MahnLab Organoid analysis with 3D holotomography, toward virtual biology

Can we tell, before treatment, whether a patient’s immune system will attack their cancer? We move from bedside to bench to digital and back, to predict immunotherapy response in lung cancer.

1. 

   01

   Building TOME

   TOME stands for Tumor Organoid MicroEnvironment. From a single lung cancer biopsy or malignant pleural effusion, we dissociate and size-filter tumor tissue and embed it in extracellular matrix to grow patient-derived lung cancer organoids. We confirm with single-cell transcriptomics that the organoids keep the tumor identity of the original tissue.

   From the same patient’s blood we isolate PBMCs, sort and expand autologous T cells, and add them back to the organoids. The result is a personalized tumor–immune system in a dish, where we can watch how a patient’s own immune cells respond to their own cancer under immune checkpoint inhibitors.

2. 

   02

   Multimodal data integration

   Conventional assays look at fixed, two-dimensional tissue at a single time point. We image living TOME cultures with label-free 3D holotomography, which measures refractive index instead of stains, and follow tumor organoids and infiltrating lymphocytes for hours during checkpoint blockade.

   Segmentation and tracking models capture how immune cells approach, contact and attack the tumor. A 3D deep learning model turns these morphodynamics into latent features, which we connect to single-cell RNA sequencing of resting and activated T cells and to each patient’s clinical response, placing image, transcriptome and outcome in one shared space.

3. 

   03

   Virtual organoid

   Organoid experiments are powerful but slow and costly, and each patient sample is limited. We use the integrated imaging, transcriptomic and clinical data to train a virtual organoid: a digital twin of each patient’s tumor and immune microenvironment.

   In this virtual model we can simulate immunotherapy, compare treatment options and predict whether a patient is likely to respond, then check those predictions against real organoids and clinical outcomes. Our goal is a loop from bedside to bench to digital and back to the bedside, giving clinicians an answer before treatment begins.