About

Deshka Foster, MD, PhD, is an assistant professor of surgery at Stanford University School of Medicine, specializing in surgical oncology and hepatopancreatobiliary (HPB) surgery. She is a dual fellowship-trained surgeon with expertise in treating cancers of the pancreas, bile ducts, gallbladder, liver, and gastrointestinal cancers. Dr. Foster directs the Hepatic Artery Infusion (HAI) Pump Therapy Program at Stanford and offers advanced, personalized treatment options for patients with HPB malignancies. As a surgeon-scientist, Dr. Foster holds a Ph.D. in Cancer Biology from Stanford University and leads a basic science and translational research laboratory. Her research interests include the pancreatic tumor microenvironment, cancer-associated fibroblasts, and solid tumor response to immunotherapy. Her laboratory specifically investigates cells that influence tumor behavior and resistance to treatment, focusing on cholangiocarcinoma and pancreatic cancers. Additionally, her research encompasses abdominal adhesion fibrosis, which involves internal scar tissue formation after surgical procedures. Dr. Foster has published extensively in peer-reviewed journals and presented her work at national and regional meetings. She is actively involved in professional organizations related to surgical oncology and hepatobiliary surgery, serving on research committees and holding memberships in several societies.

Research topics

• Medicine
• Genetics
• Biology
• Cell biology
• Immunology
• Internal medicine
• Pathology
• Surgery
• Computational biology
• Radiology
• Cancer research
• Anatomy

Selected publications

• Abstract 6202: Integrated multiomic atlas of pancreatic solid pseudopapillary neoplasms suggests acinar cells as a potential cell-of-origin

  Cancer Research · 2026-04-03

  article

  Abstract Introduction: Solid pseudopapillary neoplasms (SPNs) of the pancreas are rare, low-grade tumors that typically affect young women and occasionally recur or metastasize. While the histological features and mutational profile of SPNs are well documented, their cell-of-origin and spatial organization remain incompletely understood. The purpose of this study was to delineate the development and spatial transcriptomic landscape of SPNs. Methods: Regions of interest from 10 surgically resected SPN specimens collected between 2020 and 2025 were profiled using the G4X Singular In Situ Multiomic platform. Gene expression data were integrated and analyzed to generate spatial embeddings, followed by unsupervised clustering. Cell types were annotated using canonical markers, and proteomic data were integrated for further resolution. Results: Spatial transcriptomic profiling of 30 SPN sections from 10 patients yielded 3.6 million high-quality cells. The cohort was entirely female, with a median age of 31.5 years. Integrated analysis demonstrated a heterogeneous tumor microenvironment composed of exocrine pancreatic cells, immune cells, fibroblasts, endothelial cells, and tumor cells. Tumor cells demonstrated intra- and inter-tumoral heterogeneity, with five distinct subpopulations: (1) CPB1+ tumor cells, (2) MYH11+ tumor cells, (3) FN1+ VIM+ MMP2+ tumor cells, (4) IGHG1+ FCGR1A+ TBX21+ tumor cells, and (5) STAT1+ TAP1+ CD74+ tumor cells. Mapping these tumor transcriptional states back onto the tissue sections revealed distinct microenvironmental niches. IGHG1+ FCGR1A+ TBX21+ tumor cells and MYH11+ tumor cells were the most prevalent tumor cells across samples. STAT1+ TAP1+ CD74+ tumor cells, characterized by elevated antigen-presentation and interferon-signaling programs, were localized in areas with high immune-infiltration, which was further corroborated and delineated proteomically. CPB1+ tumor cells showed relatively increased expression of acinar cell markers and formed spatial gradients at tumor-acinar interfaces, suggesting that SPN tumor cells may arise from an acinar lineage. Conclusion: We present the first multiomic analysis of SPNs of the pancreas, establishing a high-resolution atlas of their cellular and spatial architecture. We identify conserved tumor transcriptional and proteomic states that occupy distinct spatial niches. Spatial gradients in CPB1+ tumor cells suggest a potential acinar origin for SPN tumor cells, warranting further lineage-focused investigation. These findings collectively refine our understanding of SPN tumor biology. Citation Format: Biren Reddy, Maria Korah, James P. Agolia, Rosyli Reveron-Thornton, Maggie Lam, Deshka Foster, Michael T. Longaker, Daniel Delitto. Integrated multiomic atlas of pancreatic solid pseudopapillary neoplasms suggests acinar cells as a potential cell-of-origin [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6202.

  PublisherDOI

• A Pan-Cancer Single-Cell Atlas to Evaluate Tumor Identity, Cell Line Concordance, and Dependency Mapping

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-02-17

  articleOpen access

  Bulk RNA sequencing enables pan-cancer transcriptional analyses, but obscures cancer cell-specific programs due to admixture with nonmalignant cells, thereby limiting direct comparison between experimental models and primary tumors. Single-cell RNA sequencing (scRNA-seq) overcomes these limitations; however, the biological interpretability of public datasets is often compromised by variable data quality, inconsistent annotation, and atlas-scale aggregation strategies that prioritize data volume over biological coherence. We therefore developed a stringent integration framework that prioritizes representative malignant transcriptional states. Using Mahalanobis distance-based selection within batch-corrected latent space, we constructed a pan-cancer atlas comprising 135,424 high-quality malignant cells from 499 samples across 36 adult and pediatric cancers. Atlas-derived cancer signatures were used to determine tumor-cell line concordance and project ElasticNet models trained on DepMap CRISPR screens to infer cancer-specific gene dependencies. The scTumor Atlas establishes a scalable framework for tumor identity inference, cancer cell line benchmarking, and systematic identification of genetic vulnerabilities.

  PublisherDOI

• Abstract 2201: Multidimensional transcriptomic alas of recurrent uterine leiomyosarcomas uncovers stem-like hormonal cells with high drug sensitivity and improved patient outcomes.

  Cancer Research · 2026-04-03

  article

  Abstract INTRODUCTION: Uterine leiomyosarcomas (ULMS) are rare, aggressive tumors with profound genomic heterogeneity, which has precluded the identification of effective targeted therapies. The overall purpose of this study was to investigate the transcriptomic and spatial landscape of ULMS to identify new therapeutic avenues. METHODS: We performed single-cell RNA sequencing (scRNA-seq) on fresh ULMS tumors using the 10X Genomics platform and spatial transcriptomics on FFPE sections using the Singular G4X platform. Resulting cells were integrated with scVI or resolVI and annotated using canonical markers. Tumor signatures were correlated with patient outcomes using bulk RNA sequencing data and novel drugs predictions for each tumor cell subtype were computed using the scIDUC algorithm. RESULTS: ScRNA-seq was performed on 15 recurrent, metastatic ULMS tumors from 13 patients at the time of debulking surgery, which yielded 204,250 high quality cells. Spatial transcriptomics was performed on 29 sections from 10 patients, which yielded 2.3 million spatially resolved cells. Both assays revealed heterogenous tumor microenvironments comprised of several types of cells, including myeloid cells, lymphoid cells, endothelial cells, and tumor cells. Among the tumor cells, three prevalent subtypes emerged: 1) interferon-signaling cells that were distributed throughout sections, 2) mesenchyme-like cells arranged in nest-like configurations and around blood vessels, 3) ischemic cells with a high ribosomal signature organized around necrotic penumbras. Additionally, dedifferentiated stem-like cells with high expression of ESR1, PGR, and AR, along with Wnt signaling pathway markers, were scattered throughout tumor sections. These hormonal cells clustered into two subtypes on scRNA-seq: 1) those with high ESR1 and low AR/PGR and 2) those with low ESR1 and higher AR/PGR expression. These cells were the most sensitive to all interrogated drugs, including current first-line therapies gemcitabine, docetaxel, and doxorubicin. All other tumor subtypes were highly resistant to most drugs, though several alternative drug candidates uniquely targeting each tumor subtype were identified. Furthermore, by correlating with bulk RNA sequencing, each tumor subtype was shown to be associated with unique patient outcomes. AR and PGR hormone receptor-expressing cells correlated with improved patient survival, whereas ischemic cells correlated with the worst survival outcomes. CONCLUSIONS: We present a comprehensive atlas of ULMS, identifying several transcriptomic subtypes of tumor cells, including dedifferentiated stem-like hormonal cells that correlate with better drug sensitivity and improved patient outcomes. Findings from this study may further facilitate patient prognostication and guide targeted therapeutic management. Citation Format: Maria Moozhiyil Korah, James P. Agolia, Biren Reddy, Renceh A. Flojo, Amanda Gonçalves, Lilin Wang, Rosyli F. Reveron-Thornton, Chuner Guo, Beatrice Sun, Amanda R. Kirane, George Poultsides, Deshka Foster, R. Stephanie Huang, Gregory W. Charville, Michael T. Longaker, Daniel Delitto. Multidimensional transcriptomic alas of recurrent uterine leiomyosarcomas uncovers stem-like hormonal cells with high drug sensitivity and improved patient outcomes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2201.

  PublisherDOI

• Abstract 1426: A single-cell tumor atlas defines robust pathway and gene signatures enabling cancer cell-line fidelity assessment.

  Cancer Research · 2026-04-03

  article

  Abstract The aim of this study was to determine whether rigorous quality control applied across multiple single-cell RNA (scRNA-seq) sequencing datasets could generate reproducible transcriptional signatures that accurately reflect tumor biology and support evaluation of cancer model fidelity. We aggregated publicly available scRNA-seq datasets and processed all samples through a high-stringency quality-control pipeline that included thresholds of >5,000 counts, <10% mitochondrial content, removal of samples with < 200 cells, and doublet identification using Scrublet. The resulting atlas included 135,441 high-quality tumor cells across 494 samples representing 36 adult and pediatric tumor types. We identified tumor specific gene signatures through differential expression analysis and computed hallmark pathways analysis. Strict QC markedly improved the clarity and biological coherence of tumor-specific signatures enabling us to group otherwise unrelated primary tumors into reproducible transcriptional archetypes (proliferative, immune-signaling, and metabolic) states. These atlas-derived gene signatures showed strong concordance with independent bulk RNA-seq datasets and spatial transcriptomic signatures validating the approach/model. To examine the utility of these signatures, we projected gene expression profiles from established cancer cell lines onto the atlas-derived signatures. This analysis scored cell lines based on how representative they remained to their tumor of origin. Culture adaptation, metabolic drift, or the loss or gain of hallmark pathways, are known causes of transcriptional divergence in in-vitro models. These findings demonstrate that rigorous QC enables construction of a reproducible, pan-cancer single-cell atlas that yields stable transcriptomic signatures suitable for more reliable tumor characterization than offered by the publicly resources (HTAN, EcoTyper, DepMap, Cancer SCEM etc) which vary significantly in their QC measures. This atlas provides a high-quality reference for tumor biology and a framework for evaluating the fidelity of cancer cell lines, with implications for model selection, assessment of therapeutic vulnerabilities, and translational research. Citation Format: Rosyli F. Reveron-Thornton, Chuner Guo, James P. Agolia, Maria Moozhiyil Korah, Peter Yuxin Xie, Andrea Delitto, Amanda Gonçalves, Angela Tabora, Biren Reddy, Wesley Bobst, Amanda R. Kirane, Monica Dua, Brendan Visser, Byrne Lee, George Poultsides, Jeffrey A. Norton, Derrick C. Wan, Michael T. Longaker, Deshka Foster, Daniel Delitto. A single-cell tumor atlas defines robust pathway and gene signatures enabling cancer cell-line fidelity assessment [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1426.

  PublisherDOI

• Abstract 2270: A MUC1-overexpressing epithelial population is associated with CDH1 loss-of-function gastric adenocarcinoma.

  Cancer Research · 2026-04-03

  article

  Abstract Germline mutations in CDH1 are associated with early-onset diffuse gastric adenocarcinoma. There is no gold standard for screening and management, and endoscopic surveillance is often insufficient for detecting early disease. Prophylactic gastrectomy is maximally preventive but is a major operation associated with significant morbidity and extensive lifestyle modifications. Therefore, there is a need to better understand the pathophysiological features underlying carcinogenesis in the setting of CDH1 mutation. To address this gap, we performed spatial transcriptomic profiling on 29 formalin-fixed paraffin-embedded sections from 8 patients with various stages of gastric adenocarcinoma, including one patient with a germline CDH1 mutation (c.1936+5G>A) found to have multifocal intramucosal poorly cohesive signet-ring cell carcinoma. Image-based cell segmentation and subcellular transcript detection enabled identification of diverse epithelial cell types, as well as their stromal and immune neighborhoods. Epithelial cells from the specimen harboring the germline CDH1 mutation expressed lower levels of CDH1 compared with other samples. Although morphologically normal-appearing, mucosal layers in these sections exhibited altered transcriptomic profiles characterized by an expanded epithelial population with high MUC1 expression. MUC1 is a transmembrane glycoprotein frequently upregulated in various cancers, with a cytoplasmic domain capable of interacting with several intracellular signaling partners, particularly β-catenin. Gene regulatory network analysis revealed a gradient of CTNNB1/β-catenin activation program, suggesting a potential interplay between CDH1 loss-of-function, MUC1 overexpression, and β-catenin pathway dysregulation. These findings reveal a novel MUC1-overexpressing epithelial population associated with CDH1 mutation and early-onset gastric adenocarcinoma, providing new rationales for developing surveillance biomarkers and targeted therapies. Citation Format: Chuner Guo, Rosyli F. Reveron-Thornton, Xiaomo Li, James P. Agolia, Maria Moozhiyil Korah, Peter Yuxin Xie, Amanda Gonçalves, Angela Tabora, Lily Xia, Natali Barakat, George Poultsides, Byrne Lee, Amanda R. Kirane, Deshka Foster, Michael T. Longaker, Gregory W. Charville, Daniel Delitto. A MUC1-overexpressing epithelial population is associated with CDH1 loss-of-function gastric adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2270.

  PublisherDOI

• Abstract 2222: Mapping lymphoid responses to tumor growth and lymph node metastasis: multiomic spatial analysis of pancreatic ductal adenocarcinoma reveals tertiary lymph node structures.

  Cancer Research · 2026-04-03

  article

  Abstract Introduction: Within the next five years, pancreatic ductal adenocarcinoma (PDAC) will become the second-highest cause of cancer death in the United States; even with surgical resection and chemotherapy, median overall survival remains less than three years. Spatial transcriptomic and proteomic analysis is one pathway toward therapeutic target discovery for PDAC. While previous spatial transcriptomic studies have uncovered remarkable intratumoral heterogeneity in primary and metastatic samples, most lack subcellular resolution and multiomic scale. Methods: Using the Singular G4X platform, we performed spatial multiomic analysis of 31 regions of interest in formalin-fixed, paraffin-embedded matched primary tumor and metastatic lymph nodes from eight patients. All patients had a pathological diagnosis of PDAC and received neoadjuvant chemotherapy at a single cancer center, and 7/8 patients showed pathological treatment effect with partial response to therapy. Data were analyzed in python. Results: After quality control, 2.4 million cells were integrated. Major cell types were manually annotated: acinar cells (CPB1+ GATM+), adipocytes (ADIPOQ+ PLIN1+), B cells (MS4A1+ HLA-DRA+), cancer-associated fibroblasts (COL1A1+ LUM+), plasma cells (JCHAIN+ IGHA1/IGHM+), macrophages (CD163+ CD68+), mast cells (KIT+), neuroendocrine cells (NRXN1+ NCAM1+), pericytes/endothelial cells (PECAM1+ RGS5+), smooth muscle cells (TAGLN+ MYH11+), and T cells (CD3D+ IL7R+). Tumor cells formed three distinct clusters: KRT19+ MUC1+ ductal cells, PIGR+ GATM+ ductal cells, and STAT1+ CD44+ cells (from one patient with sarcomatoid dedifferentiation). The predominant cell type in the tumor microenvironment, cancer-associated fibroblasts (CAFs) could be subclustered into inflammatory CAF (CXCL12-high IL6-high), mechanoresponsive CAF (PDGFRB-high ACTA2-high), and steady-state CAF phenotypes. Although we did not select regions of interest explicitly with tertiary lymphoid structures (TLSs) in mind, we found that 7/8 patients and 9/19 tumor sections contained putative TLSs, with 4 sections containing more than one TLS area. TLS morphology appeared similar to that of lymphoid aggregates in tumor-positive lymph nodes. A predominance of CAFs was also noted in the involved lymph nodes. B cells within TLSs stained positive for CD20 protein and expressed CXCR5 transcripts; they were surrounded by T cells that stained positive for CD3 and CD4/CD8 proteins. CXCR5 colocalized closely with its binding partner CXCL13, a known trigger of TLS formation. Conclusion: In a cohort of PDAC patients with lymph node metastasis, multiomic spatial analysis revealed TLSs in nearly all patients, suggestive of immune potential within the PDAC stroma. The CXCR5-CXCL13 axis should remain a target of active investigation in TLS formation. Citation Format: James P. Agolia, Chuner Guo, Rosyli F. Reveron-Thornton, Xiaomo Li, Maria Moozhiyil Korah, Angela Tabora, Byrne Lee, Amanda R. Kirane, George Poultsides, Brendan Visser, Gregory W. Charville, Michael T. Longaker, Deshka S. Foster, Daniel Delitto. Mapping lymphoid responses to tumor growth and lymph node metastasis: multiomic spatial analysis of pancreatic ductal adenocarcinoma reveals tertiary lymph node structures [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2222.

  PublisherDOI

• A Translational Surgical Porcine Model for Postoperative Intra-Abdominal Adhesion Formation

  Journal of Visualized Experiments · 2026-03-13

  articleSenior author

  A porcine model of postoperative intra-abdominal adhesion formation was established using Yucatan mini pigs. The protocol combines midline laparotomy, small bowel resection with two-layer primary anastomosis, and a unilateral, parietal peritoneal abrasion in the format of an open abdominal surgical procedure. Adhesion formation was assessed four weeks postoperatively using established gross and histologic scoring criteria, with evaluations performed by blinded observers. Adhesions developed in all animals using this model and were multifocal, involving bowel loops, between the bowel and abdominal wall, involving the peritoneum overlying other organs in the abdomen (e.g. the liver), and operative sites, with variable severity. Histological analysis at four weeks demonstrated adhesions composed predominantly of extracellular matrix, fibroblasts, and blood vessels, consistent with a remodeling-phase wound healing tissue phenotype. This model is relevant for the study of abdominal adhesion fibrosis biology and/or the translational evaluation of candidate anti-adhesion therapeutics. By integrating a clinically relevant intestinal surgical procedure with a defined peritoneal injury and a standardized assessment strategy, this protocol provides a reproducible approach for inducing and evaluating postoperative intra-abdominal adhesions in a large animal model.

  PublisherDOI

• JoVE Video Dataset

  2026-03-13

  databaseSenior author

  A porcine model of postoperative intra-abdominal adhesion formation was established using Yucatan mini pigs. The protocol combines midline laparotomy, small bowel resection with two-layer primary anastomosis, and a unilateral, parietal peritoneal abrasion in the format of an open abdominal surgical procedure. Adhesion formation was assessed four weeks postoperatively using established gross and histologic scoring criteria, with evaluations performed by blinded observers. Adhesions developed in all animals using this model and were multifocal, involving bowel loops, between the bowel and abdominal wall, involving the peritoneum overlying other organs in the abdomen (e.g. the liver), and operative sites, with variable severity. Histological analysis at four weeks demonstrated adhesions composed predominantly of extracellular matrix, fibroblasts, and blood vessels, consistent with a remodeling-phase wound healing tissue phenotype. This model is relevant for the study of abdominal adhesion fibrosis biology and/or the translational evaluation of candidate anti-adhesion therapeutics. By integrating a clinically relevant intestinal surgical procedure with a defined peritoneal injury and a standardized assessment strategy, this protocol provides a reproducible approach for inducing and evaluating postoperative intra-abdominal adhesions in a large animal model.

  PublisherDOI

• Protocol for orthotopic implantation of a collagen hydrogel to model pancreatic ductal adenocarcinoma in mice

  STAR Protocols · 2026-01-13

  articleOpen accessCorresponding

  Available mouse models for pancreatic ductal adenocarcinoma (PDAC) are limited by slow tumor development and failure to recapitulate key stromal and immune characteristics. Here, we present a protocol for generating a collagen hydrogel mouse model for orthotopic PDAC. We describe steps for embedding mouse pancreatic cancer cells in a dense collagen hydrogel and surgically implanting it into the mouse pancreas. Mouse PDAC tumors typically reach 1 cm in diameter by 10 days after implantation and show immune and stromal cell recruitment. For complete details on the use and execution of this protocol, please refer to Korah et al. 1 • Protocol for embedding and culturing KPC cells in a collagen gel matrix • Surgical procedure for implanting tumor hydrogels into the mouse pancreas • Guidance on characterization of the hydrogel and measurement of mouse tumors • Instructions for limiting variability and ensuring consistent tumor growth Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics. Available mouse models for pancreatic ductal adenocarcinoma (PDAC) are limited by slow tumor development and failure to recapitulate key stromal and immune characteristics. Here, we present a protocol for generating a collagen hydrogel mouse model for orthotopic PDAC. We describe steps for embedding mouse pancreatic cancer cells in a dense collagen hydrogel and surgically implanting it into the mouse pancreas. Mouse PDAC tumors typically reach 1 cm in diameter by 10 days after implantation and show immune and stromal cell recruitment.

  PublisherDOI

• NFAT mediates pro-tumorigenic inflammation in cancer-associated fibroblasts in pancreatic ductal adenocarcinoma

  Cell Reports · 2026-01-01 · 3 citations

  articleOpen access

  CAFs activate an inflammatory phenotype associated with elevated NFAT motif activity and gene expression. In vivo, NFAT inhibition in a mouse model of PDAC significantly reduces tumor weight and fibrosis, supporting its pro-tumorigenic role. Our findings suggest that NFAT mediates IL-1-induced inflammation in PDAC, highlighting its potential as a therapeutic target. This study demonstrates the power of multi-omic analyses to uncover therapeutic targets within the complex TME.

  PublisherDOI

Frequent coauthors

• Michael T. Longaker

  Stanford University

  148 shared

• Shamik Mascharak

  Stanford University

  88 shared

• Malini Chinta

  Stanford University

  68 shared

• Jeffrey A. Norton

  Stanford University

  64 shared

• Ruth Ellen Jones

  Edinburgh Royal Infirmary

  58 shared

• Heather E. desJardins-Park

  Stanford University

  58 shared

• Ankit Salhotra

  Stanford University

  54 shared

• Derrick C. Wan

  Stanford University

  52 shared

Labs

• Foster LaboratoryPI

Education

• M.D.

  Stanford University

• Ph.D.

  Stanford University

Awards & honors

• Career Development Award, Society for Surgery of the Aliment…
• Medical Student Teaching Award, Weill Cornell Department of…
• Best Basic/Translational Oral Presentation, Holman Research…
• The Resident Research Award, Stanford Department of Surgery…
• Dr. Hilary Sanfey Outstanding Resident Award, Association of…