SpatialX

Overview

SpatialX is an integrated platform for multi-image, multi-technology spatial data (CosMx, Xenium, Visium). It provides algorithms and tools for in-context analysis, optimized computation, and an updated cell-annotation database.

In less biology-heavy language: SpatialX is BioTuring’s product for exploring and analyzing tissue spatial data through an interactive, visual interface.

My role sat between product design and front-end engineering, which let me move quickly from problem framing → UI decisions → implementation.

This page isn’t a full case study yet, it’s a work log while I’m still actively working on the product. I joined the project in its early days, so there have been many iterations, and we’re continuing to add features.

Application features (non-biologist explanation)

At a high level, the app is structured like this:

• There are multiple studies.
• Each study contains samples.
• Each sample contains data records.
• Each data record includes images and cell information. Each data panel can be viewed in Image view or Cell view.

There are also several tools to analyze cells and images:

• Cell type prediction: predict cell types using our database or a user-provided data source for all cells in the current dataset.
• Differential gene expression: compare gene expression between two cell selections and generate a report.

  

  

• Marker genes: identify genes that stand out in a selected region.
• Proteomics: view images by protein channels, with the ability to adjust thresholds.

and more. See the work log to track what I’m working on.

Challenges

This has been one of the most complex products I’ve worked on. It was my first time working on a canvas-based application, and it required handling very large data responses and graphics-heavy implementation.

The biggest challenge was the biology and bioinformatics concepts—they were difficult to grasp at first. I had many discussions with product specialists (with biology backgrounds) to build understanding.

On the engineering side, I learned to use Deck.gl to create a Google Earth–like experience. Along the way, I also had to pick up a lot of math and new concepts through self-study.

Related posts

Building a Figma-like Transformation Layer for DeckGL: Harder Than It Looks | AU DUONG TUAN

Building object transformation controls seemed straightforward—just drag to resize, grab a corner to rotate. How hard could it be? Complex math, custom cursors, and deep dives into coordinate transformations. Sometimes simple-looking features are the hardest to build right.

Work log

Coming soon…