Biomarker & Cell Type Discovery
Summary
Researchers needed a seamless way to explore biomarkers and cell-type information, but previously had no way to browse, compare, or interpret this data across datasets.
I led the 0→1 design of a new discovery experience by defining the information architecture, workflows, and visualizations while navigating complex, multi-API data constraints.
The result is a cohesive system that makes biomarker and cell-type exploration clear, comparable, and accessible directly within the portal.
Role
UX Research, UI Design, Data Visualization
End Users
Experimental & Computational Biologists
Project Duration
2021 - Present
Highlights
2025: scellop Preprint
Challenge
From Raw Data to Discoverable Insight
Researchers struggled with:
Biomarker and cell-type information buried inside individual datasets that require download or convoluted navigation.
No way to browse, compare, or understand these entities across multiple datasets.
Reliance on downloading data or manual inspection to answer basic biological questions.
The core challenge:
How do we create a unified, intuitive way for researchers to explore biomarkers and cell types, without downloading data, switching tools, or navigating dataset by dataset?
Key Features Designed
I designed a cohesive biomarker and cell-type discovery system that brings together a stand-alone advanced search tool with newly created pages for exploring genes, proteins and cell types. The final experience creates a clear, connected model for navigating molecular relationships that were previously buried inside individual datasets.
Advanced Biomarker & Cell Type Search
Advanced query interface for datasets containing specific genes, proteins, pathways, or cell types. Supports multiple query methods across RNA-seq, ATAC-seq, and proteomic data, returning statistical summaries and interactive visualizations.
Comprehensive Biomarker Pages
Detailed gene profiles aggregating descriptions with external references to HUGO and Ensembl, anatomical context, associated cell types, and expression visualizations across datasets.
Cell Type Explorer
Detailed cell type pages showing descriptions, distribution across organs, key marker genes, and associated datasets. Interactive visualizations compare cell type distributions across organs.
Design Timeline
2021: Foundation & Early Research
Conducted interviews, competitive analysis, and early use-case mapping for biomarker and cell-type exploration.
Validated HuBMAP’s Cells API queries to ensure they supported meaningful biological questions researchers needed to answer.
Identified the need for advanced molecular search capabilities.
2022, MVP Design & Implementation
Designed and delivered the MVP molecular & cellular query interface.
Collaborated with the API team to validate feasibility of planned biological queries.
Presented the design process and MVP at the International DESIGN Conference.
2023, Biological Knowledge
Collaborated with the UBKG (Unified Biomedical Knowledge Graph) API team to provide biological context to data exploration.
Validated gene, protein and cell type pages, and cell population plot visualizations with biologists.
Released biomarker profile pages (beta) with descriptions, external references (e.g. HUGO, Ensembl), and visualizations.
Cell type pages deferred due to API limitations at the time.
2024 - 2025 Full System
Partnered with the scFind API team to integrate statistical marker gene discovery and gene to cell type significance, fulfilling a major user need that earlier APIs could not support
Implemented complete biomarker and cell type pages with unified navigation, visualizations, and dataset linkage.
Delivered the final end-to-end discovery experience connecting biomarkers → cell types → tissues → datasets.
Impact
Made molecular and cellular context accessible without downloading data
Enabled new scientific workflows (marker-gene discovery, cell-type significance) that were previously impossible
Delivered cross-dataset visualizations that connect biomarkers, cell types, tissues and metadata, giving researchers a unified way to understand molecular context without downloading data.
