From abstract to a mechanism figure you can check
Paste a biomedical abstract. Cellfig draws the pathway, ties every line to the sentence behind it, and flags what the text does not say. You export real text, not a picture of text.
Check against your text
To review (3)
-
Mechanistically, PGC1α cooperates with estrogen-related receptors (ERRs) to regulate chromatin accessibility and gene transcription.
The text says cooperates with, not binds
-
Among these, Gys2 and Ppp1r3c are rapidly induced following acute β3-adrenergic receptor activation.
The text never says β3-AR activates PKA. It names PKA-CREB signaling separately, so the line stays dashed
-
In contrast, sustained expression of glycogen metabolism genes under chronic β3-adrenergic activation requires the coactivator PGC1α, whose loss blunts glycogen accumulation and thermogenic capacity.
The text says the program requires PGC1α under chronic activation, not that β3-AR activates PGC1α directly
Matches the text (7)
-
We identify Gys2 as a direct transcriptional target of PKA-CREB signaling.
-
We identify Gys2 as a direct transcriptional target of PKA-CREB signaling.
-
Among these, Gys2 and Ppp1r3c are rapidly induced following acute β3-adrenergic receptor activation.
-
Chromatin profiling confirm that ERRs directly control the glycogen metabolic program in beige adipocytes.
-
Although deletion of ERRα is compensated by ERRγ, combined deletion of ERRα/β/γ abolishes expression of glycogen metabolism and thermogenic genes.
-
…the coactivator PGC1α, whose loss blunts glycogen accumulation and thermogenic capacity.
-
…the coactivator PGC1α, whose loss blunts glycogen accumulation and thermogenic capacity.
Abstract: “Multi-layered transcriptional control of glycogen metabolism coordinates thermogenic remodeling of white adipocytes in male mice”, Nat Commun 2025, doi:10.1038/s41467-025-67515-9, CC BY 4.0. The relations were extracted from this abstract by Cellfig. We confirmed the flags on this example by hand, and the figure is ours, not the authors’.
The figure stays answerable to the abstract
An omission leaves nothing on the figure to catch, so Cellfig also marks the text it left out.
Thermogenic activation of subcutaneous white adipocytes requires glycogen synthesis and turnover1. Here we show that β-adrenergic stimulation induces a distinct glycogen metabolism gene program in inguinal white adipose tissue2 in a cell-autonomous and adipocyte-specific manner. Among these, Gys2 and Ppp1r3c are rapidly induced following acute β3-adrenergic receptor activation. We identify Gys2 as a direct transcriptional target of PKA-CREB signaling. In contrast, sustained expression of glycogen metabolism genes under chronic β3-adrenergic activation requires the coactivator PGC1α, whose loss blunts glycogen accumulation and thermogenic capacity. Mechanistically, PGC1α cooperates with estrogen-related receptors (ERRs) to regulate chromatin accessibility and gene transcription3. Although deletion of ERRα is compensated by ERRγ4, combined deletion of ERRα/β/γ abolishes expression of glycogen metabolism and thermogenic genes. Chromatin profiling confirm that ERRs directly control the glycogen metabolic program in beige adipocytes. Together, our results identify a multilayered transcriptional axis that sustains glycogen metabolism during β-adrenergic activation in male mice.
Left out of this figure (4)
- Glycogen synthesis is required for thermogenic activation
- β-adrenergic stimulation induces glycogen metabolism genes. This summarizes the two arms already drawn
- PGC1α with ERRs regulates chromatin accessibility
- ERRγ compensates for deletion of ERRα. No relation was extracted for this one
Add any of them to the figure, or leave them out and know that you did.
Exported text stays text
Labels leave Cellfig as text objects, not outlines. Open the SVG in Illustrator or Inkscape, or the file in PowerPoint, and edit a label the way you edit any other text. Icons are separate groups.
Before you download, Cellfig counts the text on the figure against the text in the file, so you know nothing was flattened on the way out.
- SVG
- Real text, embedded icons, no external links
- PowerPoint
- Editable text and icons
- PNG
- Up to 300 dpi at 180 mm wide
<g class="node" data-node="pka">
<g class="icon">…</g>
<text font-size="13" font-weight="500">
<tspan x="168.2" y="321.3">PKA</tspan>
</text>
</g>
<g class="node" data-node="gys2">
<rect class="gene-bar" … />
<text font-size="13" font-style="italic">
<tspan x="168.2" y="573">Gys2</tspan>
</text>
</g>
Every icon records where it came from
An icon has one of three sources, and the source travels with it into your export. The AI use statement and the credits are assembled from the icons actually on your figure.
-
β3-AR
Drawn from structural parameters, 7 transmembrane helices
-
GLUT4
Drawn from structural parameters, 12 transmembrane helices
-
Macrophage
Open library: Servier Medical Art, CC BY 3.0, credit included
-
AKT
AI-generated, and labeled that way everywhere it appears
Use non-AI icons only is one switch on the figure. It swaps each AI-generated icon for a version drawn from parameters or from an open library, and it names any icon that has no such version. It covers icons only, so the statement still reports the AI used for extraction and layout.
Journal rules differ, so check your target journal’s policy before you submit.
Entities and relations were extracted from the source text with AI assistance (Cellfig), and the layout was generated automatically. The authors have not yet checked the relations against the text. Icons: 5 drawn from structural parameters.
Built only for biomedical mechanisms
Cellfig does one kind of figure, so it can follow the conventions your readers already use.
- Lines carry the meaning
-
- Activates
- Inhibits
- Binds
- Phosphorylates
- Translocates
- Not stated in the text
Style, never color, so a figure still reads in grayscale print.
- Gene and protein type
Gys2 is a mouse gene, so it is italic. PKA is a protein, so it is roman.
- Official symbols
The label says β3-AR. The record keeps Adrb3, so you can see which entity Cellfig took it to be.
- Compartments
Receptors sit in the bilayer, targets on the DNA line, and transcription factors inside the nucleus. A second layout gives you the same mechanism as a linear chain.
Questions
Can I use it today?
Not yet. Cellfig is being built now, and the figure on this page comes from our working prototype. Leave your email and we will write when you can try it on your own abstract.
What does the extraction get wrong?
It can miss a relation or type one wrongly, as it did with PGC1α and ERRs above. That is why the first draft is never the last word: each line shows its sentence, and text left out of the figure stays marked.
Will my target journal allow a figure made this way?
We cannot answer that for you. Journal rules differ and they change. Cellfig gives you a record of how the figure was made and where each icon came from, so you can follow your journal’s disclosure policy.
What does it cost?
Pricing is not set yet. People on the early access list will hear first.