Biology subjects

Bang, S.

Publications and source records attributed to Bang, S..

At least 19 recordsLinked to original sources

bioRxiv2026

Microbial-associated acylated putrescines as immunomodulatory molecules in inflammatory bowel diseases

The gut microbiota regulates intestinal immunity through metabolite production, yet most disease-associated metabolites remain functionally uncharacterized. In inflammatory bowel diseases (IBD), where the microbial metabolome is profoundly altered, we identify N-acyl putrescines as microbiome-associated metabolites enriched across two independent cohorts. N-oleoylputrescine (NOP) emerges as the primary immunomodulatory candidate, inducing robust transcriptional responses in dendritic cells and colonic organoids. Enterocloster species harboring nonribosomal peptide synthetase gene clusters synthesize NOP, confirmed by isotope-tracing in vitro and germ-free mouse colonization in vivo. NOP suppresses core IBD inflammatory pathways in mouse dendritic cells and human monocytes, reducing signatures of histologic inflammation and therapy non-response. NOP dampens inflammation in four colitis models, decreases myeloid cell NF-{kappa}B activation, and suppresses type 1 immune responses through a T cell-intrinsic mechanism. That NOP accumulates in IBD despite its anti-inflammatory properties reveals a holobiont defense strategy: the gut microbiota deploys immunomodulatory metabolites as a compensatory response to restore homeostasis.

immunology↗

bioRxiv2026

Collective cancer cell calcium activity drives brain metastasis

Communication in multicellular networks is a cancer-intrinsic neural feature and crucial for primary brain tumor growth and resistance, but it is unclear whether brain metastases (BrM), the most common and deadliest brain malignancy, are also driven by communicating cancer networks. Using intravital two-photon microscopy in awake mice, clinical specimens, and Ca2+ integrators, we demonstrate that brain-colonizing breast and lung cancer and melanoma cells display gap-junction-dependent, coordinated Ca2+ activity in multicellular, cancer-cell intrinsic networks, which drives their proliferation. Mechanistically, Ca2+ oscillations induce transcription of immediate early genes, adoption of a neuronal expression profile, and cell cycle progression. While many of those features are enriched in BrM, all investigated cancer cell lines showed collective Ca2+ activity. Therapeutically, blocking Ca2+ activity with gap junction inhibitors reduces BrM burden in mouse models. Here we show communicating cancer cell syncytia as drivers of BrM growth, pointing to a targetable pathomechanism, and potentially a new pan-cancer hallmark. Graphical AbstractIn brief Brain metastases form gap-junction-coupled networks exhibiting spontaneous, coordinated Ca2+ activity linked to immediate early gene activation, neuronal gene programs, and cell cycle progression. Disrupting Ca2+ network communication with gap junction inhibitors induces cell cycle arrest and reduces brain metastatic burden in vivo. O_FIG O_LINKSMALLFIG WIDTH=195 HEIGHT=200 SRC="FIGDIR/small/723715v1_ufig1.gif" ALT="Figure 1"> View larger version (88K): org.highwire.dtl.DTLVardef@1130ea0org.highwire.dtl.DTLVardef@254dcdorg.highwire.dtl.DTLVardef@117021forg.highwire.dtl.DTLVardef@18c4f1e_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsO_LIBrain metastases display collective Ca2+ activity in gap-junction-coupled networks C_LIO_LICa2+ co-activity is conserved across cancers but enriched in brain metastasis C_LIO_LICa2+ oscillations robustly induce neuronal gene programs and cell cycle progression C_LIO_LIGap junction inhibition reduces brain metastasis burden in mice C_LI

cancer biology↗

bioRxiv2026

Generative AI designs functional thiolation domains for reprogramming non-ribosomal peptide synthetases

Large language models and generative protein design promise to accelerate biotechnology, but it remains unclear whether they can engineer dynamic megasynth(et)ases whose activity depends on transient, context-specific domain interfaces. Non-ribosomal peptide synthetases (NRPSs) are an especially demanding target, yet a high-value one because they produce many clinically important natural products and offer a route to analogs that are often difficult or impractical to access by chemical synthesis. Here we integrate pretrained generative models (ESM3, ProteinMPNN and EvoDiff) with design-build-test-learn cycles and data-guided prioritization to generate 76 de novo thiolation (T) domains. We built and tested 578 recombinant NRPS variants in vivo spanning minimal, full-length and hybrid assembly lines. AI-designed T-domains supported product formation across architectures, enabled catalytically active hybrids at recombined junctions and increased yields by up to [~]3-fold relative to NRPSs carrying the native T-domain. A representative design showed improved soluble expression, refolding, and a 12 {degrees}C higher melting temperature, while molecular dynamics simulations indicated preserved global stability but reshaped, state-dependent interdomain contact networks. Together, these results establish generative design as an effective route to context-conditioned optimization and reprogramming of biosynthetic assembly lines.

synthetic biology↗

bioRxiv2025

Multiparametric Assessment of TNNI3 Variant Phenotypes in Human iPSC-Cardiomyocytes Correlates with Disease Severity in Patients

BackgroundThe routine genetic testing of cardiomyopathy patients has significantly accelerated the identification of causative cardiomyopathy variants. However, translating these genetic insights into effective patient management poses significant challenges, since the impact of gene variants on physiological function and clinical outcomes is not yet fully understood. Therefore, there is an urgent need for large-scale methods to assess the effects of genetic variants on cardiomyocyte physiology and to establish correlations between functional phenotypes and clinical severity. MethodsWe developed a high throughput imaging platform to measure force generation and calcium handling throughout the cardiac cycle of human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs). By expressing variants of a sarcomeric protein [cardiac Troponin-I (TNNI3)] in a healthy genetic background, we were able to assess sarcomeric calcium sensitivity as well as systolic and diastolic function. Analysis of these parameters distinguished subgroups of variants, and permitted the correlation of in vitro physiological effects with a measure of disease severity in a single-center cardiomyopathy cohort. ResultsCombining contractile force and calcium cycling measurements accurately distinguished known pathogenic from non-pathogenic TNNI3 variants and also revealed pathogenicity of two variants of unknown significance (VUS) that occurred in two families, suggesting the ability to prospectively discern pathogenicity. Clustering of TNNI3 variants based on quantitative physiological phenotypes identified subgroups that correlated with age of disease onset across a well-characterized cardiomyopathy patient cohort, showing clinical relevance of the in vitro phenotypes. Interestingly, normalized measures of in vitro diastolic function correlated with age of onset (R2 = 0.6), but calcium sensitivity, which accurately predicted pathogenicity, did not translate into disease severity. ConclusionsA high throughput in vitro platform that measures multidimensional cardiomyocyte function can link subgroups of human genetic variants in TNNI3 with differential patient outcomes. Comprehensive determination of variant effects on disease-relevant cardiomyocyte function will help classify variants into different pathogenic mechanisms leading to variable disease severity, and potentially lead to class-targeted ameliorative strategies.

genetics↗

bioRxiv2025

p53 and YAP/TAZ-TEAD activities determine metaplastic heterogeneity in pancreatic cancer

BACKGROUND and AIMSPancreatic ductal adenocarcinoma (PDAC) initiation and progression is characterized by lineage plasticity across a continuum of metaplastic and malignant cell fates. This molecular heterogeneity influences disease progression and therapeutic response, yet the genetic and molecular interactions underlying cellular plasticity during PDAC progression remain poorly understood. Here we studied how p53, the most frequently inactivated tumor suppressor in PDAC, interacts with the YAP/TAZ-TEAD transcriptional effector arm of the Hippo pathway to determine cell fate and neoplastic heterogeneity in PDAC development. METHODSWe characterized differentiation states in mouse PDAC and pre-malignant cells in response to p53 activation or inactivation, respectively, and upon genetic, pharmacological, and physiological modulation of YAP/TAZ-TEAD activity. Differentiation was analyzed in vivo in an orthotopic PDAC model permitting independent, inducible control of p53 and TEAD function. We analyzed the effect of p53 and YAP/TAZ-TEAD activity on cell fate in mouse and human PDAC and analyzed markers of clinically relevant molecular subtypes in human PDAC cell lines treated with small molecule TEAD inhibitors. RESULTSRe-engaging wildtype p53 function in KrasMut pancreatic cancer cells increases TEAD activity through accumulation of its co-activator TAZ. Remarkably, increased YAP/TAZ-TEAD activity broadly suppresses the expression of overlapping markers of gastric pit-like (GPL) differentiation and the classical PDAC subtype that are promoted by p53. Accordingly, p53 and YAP/TAZ-TEAD activity inversely correlate with GPL and classical differentiation during PDAC development and modulation of both pathways determines expression of GPL and classical markers in vitro and in vivo. GPL differentiation is acutely sensitive to YAP/TAZ-TEAD activity including oncogenic perturbations of the Hippo pathway, actin dynamics, cell-to-cell contact, and adaptative YAP/TAZ accumulation in response to Ras/MAPK pathway inhibition. Functional repression of gastric-classical differentiation by TEAD persists in human PDAC where potent TEAD inhibition increases expression of GPL and classical markers. CONCLUSIONSWe find that p53 and YAP/TAZ-TEAD activities determine gastric plasticity during PDAC development highlighting the role that key cancer drivers and therapeutically targetable signaling pathways play in active maintenance of cellular identity during pancreatic cancer development.

cancer biology↗

bioRxiv2025

Protectin DX resolves fracture-induced postoperative pain in mice via neuronal signaling and GPR37-activated macrophage efferocytosis

Protectin DX (PDX) is a member of the superfamily of specialized pro-resolving mediators (SPMs) and exerts anti-inflammatory actions in animal models; but the specific signaling pathways involved in its actions have not been fully elucidated. Here we demonstrate the analgesic actions of PDX in a mouse model of tibial fracture-induced postoperative pain (fPOP) via activation of GPR37. Intravenous early-phase and late-phase treatment of PDX (100 ng/mouse) effectively alleviated fPOP. Compared to protectin D1/neuroprotectin D1, DHA, steroid, and meloxicam, PDX produced superior pain relief. While dexamethasone and meloxicam prolonged fPOP, PDX shortened the pain duration. The analgesic effects of PDX were abrogated in Gpr37-/- mice, which displayed deficits in fPOP resolution. PDX binds GPR37 and induces calcium responses in peritoneal macrophages. LC-MS-MS-based lipidomic analysis revealed that endogenous PDX levels were [~]10-fold higher than PD1 in muscle at the fracture site. PDX promotes macrophage polarization via GPR37-dependent phagocytosis and efferocytosis through calcium signaling in vitro, and it further enhances macrophage viability and efferocytosis in vivo via GPR37. Finally, PDX rapidly modulates nociceptor neuron responses by suppressing C-fiber-induced muscle reflex in vivo and calcium responses in DRG neurons ex vivo and reducing TRPA1/TRPV1-induced acute pain and neurogenic inflammation in vivo. Our findings highlight multiple benefits of PDX to manage postoperative pain and promote perioperative recovery.

neuroscience↗

bioRxiv2025

p53 activity licenses transcriptional regulation by YAP/TAZ-TEAD to shape expression landscapes during tumorigenesis

The tumor suppressor p53 is the most frequently disabled gene in cancer, where its inactivation promotes the transition between transcriptionally distinct pre-malignant and malignant disease. How p53 action shapes these stage-dependent expression landscapes is poorly understood. Using a mouse model where p53 function is restored in advanced pancreatic cancer, we demonstrate that p53 dependent gene expression is determined by regulation of YAP/TAZ-TEAD transcriptional complexes at multiple levels of molecular control. Restoration of p53 activity in pancreatic cancer cells expands transcriptional regulation by reorganizing TEAD binding at newly accessible enhancer landscapes, resulting in distinct TEAD-driven gene regulatory networks when p53 is inactivated versus engaged. Furthermore, p53 potentiates expression of newly licensed TEAD targets by increasing TAZ levels in response to the actin remodeling associated with cell cycle arrest and senescence. Thus, p53 activity controls the qualitative and quantitative output of YAP/TAZ-TEAD, including secretory programs involved in tumor-immune communication and remodeling of the microenvironment. Our work nominates regulation of YAP/TAZ-TEAD via p53 as a mechanism that determines stage dependent transcriptional landscapes during stepwise tumorigenesis.

cancer biology↗

bioRxiv2025

Asymmetric distribution of mitochondrial Ca2+ regulators specifies compartment-specific mitochondrial function and neuronal development

Neuronal polarization is essential for functional compartmentalization, enabling dendritic synaptic integration and axonal action potential generation. While structural differences in mitochondria across compartments have been identified, their functional distinctions remain unclear. Here, we uncovered compartment-specific mitochondrial Ca2+ dynamics and their molecular determinants. In axonal mitochondria, Ca2+ uptake through MCU occurs independently of ER-stored Ca2+ release, with faster matrix Ca2+ clearance than dendritic mitochondria, where Ca2+ uptake predominantly originates from ER Ca2+. The ER-independent mitochondrial Ca2+ uptake in axonal mitochondria is mediated by enriched MCU-regulating proteins, MICU1 and MICU2, while higher NCLX expression facilitates rapid Ca2+ clearance. Moreover, NCLX knockdown, which functionally mimics a mental retardation-associated mutation, caused more significant axonal branching defects compared to dendrites in vivo, aligning with its enrichment in axons. These findings highlight fundamental Ca2+-modulating features and developmental importance of neuronal mitochondria in a compartment-specific manner and reveal the key underlying molecular mechanisms.

neuroscience↗

bioRxiv2025

Self-Contemplating In-Context Learning Enhances T Cell Receptor Generation for Novel Epitopes

Computational design of T cell receptors (TCRs) that bind to epitopes holds the potential to revolutionize targeted immunotherapy. However, computational design of TCRs for novel epitopes is challenging due to the scarcity of training data, and the absence of known cognate TCRs for novel epitopes. In this study, we aim to generate high-quality cognate TCRs particularly for novel epitopes with no known cognate TCRs, a problem that remains under-explored in the field. We propose to incorporate in-context learning, successfully used with large language models to perform new generative tasks, to the task of TCR generation for novel epitopes. By providing cognate TCRs as additional context, we enhance the models ability to generate high-quality TCRs for novel epitopes. We first unlock the power of in-context learning by training a model to generate new TCRs based on both a target epitope and a small set of its cognate TCRs, so-called in-context training (ICT). We then self-generate its own TCR contexts based on a target epitope, as novel epitopes lack known binding TCRs, and use it as an inference prompt, referred to as self-contemplation prompting (SCP). Our experiments first demonstrate that aligning training and inference distribution by ICT is critical for effectively leveraging context TCRs. Subsequently, we show that providing context TCRs significantly improves TCR generation for novel epitopes. Furthermore, we show TCR generation using SCP-synthesized context TCRs achieves performance comparable to, and sometimes surpassing, ground-truth context TCRs, especially when combined with refined prompt selection based on binding affinity and authenticity metrics.

bioinformatics↗

bioRxiv2024

HLA-E and NKG2A Mediate Resistance to M. bovis BCG Immunotherapy in Non-Muscle-Invasive Bladder Cancer

BackgroundBacillus Calmette-Guerin (BCG) is the standard of care treatment for high-risk non-muscle-invasive bladder cancer (NMIBC), yet many patients develop recurrent disease despite evidence of ongoing immune activation. We investigated mechanisms of immune escape in BCG-unresponsive tumors and evaluated the therapeutic potential of targeting the HLA-E/NKG2A axis. MethodsSingle-cell RNA sequencing, spatial immunophenotyping, proteomic profiling, and functional ex vivo assays were performed using tumors and urine samples from patients with BCG-naive and BCG-unresponsive NMIBC. ResultsBCG-unresponsive tumors were enriched for HLA-E-expressing malignant cells compared with BCG-naive tumors. Increased HLA-E expression was associated with enhanced IFN-{gamma} signaling and was induced by IFN-{gamma} stimulation in primary tumor cells and bladder cancer tumor lines. Spatial analyses demonstrated accumulation of NKG2A+ NK and CD8 T cells in proximity to HLA-Ehigh tumor cells, with increased NKG2A:HLA-E interactions in BCG-unresponsive tumors. Despite high expression of cytotoxic mediators, NKG2A+ effector cells displayed impaired degranulation. Blockade of NKG2A with monalizumab restored degranulation of and cytotoxicity by tumor-infiltrating lymphocytes in autologous tumor co-cultures. ConclusionsBCG-unresponsive NMIBC tumors are enriched for HLA-E-expressing tumor cells and NKG2A+ effector lymphocytes, with increased engagement of the HLA-E/NKG2A axis within the tumor microenvironment. These findings identify the HLA-E/NKG2A axis as a therapeutic vulnerability and provide a rationale for clinical evaluation of NKG2A blockade as a bladder-sparing strategy for patients with BCG-unresponsive disease.

cancer biology↗

bioRxiv2024

The genetic architecture of cell-type-specific cis-regulation

AbstractGene expression and complex phenotypes are determined by the activity of cis-regulatory elements. However, an understanding of how extant genetic variants affect cis regulation remains limited. Here, we investigated the consequences of cis-regulatory diversity using single-cell genomics of >0.7 million nuclei across 172 Zea mays (maize) inbreds. Our analyses pinpointed cis-regulatory elements distinct to domesticated maize and revealed how historical transposon activity has shaped the cis-regulatory landscape. Leveraging population genetics principles, we fine-mapped [~]22,000 chromatin accessibility-associated genetic variants with widespread cell-type-specific effects. Variants in TEOSINTE BRANCHED1/CYCLOIDEA/PROLIFERATING CELL FACTOR binding sites were the most prevalent determinants of chromatin accessibility. Finally, integrating chromatin accessibility-associated variants, organismal trait variation, and population differentiation revealed how local adaptation has rewired regulatory networks in unique cellular context to alter maize flowering.

genetics↗

bioRxiv2024

A spatially resolved multiomic single-cell atlas of soybean development

Cis-regulatory elements (CREs) precisely control spatiotemporal gene expression in cells. Using a spatially resolved single-cell atlas of gene expression with chromatin accessibility across ten soybean tissues, we identified 103 distinct cell types and 303,199 accessible chromatin regions (ACRs). Nearly 40% of the ACRs showed cell-type-specific patterns and were enriched for transcription factor (TF) motifs defining diverse cell identities. We identified de novo enriched TF motifs and explored conservation of gene regulatory networks underpinning legume symbiotic nitrogen fixation. With comprehensive developmental trajectories for endosperm and embryo, we uncovered the functional transition of the three sub-cell types of endosperm, identified 13 sucrose transporters sharing the DOF11 motif that were co-up-regulated in late peripheral endosperm and identified key embryo cell-type specification regulators during embryogenesis, including a homeobox TF that promotes cotyledon parenchyma identity. This resource provides a valuable foundation for analyzing gene regulatory programs in soybean cell types across tissues and life stages.

plant biology↗

bioRxiv2024

WUSCHEL-dependent chromatin regulation in maize inflorescence development at single-cell resolution

WUSCHEL (WUS) is transcription factor vital for stem cell proliferation in plant meristems. In maize, ZmWUS1 is expressed in the inflorescence meristem, including the central zone, the reservoir of stem cells. ZmWUS1 overexpression in the Barren inflorescence3 mutant leads to defects in inflorescence development. Here, single-cell ATAC-seq analysis shows that ZmWUS1 overexpression alters chromatin accessibility throughout the central zone. The CAATAATGC motif, a known homeodomain recognition site, is predominantly observed in the regions with increased chromatin accessibility suggesting ZmWUS1 is an activator in the central zone. Regions with decreased chromatin accessibility feature various motifs and are adjacent to AUXIN RESPONSE FACTOR genes, revealing negative regulation of auxin signaling in the central zone. DAP-seq of ZmWUS1 identified the TGAATGAA motif, abundant in epidermal accessible chromatin compared to the central zone. These findings highlight ZmWUS1s context-dependent mechanisms for stem cell maintenance in the inflorescence meristem.

plant biology↗

bioRxiv2024

Vaginal lactobacilli produce anti-inflammatory beta-carboline compounds.

The vaginal microbiome is a lactobacilli-dominant, human-adapted community consistently found in people around the world. Presence of lactobacilli-dominant vaginal microbial communities, apart from Lactobacillus iners, is associated with reduced vaginal inflammation and reduced levels of pro-inflammatory cytokines. Loss of lactobacilli-dominance is associated with inflammatory conditions such as bacterial vaginosis. It remains unclear if health-associated lactobacilli actively promote the anti-inflammatory immune homeostasis of the vaginal mucosa and if so, by which mechanisms. We have identified that Lactobacillus crispatus, a key vaginal bacterial species, secretes a family of {beta}-carboline compounds with anti-inflammatory activity. These compounds suppress NF{kappa}B and interferon signaling downstream of multiple pattern recognition receptors in human macrophages and significantly dampen type I interferon receptor activation. Topical vaginal application of an anti-inflammatory {beta}-carboline compound, perlolyrine, was sufficient to significantly reduce vaginal inflammation in a mouse model of genital herpes infection. Together, we identify a key family of compounds by which vaginal lactobacilli mediate host immune homeostasis and could highlight a new therapeutic avenue for vaginal inflammation.

microbiology↗

bioRxiv2023

Satellite glial GPR37L1 regulates maresin and potassium channel signaling for pain control

G protein coupled receptor 37-like 1 (GPR37L1) is an orphan GPCR and its function remains largely unknown. Here we report that GPR37L1 transcript is highly expressed compared to all known GPCRs in mouse and human dorsal root ganglia (DRGs) and selectively expressed in satellite glial cells (SGCs). Peripheral neuropathy following diabetes and chemotherapy by streptozotocin and paclitaxel resulted in downregulations of surface GPR37L1 in mouse and human DRGs. Transgenic mice with Gpr37l1 deficiency exhibited impaired resolution of neuropathic pain symptom (mechanical allodynia), whereas overexpression of Gpr37l1 in mouse DRGs can reverse neuropathic pain. Notably, GPR37L1 is co-expressed and coupled with potassium channels in SGCs. We found striking species differences in potassium channel expression in SGCs, with predominant expression of KCNJ10 and KCNJ3 in mouse and human SGCs, respectively. GPR37L1 regulates the surface expression and function of KCNJ10 and KCNJ3. We identified the pro-resolving lipid mediator maresin 1 (MaR1) as a GPR37L1 ligand. MaR1 increases KCNJ10/KCNJ3-mediated potassium influx in SGCs via GPR37L1. MaR1 protected chemotherapy-induced suppression of KCNJ13/KCNJ10 expression and function in SGCs. Finally, genetic analysis revealed that the GPR37L1-E296K variant is associated with increased chronic pain risk by destabilizing the protein. Thus, GPR37L1 in SGCs offers a new target for neuropathy protection and pain control.

neuroscience↗

bioRxiv2023

Repair oligodendrocytes demyelinating and disintegrating damaged axons after injury

After a spinal cord injury, axons fail to regrow, which results in permanent loss of function1. This is in contrast with peripheral axons that can regrow efficiently after injury2. These differences are partly due to the different plasticity of myelinating cells, Schwann cells and oligodendrocytes, in these two systems3. The molecular mechanisms underlying this different plasticity remain however poorly understood. Here, we show that the phosphatase Dusp64 is a master inhibitor of oligodendrocyte plasticity after spinal cord injury. Dusp6 is rapidly downregulated in Schwann cells and upregulated in oligodendrocytes after axon injury. Simultaneously, the MAP kinases ERK1/2 are activated and the transcription factor c-Jun is upregulated in Schwann cells5,6, but not in oligodendrocytes. Ablation or inactivation of Dusp6 induces rapid ERK1/2 phosphorylation, c-Jun upregulation and filopodia formation in oligodendrocytes, leading to mechanically-induced, fast disintegration of distal ends of injured axons, myelin clearance and axonal regrowth. Together, our findings provide understanding of the mechanisms underlying the different plasticity of Schwann cells and oligodendrocytes after injury and a method to convert mature oligodendrocytes exhibiting inhibitory cues for axonal regrowth into repair oligodendrocytes reminiscent of repair Schwann cells. We show that repair oligodendrocytes successfully increase the compatibility of the spinal cord environment with axonal regrowth after injury, suggesting a potential use of repair oligodendrocytes as future therapeutic approach to treat spinal cord injuries.

neuroscience↗

bioRxiv2023

Context-Aware Amino Acid Embedding Advances Analysis of TCR-Epitope Interactions

Accurate prediction of binding interaction between T cell receptors (TCRs) and host cells is fundamental to understanding the regulation of the adaptive immune system as well as to developing data-driven approaches for personalized immunotherapy. While several machine learning models have been developed for this prediction task, the question of how to specifically embed TCR sequences into numeric representations remains largely unexplored compared to protein sequences in general. Here, we investigate whether the embedding models designed for protein sequences, and the most widely used BLOSUM-based embedding techniques are suitable for TCR analysis. Additionally, we present our context-aware amino acid embedding models (catELMo) designed explicitly for TCR analysis and trained on 4M unlabeled TCR sequences with no supervision. We validate the effectiveness of catELMo in both supervised and unsupervised scenarios by stacking the simplest models on top of our learned embeddings. For the supervised task, we choose the binding affinity prediction problem of TCR and epitope sequences and demonstrate notably significant performance gains (up by at least 14% AUC) compared to existing embedding models as well as the state-of-the-art methods. Additionally, we also show that our learned embeddings reduce more than 93% annotation cost while achieving comparable results to the state-of-the-art methods. In TCR clustering task (unsupervised), catELMo identifies TCR clusters that are more homogeneous and complete about their binding epitopes. Altogether, our catELMo trained without any explicit supervision interprets TCR sequences better and negates the need for complex deep neural network architectures in downstream tasks.

bioinformatics↗

bioRxiv2022

Identifying transcription factor-DNA interactions using machine learning

Machine learning approaches have been applied to identify transcription factor (TF)-DNA interaction important for gene regulation and expression. However, due to the enormous search space of the genome, it is challenging to build models capable of surveying entire reference genomes, especially in species where models were not trained. In this study, we surveyed a variety of methods for classification of epigenomics data in an attempt to improve the detection for 12 members of the Auxin Response Factor (ARF) binding DNAs from maize and soybean as assessed by DNA Affinity Purification and sequencing (DAP-seq). We used the classification for prediction by minimizing the genome search space by only surveying unmethylated regions (UMRs). For identification of DAP-seq binding events within the UMRs, we achieved 93.54% accuracy, 6.2% false positive, and a 43.29% false negative rate across 12 members of ARFs of maize on average by encoding DNA with count vectorization for k-mer with a logistic regression classifier with up-sampling and feature selection. Importantly, feature selection helps to uncover known and potentially novel ARF binding motifs. This demonstrates an independent method for identification of transcription factor binding sites. Finally, we tested the model built with maize DAP-seq data and applied it directly to the soybean genome and found unacceptably high false positive rates, which accounted for more than 40% across the ARF TFs tested. The findings in this study suggest the potential use of various methods to predict TF-DNA interactions within and between species with varying degrees of success.

bioinformatics↗