Case study · rheumatoid arthritis

Rheumatoid arthritis, read in patient-derived cells.

BioGPU compared fibroblast-like synoviocytes (FLS) from a rheumatoid arthritis patient with FLS from a healthy donor, using molecular-specific visualizers. The cellular system captured known RA biology, and the same experiments pointed to activities beyond it.

Category: Target Biology Resolution
Cellular system: Primary FLS · healthy donor vs RA patient
Readouts shown: 18 visualizers in 8 panels

Illustration of a healthy joint (left) and a rheumatoid arthritis joint (right), with immune cells in the inflamed synovium — Healthy joint · RA joint (illustration). FLS line the synovium, beside chondrocytes (cartilage), osteocytes (bone) and immune cells.

Evidence · experimental images

Same visualizers, two donors, visible differences.

Choose an activity to compare the healthy donor (WT) with the RA patient (RA). Every pair is an experimental image from the study.

CSF2, healthy donor FLS — **CSF2**Inflammatory response

CSF2, RA patient FLS — **CSF2**Inflammatory response

Activities were measured in parallel wells, not in one image.

BioGPU in action

From cellular events to statements AI can reason on.

ImageHigh-content imaging with molecular-specific visualizers
Single cellsEach cell's phenotype, then vectorized
ClusterCells grouped by shared visual patterns
ProfileHow each group shifts between conditions
“Strong diffused cytoplasmic staining …”DescribePhenotype descriptions and scores, for biological interpretation

The Visual Biology Foundation Model recognizes and describes image-level phenotypes; biological reasoning across findings happens in AI agents.

One finding as a Visual Biology Token

Src moves into the nucleus in RA cells.

Src, healthy donor FLS: cytoplasmic staining — WT · healthy donor

Src, RA patient FLS: nuclear staining — RA · patient

Image evidence · model description

Test cells show a dramatic shift in staining patterns compared to reference cells. The most striking change is a massive increase in cells displaying moderate diffuse nuclear staining with rounded dark nuclear holes accompanied by moderate granular cytoplasmic staining, indicating the emergence of a phenotype virtually absent in the reference group. Concurrently, there is a near-complete loss of cells exhibiting weak diffuse nuclear staining paired with moderate diffuse cytoplasmic staining, a phenotype that predominates in the reference population.

Biological meaning · interpretation

Non-phosphorylated Src translocated in RA-FLS cells from the cytoplasm to the nucleus. Nuclear, inactive, Src scaffolds the splicing machinery, converging with the alternative splicing hypothesis implied by SRPK1-SRSF activation.

Experimental contextSrc · primary FLS · healthy donor (reference) vs RA patient (test)

What this demonstrates

A patient's disease biology, captured and read in one cellular system.

Shown in this study

A patient-derived RA cellular system run on BioGPU beside a healthy donor.
Visualizers read inflammation, metabolism, citrullination, MAPK signaling, splicing, migration and mRNA translation in the same system.
Known RA biology appears in the images, and new activity patterns appear beside it.
The foundation model turns a phenotype shift into a description, kept with its images.

What it could inform · not shown here

Which targets to prioritize in disease-relevant cells.
A disease fingerprint that compounds can be read against.
Not shown: causality, therapeutic effect, or results across many donors.

See how a program runs → Explore cellular systems →

Rheumatoid arthritis, read in patient-derived cells.

Same visualizers, two donors, visible differences.

From cellular events to statements AI can reason on.

Src moves into the nucleus in RA cells.

A patient's disease biology, captured and read in one cellular system.

Resolve your target in disease-relevant cells.