About

Arjun Raj, PhD, is a Professor of Bioengineering at the University of Pennsylvania's Perelman School of Medicine. His research focuses on RNA systems biology, particularly related to non-coding RNAs and cancer biology. The Raj lab studies the biology of single cells, aiming to understand the molecular differences in cellular behavior even among genetically identical cells. They develop and utilize experimental tools such as fluorescence microscopy of individual RNA molecules, high throughput sequencing, and molecular biology techniques to make quantitative measurements of cellular behavior. His work contributes to understanding cellular variability and its implications in biological contexts, including cancer.

Research topics

• Biology
• Genetics
• Computational biology
• Computer Science
• Artificial Intelligence
• Cell biology
• Cancer research
• Neuroscience
• Chemistry
• Evolutionary biology
• Bioinformatics
• Medicine

Selected publications

• Environmental Amino Acid Sensing Regulates the Rate of ASC Translation and NLRP3 Inflammasome Assembly

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-01-20 · 1 citations

  articleOpen access

  ABSTRACT The NOD-, LRR-, and pyrin domain-containing protein 3 (NLRP3) inflammasome is a multiprotein signaling complex that triggers pyroptotic cell death and interleukin (IL)-1 family cytokine release during infection and cell injury. Its assembly is driven by the adaptor protein, apoptosis-associated speck-like protein containing a CARD (ASC), whose filamentation forms a supramolecular speck upon NLRP3 activation to amplify inflammasome signaling. While the NLRP3 inflammasome is well appreciated as a sensor of environmental danger and damage, little is known about how homeostatic environmental factors like dietary metabolites regulate its activity. Here, we find that environmental availability of the branched-chain amino acids (BCAAs), leucine, isoleucine, and valine, controls NLRP3 inflammasome assembly. While ASC is typically viewed as a constitutively expressed, unregulated inflammasome component, we find that Toll-like receptor 4 (TLR4) activation triggers localization of ASC mRNA to the perinuclear space. Moreover, our data demonstrate that ASC undergoes TLR4-driven translational bursting from polyribosomes during inflammasome priming. This translational engagement is dependent on BCAA availability and mechanistic target of rapamycin (mTOR) activity, which regulate the kinetics of inflammasome assembly. In contrast, the translation of NLRP3 and caspase-1 is largely insensitive to these inputs. Furthermore, we find that BCAAs regulate NLRP3 inflammasome activation in both mouse and human macrophages, in the context of bacterial infection, and during lipopolysaccharide (LPS)-induced sepsis in vivo . Altogether, this work unveils a novel inflammasome priming event governed by the amino acid environment. These findings further highlight how the activity of proteins maintained in equilibrium like ASC can be dynamically regulated through rapid changes in mRNA translation.

  PublisherOA PDFDOI

• A case of cyclical cushing's disease: a diagnostic conundrum of investigations and management

  Endocrine Abstracts · 2025-02-19

  article
  PublisherDOI

• Single-cell spatial mapping reveals reproducible cell type organization and spatially-dependent gene expression in gastruloids

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-07-14 · 3 citations

  preprintOpen accessSenior authorCorresponding

  Gastruloids are three-dimensional stem-cell-based models that recapitulate key aspects of mammalian gastrulation, including formation of an anterior-posterior (AP) axis. However, we do not have detailed spatial information about gene expression and cell type organization, particularly at the level of individual gastruloids. Here, we report a spatially resolved, single-cell molecular catalog of the transcriptomes of 26 individual gastruloids. We found that cell type composition and spatial organization were remarkably consistent across gastruloids. Posterior cell types formed distinct, organized clusters, while anterior cell types were more disorganized. To distinguish progressive differentiation from cell type differences, we developed the L-metric, a parameter-free quantification of mutually exclusive gene expression. This analysis revealed spatial organization without explicit encoding, recapitulated known cell type relationships, and identified novel gene expression states and spatial subclusters within cell types. We confirmed that in gastruloids, NMP differentiation occurred through a continuous, spatially-coordinated process. We also showed that endothelial precursors exhibited unique spatial organization and had distinct gene expression profiles dependent on their association with anterior somitic or posterior endodermal tissues. This work enables the rigorous use of gastruloids as models for studying the molecular mechanisms underlying mammalian development and tissue organization, and introduces novel computational tools for analyzing spatially-resolved single-cell datasets.

  PublisherOA PDFDOI

• MedSyn: AI-driven Medical History Summarization Using EHR Data

  2025-07-18

  article

  The increasing complexity and volume of Electronic Health Records (EHRs) present ongoing challenges for health-care professionals in obtaining meaningful clinical insights. The Intention of this paper is to introduce an AI-driven solution that employs advanced natural language processing (NLP) techniques and Large Language Models (LLMs) to summarize unstructured EHR data through a question-answering framework. Our approach combines Retrieval-Augmented Generation (RAG) with the Mixtral-8×7b-v0.1 LLM, augmented by Knowledge Graphs, to provide context-aware responses to medical queries based on FHIR-standard EHR data. By leveraging knowledge graphs, the system is able to establish structured data relationships, enhancing retrieval efficiency and interpretability. Fully aligned with the Fast Healthcare Interoperability Resources (FHIR) standard, this method ensures interoperability and smooth integration into existing healthcare systems.

  PublisherDOI

• Abstract 6385: Heterogeneity in sensitivity to second-line inhibitors of therapy-resistant melanoma

  Cancer Research · 2025-04-21

  articleSenior author

  Inhibition of the mutant BRAFV600E protein in melanoma via targeted therapy causes many cells to die, but resistant populations emerge that are refractory to further treatments. Strategies to target resistance have proven difficult to find, which could be explained by the potential heterogeneity of the population. It is assumed that genetic mutations alone cause resistance, but targeting such mutations has met with limited success in melanoma. Work from our lab and others has now established that non-genetic variability also leads to resistance in vitro and in vivo: we have shown that rare clonal cells, marked by differences in gene expression, survive and form resistant colonies. We recently demonstrated in vitro and in vivo that individual resistant colonies have distinct morphological and transcriptional differences, even with a clonal parent population. These results provide a potential explanation for the difficulty in identifying effective second-line therapies: individual resistant types may have different drug susceptibilities, so an effective second-line drug would have to kill all resistant types at once. These findings raise the possibility that there may be therapeutic vulnerabilities to exploit that are invisible in the conventional view of homogeneous resistance. For instance, drug A could kill half of the resistant population while drug B kills the other half. These drugs alone would not be viable second-line treatments, but together they could be highly effective, greatly expanding the pool of drugs that could combat resistance. However, screens to evaluate all combinations of potential drugs would be prohibitively large. This search space could be simplified by determining the drug sensitivities of individual resistant types and using those drugs in combination to eliminate a larger proportion of the total resistant population. However, such per-type classification has not yet been performed in melanoma. We generated clonal cell lines from individual resistant colonies, screening four and the drug-naive line using a library of ∼2200 drugs and small molecules with known pathways and targets in cancers. These genetically identical resistant lines indeed had distinct drug sensitivity profiles, and therapy-resistant lines as a whole had novel sensitivities and resistances compared to the drug-naive line. Treating drug-naive cells with targeted therapy in combination with drugs that kill resistant lines can reduce or eliminate resistance, even when the candidate second-line drugs alone do not kill drug-naive cells. Using a panel of 24 drugs identified as of interest in our initial screens, we screened 13 resistant lines, including the original four and four resistant to a different targeted therapy. We found potential links between the transcriptional and toxicity profiles of resistant colonies, and are continuing to test pairs of drugs that kill distinct fractions of the resistant population. Citation Format: Gianna T. Busch, Ryan H. Boe, Jingxin Li, Pavithran T. Ravindran, Arjun Raj. Heterogeneity in sensitivity to second-line inhibitors of therapy-resistant melanoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6385.

  PublisherDOI

• Innate Immune Memory is Stimulus Specific

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-01-24 · 4 citations

  preprintOpen accessSenior authorCorresponding

  Innate immune memory (also termed trained immunity) is defined in part by its ability to cross-protect against heterologous pathogens, and can be generated by many different stimuli, suggesting a "universal" trained state. However, different stimuli could form distinct memories, leading to stimulus-specific trained responses. Here, we use primary human monocyte-derived macrophages to demonstrate phenotypic and epigenetic stimulus specificity of innate immune memory six days after initial exposure. Quantification of cytokine production with single-molecule RNA imaging demonstrates stimulus-specific patterns of response to restimulation at the single cell level. Differential licensing of inflammatory transcription factors is associated with encoding of specificities in chromatin. Trained cells show stronger responses to secondary stimuli that are more similar to the initial stimulus they experienced, suggesting a functional role for these stimulus-specific memories. Rather than activating a universal training state, our findings demonstrate that different stimuli impart specific memories that generate distinct training phenotypes in macrophages.

  PublisherOA PDFDOI

• Single-cell susceptibility to viral infection is driven by variable cell states

  Cell · 2025-11-14 · 8 citations

  articleSenior author
  PublisherDOI

• Evaluation and Stability Study of Hair Colourants – A Comparison

  International Journal of Pharmaceutical Sciences Review and Research · 2025-06-01

  articleOpen access
  PublisherDOI

• Lineage memory shapes viral resistance barriers in human skin

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-06-27

  preprintOpen accessSenior authorCorresponding

  Abstract Individual cells within a given population exhibit striking variability in viral susceptibility, but it remains unknown whether this heterogeneity reflects memories encoded into the cellular lineage or true probabilistic variability. We used multi-color lineage tracing in a human primary organotypic skin model to reveal that viral resistance is encoded within specific cellular lineages. These lineages create distinct boundaries that block viral spread. Our lineage analyses in vitro confirmed that viral susceptibility exhibits strong heritability across cell generations, with siblings and cousins displaying remarkably similar infection outcomes. ATAC and proteomics profiling of resistant and susceptible clones revealed distinct epigenomic and proteomic states, with the transcription factor AP-1 emerging as a potential central regulator of lineage-encoded viral resistance. Inducing AP-1 activity with PMA rendered cells resistant to viral infection, suggesting a causative role in mediating resistance memory. Our findings demonstrate that antiviral resistance in human skin cells is encoded within cellular lineages and preserved through cell divisions, revealing how cell memory may shape infection dynamics and viral containment in tissues.

  PublisherOA PDFDOI

• Wheat Disease Detection Using Deep Convolutional Neural Networks: A Machine Learning Approach to Resolve the Agricultural Intrusion

  Journal of Neonatal Surgery · 2025-02-07 · 3 citations

  articleOpen access

  A system for detecting wheat leaf diseases, specifically Septoria and stripe rust, has been developed using a Convolutional Neural Network (CNN) implemented with TensorFlow. The model was trained on a dataset of 407 images of wheat leaves and achieved 97% accuracy in differentiating between healthy leaves and those with diseases. The system is deployed as a user-friendly web application, allowing farmers to upload images of potentially infected leaves and receive instant disease predictions. The application also provides information on disease symptoms, lifecycles, and management strategies from trusted sources. This technology has the potential to revolutionize wheat disease management, enhance farm productivity, and strengthen food security in India and other wheat-producing regions. Farmers can make informed decisions, implement targeted interventions, and minimize yield losses by providing timely and accurate disease detection. The system aims to facilitate early detection and precise classification of diseases, ultimately helping farmers minimize resource wastage and prevent economic losses. The project underscores the potential of deep learning techniques for real-time disease detection and management in wheat crops. The web application is designed to be accessible to farmers in remote areas, addressing the limitations of traditional disease diagnosis methods. The system's accuracy and efficiency can help reduce the economic impact of wheat diseases, which can cause significant yield losses and impact livelihoods. Overall, the project demonstrates the potential of AI-powered solutions for improving agricultural practices and enhancing food security

  PublisherOA PDFDOI

Recent grants

• Understanding and Overcoming Resistance to BRAF/MEK Kinase Inhibitors in Melanoma

  NIH · $2.9M · 2019–2025

• A plasticity and reprogramming paradigm for therapy resistance at the single cell level

  NIH · $3.3M · 2018–2023

• An integrated approach to melanoma metastasis and therapy resistance: effects of age-related changes in the ECM and the biomechanics of the skin

  NIH · $3.0M · 2019–2025

• Decoding the bridges and barriers to cellular reprogramming and lineage identity

  NIH · $5.6M · 2019–2025

• NIH Grant U01HL129998

  NIH · $4.2M · 2021

Frequent coauthors

• Ian A. Mellis

  Columbia University

  55 shared

• Ashani T. Weeraratna

  46 shared

• Margaret C. Dunagin

  University of Pennsylvania

  44 shared

• Mitchell E. Fane

  41 shared

• Meenhard Herlyn

  41 shared

• Amanpreet Kaur

  Punjab Agricultural University

  40 shared

• Jennifer A. Wargo

  The University of Texas MD Anderson Cancer Center

  35 shared

• Rajasekharan Somasundaram

  35 shared