CellPrism: A Visual Analytics System for Exploring AI-Driven Virtual Cells in Drug Discovery

Authors

Chuhan Shi (Southeast University), Zijian Guo (Southeast University), Zelin Zang (AI), Chengbo Zheng (The University of Queensland), Ding Ding (Southeast University), Rui Sheng (The Hong Kong University of Science and Technology)

Presentation

Session

What does it mean to live, anyway?

Time

Wednesday, Nov 11, 11:00 – 11:12 (US/Eastern) · session 10:00 – 11:30

Location

Hall America center

Keywords

Drug discovery, virtual cell, gene expression data, human-AI collaboration

Abstract

Gene perturbation analysis plays a critical role in drug discovery by enabling researchers to investigate how interventions on specific genes influence global gene expression patterns within cells. Recent advances in artificial intelligence (AI)-driven virtual cell models have made it possible to predict gene expression outcomes for a wide range of perturbation strategies in silico, substantially reducing reliance on costly and time-consuming biological experiments. However, effectively exploring and interpreting the high-dimensional perturbation spaces produced by these models remains challenging due to the combinatorial nature of perturbations and the complex, cell-specific gene expression responses they generate. In this work, we present CellPrism, a visual analytics system designed to support the systematic exploration of gene perturbation strategies for drug discovery. Specifically, CellPrism integrates clustering-based overviews to summarize perturbation outcomes, a glyph-based representation to compactly encode gene expression patterns across cell types, and coordinated views that enable fine-grained comparison and interpretation of perturbation effects. We demonstrate the effectiveness of CellPrism through a real-world case study and expert interviews. This work highlights the value of visual analytics in bridging virtual cell modeling with expert-driven decision-making in drug discovery.

For Practitioners

Computational biologists, bioinformaticians, single-cell researchers, and drug discovery scientists who use virtual cell models or gene perturbation data would be particularly interested in this paper. Visual analytics researchers and practitioners developing AI-assisted tools for biological discovery may also find it useful. Practitioners can apply the proposed workflow to progressively screen observed genes, navigate large combinatorial perturbation spaces, compare predicted effects across target and protected cell types, and balance therapeutic efficacy against potential side effects. They can also use feature-contribution information as a retrieval and pruning cue to identify influential genes, refine multi-gene perturbations, and prioritize promising candidates for subsequent wet-lab validation. More broadly, the hierarchical exploration and coordinated comparison mechanisms can inform the design of visual analytics tools for other model-driven biological discovery tasks.