
Serena Tan
· Assistant Professor Of PathologyVerifiedStanford University · Rheumatology
Active 1998–2026
About
Serena Tan is an Assistant Professor of Pathology at Stanford University and is affiliated with the Center for Artificial Intelligence in Medicine & Imaging (AIMI). Her work focuses on the application of artificial intelligence in the field of medicine and imaging, contributing to the advancement of healthcare through innovative AI research. As part of AIMI, she is involved in initiatives that promote research, education, and collaboration in AI for healthcare, supporting projects and programs that aim to improve medical diagnostics and treatment through technological innovation.
Research topics
- Genetics
- Biology
- Cell biology
- Computational biology
- Medicine
- Internal medicine
- Immunology
- Pathology
Selected publications
Multiomics and deep learning dissect regulatory syntax in human development
Nature · 2026-04-08 · 1 citations
articleOpen accessAbstract Transcription factors establish cell identity during development by binding regulatory DNA in a sequence-specific manner, often promoting local chromatin accessibility and regulating gene expression 1 . Mapping accessible chromatin offers critical insights into transcriptional control, but available datasets for human development are restricted to bulk tissue, single organs or single modalities 2 . Here we present the Human Development Multiomic Atlas, a single-cell atlas of chromatin accessibility and gene expression from 817,740 fetal cells across 12 organs, spanning 203 cell types and more than 1 million candidate cis -regulatory elements, many of which exhibit organ-specific in vivo enhancer activity. Deep learning models trained to predict accessibility from local DNA sequence unravel a comprehensive lexicon of motifs that influence accessibility, including composite motifs exhibiting distinct syntactic constraints that are predicted to mediate transcription factor cooperativity. We identify ‘hard’ syntactic rules requiring precise motif spacing and orientation, ‘soft’ rules allowing flexible motif arrangements, and ubiquitous motifs inhibiting accessibility. Model-based interpretation of genetic variants reveals that disruption of motifs with positive and negative effects is associated with concordant effects on gene expression. Our work delineates how motif syntax governs cell-type-specific chromatin accessibility and provides a foundational resource for decoding cis -regulatory logic and interpreting genetic variation during human development.
Clinical Cancer Research · 2025-10-27 · 2 citations
articlePURPOSE: Historically, neuroblastoma risk stratification has been performed with clinical stage as the starting point and successively adding other prognostic factors thereafter. This study takes an alternative approach to define risk groups of patients with neuroblastoma by starting with the International Neuroblastoma Pathology Classification (INPC). EXPERIMENTAL DESIGN: The cohort of patients with neuroblastoma previously used for developing the Children's Oncology Group-Revised Neuroblastoma Risk Classification (RNRC) system was reanalyzed by survival tree regression analysis, starting with the INPC distinguishing favorable-histology and unfavorable-histology categories. The resultant two branches were further divided first by the International Neuroblastoma Risk Group Staging System and successively by other prognostic factors. RESULTS: This new stratification system, the INPC-Risk Grouping (INPC-RG), is simpler than the RNRC system, eliminating unnecessary decision trees, and distinguishes four risk groups (groups I-IV). Using only INPC (unfavorable histology) and International Neuroblastoma Risk Group Staging System (stage M), INPC-RG defines patients with highly aggressive group IV tumors, whose 5-year event-free survival was worse than that of the RNRC high-risk group. Additionally, it identifies group III patients whose 5-year event-free survival spanned 50% to 80%, which was not identified by the RNRC. CONCLUSIONS: The benefits of using this new INPC-RG system are fourfold: (1) it allows for the rapid identification of group IV patients, (2) it lays the foundation for further refinement of group III, (3) it can stratify patients when the amount of tumor tissue is limited, and (4) it allows patients in resource-limited areas to be appropriately stratified, potentially improving the worldwide treatment of patients with neuroblastoma.
Comparative analysis of RNA expression identifies effective targeted drug in myoepithelial carcinoma
npj Precision Oncology · 2025-05-16 · 1 citations
articleOpen accessMyoepithelial carcinoma is an ultra-rare pediatric solid tumor with no targeted treatments. Clinical implementation of tumor RNA sequencing (RNA-Seq) for identifying therapeutic targets is underexplored in pediatric cancer. We previously published the Comparative Analysis of RNA Expression (CARE), a framework for incorporating RNA-Seq-derived gene expression into the clinic for difficult-to-treat pediatric cancers. Here, we discuss a 4-year-old male diagnosed with myoepithelial carcinoma who was treated at Stanford Medicine Children's Health. A metastatic lung nodule from the patient underwent standard-of-care tumor DNA profiling and CARE analysis, wherein the patient's tumor RNA-Seq profile was compared to over 11,000 uniformly analyzed tumor profiles from public data repositories. DNA profiling yielded no actionable mutations. CARE identified overexpression biomarkers and nominated a treatment that produced a durable clinical response. These findings underscore the utility of data sharing and concurrent analysis of large genomic datasets for clinical benefit, particularly for rare cancers with unknown biological drivers.
Cancer Research · 2025-04-21
articleAbstract Neuroblastoma (NB) is the most frequent extracranial childhood tumor. The majority of NB tumors arise in the adrenal gland with bone marrow (BM) metastases present in 70.5% of Stage IV disease. A recent study using 23 most commonly used two-dimensional (2D) adherent NB cell lines found that expression of GD2, a target for immunotherapy, is significantly correlated with adrenergic (ADRN) and mesenchymal (MES) cell lineages through epigenetic regulation of ganglioside synthesis enzyme ST8SIA1. We aimed to establish primary NB patient cell line from adrenal NB or BM metastasis to determine whether ADRN and MES lineage specific phenotypes exist in primary NB patient cells and are also associated with distinct GD2 and ST8SIA1 expression. Using our in vitro primary cell culture conditions, all NB cells derived from four individual patients’ BM metastases spontaneously assembled into three-dimensional (3D) spheroids in scaffold-free regular culture plates with a predominant GD2 high ADRN cell type as determined by flow cytometric and immunofluorescent analysis of ADRN marker genes. Cells maintained self-assembling ability for multiple passages (up to 30) within 5-8 months of prolonged in vitro cultures at a sustainable growth rate. In contrast to the BM-derived NB cells, a majority of the adrenal tumor-derived NB cells became 2D adherent within 7-10 days and displayed marked changes in immuno-phenotypes after several weeks, shifting from predominantly GD2-high to a mixture of high and low GD2 expression. Forcing the adherent NB cells with mixed GD2 level into suspension cultures using Corning’s 3D Ultra-Low Attachment (ULA) plate resulted in a marked enrichment of NB spheroids with GD2-high, which regrew as 2D-adherent cells after re-seeding into regular culture plate, and retained the GD2-high and ST8SIA1-high ADRN phenotype for several months. In contrast, the NB adherent cells cultured in parallel without ULA-selection showed predominant GD2-low, ST8SIA1-low and vimentin (VIM)-high MES-like phenotype as characterized by GD2 flow and Real-time quantitative PCR of ADRN and MES marker gene expression. Our novel primary cell culture conditions allows long term in vitro spheroid growth of patient-derived NB cells from BM in regular culture plate, which is scaffold-free, faster, less expensive, less labor intensive, and mimics the in vivo NB cell clusters or aggregates seen in NB patient’s BM aspirates with NB infiltration. Our culture conditions in combination with the 3D ULA-selection strategy enabled us to establish paired ADRN and MES-like NB sublines from two primary NB tumors derived from adrenal and BM sites. These patient-matched, paired NB cell lines in ADRN and MES-like cell states will be useful to study NB lineage specific biology and implications of cell states on an individual patient’s response to targeted immunotherapy. Citation Format: Min Huang, Robbie G. Majzner, Nathan Chiang, Alejandro Solis, Bill Chiu, Serena Tan, Emon Nasajpour, Kathleen M. Sakamoto, Norman J. Lacayo, Claudia K. Petritsch, Sheri L. Spunt, Raya Saab. Novel scaffold free method to establish 3D spheroid and 2D adherent primary neuroblastoma cell lines with adrenergic and mesenchymal-like phenotypes [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 3950.
Comparative analysis of RNA expression identifies effective targeted drug in myoepithelial carcinoma
Research Square · 2025-04-16
preprintOpen accessAnnals of Surgery · 2025-03-20
articleOBJECTIVE: To quantify liver fibrosis in infants with biliary atresia (BA) through automated analysis of collagen extracellular matrix (ECM) ultrastructure in index liver biopsies and use a composite fibrosis architecture score to predict native liver survival. SUMMARY BACKGROUND DATA: Despite early management with Kasai portoenterostomy , BA remains the leading indication for pediatric liver transplantation. There is no established method for quantitatively assessing liver fibrosis in patients with BA, and no factors to accurately predict which patients will ultimately require transplantation early versus late. METHODS: Index liver biopsies from 12 BA patients were retrieved from our pathology archives Masson's Trichrome-stained biopsies were scanned, tiled, binarized, and quantified for 147 ECM features. These features were reduced by Uniform Manifold Approximation and Projection. Pseudotime analysis was applied to summarize global variations in architecture and assign BA-ECM scores to all biopsy images. Retrospective chart review was performed to correlate clinical characteristics with BA-ECM score. RESULTS: BA-ECM score, a multi-dimensional fibrosis architecture score, was significantly higher for biopsies from listed patients compared to non-listed patients (35.9 vs. 22.9, *P<0.0001). High BA-ECM score was characterized by thick, patchy, irregular ECM, while low BA-ECM score was associated with large-volume thin, porous collagen fibers. Survival analysis stratified by the third quartile BA-ECM score of all data points demonstrated a significant difference in native liver survival (*P=0.02). CONCLUSIONS: We present the application of an automated ECM ultrastructure analysis tool designed to capture and quantify 147 aspects of fibrotic tissue heterogeneity. These manifold features are summarized using a multi-dimensional BA-ECM score that could be used to prognosticate disease course for BA patients.
Proceedings of The Nutrition Society · 2025-04-01
reviewOpen accessSenior authorTelomere length is a biomarker of ageing (1) . A shorter telomere length is associated with an increased risk of age-related diseases and mortality. Oxidative stress and inflammation are predominant mechanisms leading to telomere shortening (2) . Diets and food groups high in antioxidant and anti-inflammatory properties are shown to be protective against telomere shortening (3) . The nut and seed food group is rich in nutrients such as unsaturated fats, vitamins, and minerals, and contains antioxidants and anti-inflammatory phytochemicals. Evidence is emerging on the beneficial effects of nuts and seeds in the prevention and management of age-related chronic conditions. This review aims to evaluate the role of nut and seed intake on telomere length in humans using the evidence from observational and interventional studies. Four databases, including Medline, CINAHL, Embase and Web of Science, were systematically searched from inception to 12 March 2024 for observational and interventional studies assessing the intake of nut or seed or applied nut or seed interventions and measured telomere length as an outcome in adult human participants (age ≥ 18 years). The quality assessment of the included studies was performed using the Academy of Nutrition and Dietetics Evidence Analysis Library® November 2022: Quality Criteria Checklist. Nine observational and four interventional studies were included. A positive association between nut and seed intake and telomere length was reported in three of the nine observational studies. None of the interventional studies reported a significant positive effect of nuts on telomere length. Three of the observational and interventional studies were classified as high quality, and the remaining studies were of neutral quality. Meta-analysis was not warranted due to the high heterogeneity in the telomere length measurements across the studies. The findings are inconsistent across these studies, and the evidence is insufficient to establish a beneficial role of nut and seed intake on telomere length. Larger epidemiological studies and adequately powered long-term randomised controlled trials are needed to establish the positive role of nut and seed on telomere length. However, nut and seed should continue to be recommended as a part of a healthy diet, given their proven benefits against age-related conditions.
Malignant peripheral nerve sheath tumors: a report from Children’s Oncology Group study ARST0332
JNCI Journal of the National Cancer Institute · 2025-12-08
articleOpen accessBACKGROUND: The cornerstone of the treatment of malignant peripheral nerve sheath tumors is surgical resection. Radiation and chemotherapy are variably employed. The optimal treatment remains uncertain, particularly for unresectable or metastatic disease and patients with neurofibromatosis type 1 (NF-1). METHODS: We present data for 58 patients with newly diagnosed malignant peripheral nerve sheath tumors enrolled on the Children's Oncology Group study ARST0332. Patients were treated with risk-adapted therapy including surgery with or without radiotherapy and ifosfamide and doxorubicin chemotherapy. RESULTS: Most patients had primary tumors that were greater than 5 cm (86%), deep (95%), and invasive (74%), and 10% had distant metastases. Of the patients, 32 (55%) had germline NF-1, and 26 (45%) did not. Among patients, 31 received neoadjuvant therapy, and 22 were evaluable for response with 5 (23%) attaining an objective response, 10 (45%) stable disease, and 7 (32%) progressive disease. Estimated 5-year event-free survival was 87%, 52%, and 0% for the low- (n = 8), intermediate-, (n = 44), and high-risk (n = 6) patients, respectively. In univariate analysis, event-free survival and overall survival differed by sex, presence or absence of metastatic disease, risk group, and achievement of upfront or delayed R0 and/or R1. There was no difference in event-free survival or overall survival based on germline NF-1 status. CONCLUSION: The treatment strategy in ARST0332 achieved excellent outcomes for low-risk malignant peripheral nerve sheath tumors. Patients with high-risk (metastatic) malignant peripheral nerve sheath tumors have poor outcomes, and novel treatments are needed (NCT00346164).
Benchmarking cell type and gene set annotation by large language models with AnnDictionary
Nature Communications · 2025-10-28 · 4 citations
articleOpen accessWe develop an open-source package called AnnDictionary to facilitate the parallel, independent analysis of multiple anndata. AnnDictionary is built on top of LangChain and AnnData and supports all common large language model (LLM) providers. AnnDictionary only requires 1 line of code to configure or switch the LLM backend and it contains numerous multithreading optimizations to support the analysis of many anndata and large anndata. We use AnnDictionary to perform the first benchmarking study of all major LLMs at de novo cell-type annotation. LLMs vary greatly in absolute agreement with manual annotation based on model size. Inter-LLM agreement also varies with model size. We find that LLM annotation of most major cell types to be more than 80-90% accurate, and will maintain a leaderboard of LLM cell type annotation. Furthermore, we benchmark these LLMs at functional annotation of gene sets, and find that Claude 3.5 Sonnet recovers close matches of functional gene set annotations in over 80% of test sets. Cell type labelling in single-cell datasets remains a major bottleneck. Here, the authors present AnnDictionary, an open-source toolkit that enables atlas-scale analysis and provides the first benchmark of LLMs for de novo cell type annotation from marker genes, showing high accuracy at low cost.
American Journal of Respiratory and Critical Care Medicine · 2025-05-01
articleAbstract Introduction Hereditarypulmonary arterial hypertension (HPAH) and hereditary hemorrhagic telangiectasia(HHT) are two rare genetic diseases affecting the pulmonary vasculature. Theyare both characterized by haploinsufficiency in different components of the bonemorphogenetic protein receptor type 2 (BMPR2) pathway. Despite shared genetics, the vascular phenotype differs. HPAH causes plexiform lesions, convolutes ofvascular channels that are considered a pathological hallmark of PAH. Incontrast, HHT can lead to pulmonary arteriovenous malformations (pAVMs), directconnections between arteries and veins. The pathobiology of both plexiformlesions and pAVMs is incompletely understood. In HHT-associated vascularmalformations in the skin and some solid organs, there is evidence that loss ofheterozygosity caused by somatic loss-of-function mutations in the functionalallele of the HHT gene results in clonally expanding endothelial cells. Inplexiform lesions, clonal vascular expansion has been described as well. Wetherefore aimed at detecting somatic mutations in pulmonary vascular lesions ofpatients with HPAH and a patient with HHT and PAH (HHT/PAH). Methods Genomic DNA wasextracted from vascular lesions of 2 patients with HPAH and 1 patient with HHT/PAH.Whole exome sequencing was performed on 2 plexiform lesions of the HPAHpatients. A targeted deep sequencing panel including 3 HHT causing genes, and 11VM associated genes was applied to 4 pAVMs and 14 plexiform lesions of thepatient with HHT/PAH. Results Germline variants in BMPR2 andSMAD9 were confirmed in the 2 HPAH patients, and a germline variant in Endoglinin the HHT/PAH patient. Interestingly, a somatic mutation in BMPR2 was found in theplexiform lesion of the patient with HPAH due to haploinsufficiency in BMPR2, whereas a somatic mutationin SMAD9 was present in the plexiformlesion of the SMAD9 patient. In the HHT/PAH patient, somatic mutations in VMassociated genes occurred in 4 out of 14 plexiform lesions. These included asomatic activating mutation in Phosphatidylinositol-4,5-Bisphosphate 3-KinaseCatalytic Alpha (PIK3CA) that was previously described in breast cancer andfunctionally validated as a moderately potent oncogenic mutation. Furthermore, no somatic mutations were found in pAVMs of this patient. Conclusions We identified localsomatic mutations in PAH causing genes in plexiform lesions of HPAH patientsand a somatic activating mutation in PIK3CA in a plexiform lesion of an HHT/PAHpatient. Both events likely result inaltered signaling that might contribute to changed endothelial cell behaviorand excessive proliferation in plexiform lesions.
Frequent coauthors
- 19 shared
Mark A. Krasnow
Howard Hughes Medical Institute
- 15 shared
Jingsi Ming
East China Normal University
- 12 shared
Stephen R. Quake
Stanford University
- 12 shared
Antoine de Morrée
Stanford University
- 10 shared
Sheri L. Spunt
- 10 shared
Roozbeh Dehghannasiri
Stratford University
- 9 shared
Julia Olivieri
Stanford University
- 9 shared
SoRi Jang
Howard Hughes Medical Institute
- Resume-aware match score
- Save to shortlist
- AI-drafted outreach
See your match with Serena Tan
PhdFit ranks faculty by your research interests, methods, and publications — grounded in their actual work, not templates.
- Free to start
- No credit card
- 30-second signup