Yutong (Irina) Zhu

Hi there! I'm glad you made it to my digital portfolio :)

I am a 2nd-year Bioengineering PhD student at University of Pennsylvania, advised by Derek Oldridge.

I am interested in developing statistical and AI-driven analytical tools that integrate spatial-omic technologies with conventional pathology imaging to study spatial biology. My goal is to enhance our understanding of the molecular, cellular, and architectural foundations of human health and disease in immunological context .

Email  /  CV  /  LinkedIn  /  GitHub  /  Google Scholar

Using multimodal single-cell and spatial transcriptomic profiling of human tonsils and blood, we define distinct CD4 T follicular helper (Tfh) cell states and identify GNG4 as a hallmark of activated, germinal center–associated Tfh cells. GNG4 expression more specifically marks GC-localized Tfh than canonical markers, distinguishing activated GC Tfh from resting, non-GC, Th17-polarized Tfh states.

We developed an AI-based histopathology model to distinguish long-term disease-free PanNET patients from those who later recurred, using 3,888 H&E images from 926 resections. The pipeline achieved high accuracy (87–89%) across recurrence thresholds and attention maps revealed that small regions, often at invasive edges or vascular invasion—drove prediction. These findings highlight AI as a promising tool for postoperative risk stratification in PanNETs.

The study involves quantitative assessment of human pancreatic intraepithelial neoplasia (PanIN) using CODA, a novel machine-learning pipeline for 3D image analysis that generates quantifiable models of large pieces of human pancreas with single-cell resolution. By leveraging CODA, for the first time, we are able to describe the spatial and genetic multifocality of human PanINs, providing important insights into the initiation and progression of pancreatic neoplasia.

This study reevaluates the conventional clinical classification of precursor lesions in Pancreatic Ductal Adenocarcinoma (PDAC), namely PanIN (Pancreatic Intraepithelial Neoplasia) and IPMN (Intraductal Papillary Mucinous Neoplasms). It reveals the inadequacy of the traditional classification when assessing high-resolution 3D images. By constructing cellular-resolution 3D maps, the research uncovers instances where some PanINs grow to macroscopic sizes in narrow ducts, meeting the criteria for IPMNs.

The study focuses on developing a Scaffold-supported Platform for Organoid-based Tissues (SPOT) platform compatible with generic 96/384 platforms, using 3D design and bioprinting techniques. SPOT enables a 3D heterogeneous environment that mimics the tumor microenvironment, serving as a useful tool to assess the drug sensitivity of patient-derived organoids and can be easily integrated into the drug discovery pipeline.

The study involves the construction of 3D, multi-organ volumes of the normal reproductive system of mice at different age groups using CODA, a novel machine-learning pipeline for 3D image analysis, to qualitatively and quantitatively identify major changes to ovary, fallopian, cervix, and vagina morphology and immune infiltration with age.

The study focuses on the distinct adhesion mechanism employed by freshwater mussels. By quantitatively analyzing localization profiles and post-translational modifications of the adhesive proteins using a staining-based approach, this work provides a well-established role model for developing novel bioadhesives such as self-healing materials and advanced coatings in aqueous environments.

Designed and developed a library of landing pad cell lines targeting pSH231 and AAVS1 safe harbor sites. The library provides an efficient approach for recombinase-mediated cassette exchange with a specific promoter and gene of interest.

Designed and simulated a device that prepares the subtalar joint for fusion during tibiotalocaneal arthrodesis, which consists of 2 modular components: a cutting tool and a guide tube.

Discovered great potential in the combination of platelet inhibition and natural killer (NK) cell enhancement, where platelet inhibition sensitizes tumor cells to NK cell activity and NK enhancement increases cytotoxic effects.

Engineered C. albicans to express GFP with specific genes and developed a statistical method for quantifying this gene expression following long-time space mission.

Designed and fabricated of a fully autonomous mobile robot that can deploy miniature traffic cones based on the detection of "cracks" and "holes". The robot was constructed with omnidirectional mecanum wheels, infrared sensors, as well as an interactive LCD and keyboard user interface.