumor-associated macrophages across numerous cancer types in vivo. Together, this work identifies an anti-inflammatory and immunosuppressive epigenetic program in macrophages following ADCP upon mAb treatment, and expands our understanding of how phagocytosis influences macrophage heterogeneity in the tumor microenvironment.

cancer biology↗

bioRxiv2025

The human transcription factor occupancy landscape viewed using high-resolution in situ base-conversion strand-specific single-molecule chromatin accessibility mapping

Chromatin accessibility profiling is a key tool for mapping the location of cis-regulatory elements (cREs) in the genome and tracking chromatin state dynamics during development, in response to various external and internal stimuli, and in disease contexts. Single-molecule footprinting (SMF) methods that rely on the labeling of individual accessible DNA bases have emerged in recent years as a powerful chromatin accessibility mapping approach, as they provide not just an average readout of accessibility over a given genomic position but also the distribution of accessibility states within a population at the level of individual original DNA molecules. However, methylation-based SMF approaches have often been limited either in their resolution or in labeling readout accuracy. We developed a deaminase-based high-resolution strand-specific single molecule footprinting approach that uses highly active sequence context-independent endogenous methylation-insensitive double-strand DNA (dsDNA) deaminases (CseDa01 and LbDa02 C[->]U/T-ssSMF), which convert accessible cytosines into uracils (which are converted to thymine after PCR amplification). We demonstrate the application of the method to mapping single-molecule accessibility states in human cell lines in both a short and a long-read format, and quantifying the occupancy states of individual transcription factors (TFs) as well as TF co-accessibility and strand-specific accessibility patterns.

genomics↗