Seminar Series: Nadia Lanman

Nadia Lanman, research associate professor at the Purdue Institute for Cancer Research (PICR) and manager of the Collaborative Core for Cancer Bioinformatics, will be a speaker for the PICR Seminar Series.

Toward Building Cancer Patient Digital Twins: Learning Cellular States and Enabling Scalable Annotation With AI

Cancer patient digital twins aim to model disease progression and therapeutic response at the level of individual patients, but achieving this goal requires both biologically meaningful representations of tumor state and scalable approaches for generating high-quality cellular data. We present two complementary AI-based efforts that address these challenges at the cellular scale and represent steps toward building digital twins for cancer patients.

The first, scMeta, is a graph-based deep learning framework that learns latent metastatic states from single-cell RNA sequencing data. Using a pan-cancer atlas of malignant cells across multiple tumor types, scMeta demonstrates that single-cell transcriptomes contain reproducible signals associated with metastatic potential. Learned representations organize cells along biologically interpretable trajectories, consistent with a continuum of disease states, and enable improved patient-level risk stratification across studies.

The second effort, AnnotateAnyCell, addresses a key practical bottleneck in cellular-scale modeling: the need for expert annotation. This semi-supervised, human-in-the-loop framework integrates cell segmentation, latent space visualization, and active contrastive learning to substantially reduce pathologist annotation time while achieving expert-level accuracy for well-defined nuclear morphology features in whole-slide histopathology images. Together, these efforts highlight how latent cellular-state modeling and scalable expert-guided annotation can support the long-term development of cancer patient digital twins, while underscoring the remaining challenges in integrating multimodal data into predictive patient-level models.

Computational biology expert Nadia Lanman helps cancer researchers distill solutions from massive datasets. (Purdue University photo/Kelsey Lefever)

This seminar is hosted by Majid Kazemian.