Skip to content
BioLayers SI
BioLayers SI
Cancer Research Infrastructure

From papers to living cells

BioLayers SI connects biomedical literature, biological mechanisms, real microscopy, and cellular phenotypes into evidence-grounded maps for cancer research.

Every statement carries its source: a PubMed ID, a dataset accession, or a measurement you can check in the image yourself.

Map a paperBrowse the cell atlas

On 200 benchmark images, the segmentation model matched 22,212 hand-annotated nuclei with a pixel overlap (IoU) of 0.942. Read the validation report

Fluorescence micrograph of about a hundred U2OS cell nuclei stained with Hoechst.
  • No flag 77
  • Abnormal shape 4
  • Possibly dying 2
  • Cut by image edge 25

Nucleus 41

Observed in the image
Area
282 µm²
Circularity
0.72
Solidity
0.94
Inferred by a rule or model

No flag raised. Whether it migrates or divides needs a time-lapse, not one frame.

Model score 0.96. This is Cellpose’s own confidence, not a calibrated error rate.

U2OS nuclei, Hoechst stain, 20×, shown with a blue lookup table. Image from BBBC039 (Broad Institute, CC0). Outlines by Cellpose with no fine-tuning. Click a nucleus to read it.
Trusted by

Guided by leading cancer researchers

Mentors, partners and press behind BioLayers SI.

  • MD Anderson Cancer CenterMentor
  • Weill Cornell MedicineMentor
  • Nationwide Children's HospitalMentor
  • The Ohio State UniversityMentor
  • Hollings Cancer Center, MUSCMentor
  • NXT HorizonPartner
  • STEMisePartner
  • BioMedizonePartner
  • Computational Biologists Society at UCLAPartner
  • HundrEDFeatured by
  • AI BusinessAI BusinessFeatured by
  • MD Anderson Cancer CenterMentor
  • Weill Cornell MedicineMentor
  • Nationwide Children's HospitalMentor
  • The Ohio State UniversityMentor
  • Hollings Cancer Center, MUSCMentor
  • NXT HorizonPartner
  • STEMisePartner
  • BioMedizonePartner
  • Computational Biologists Society at UCLAPartner
  • HundrEDFeatured by
  • AI BusinessAI BusinessFeatured by

Mentors are affiliated with these institutions as individuals. Listing does not imply institutional endorsement.

One mechanism, followed from paper to cell

Our flagship case links a glioblastoma cell we tracked for 28.5 hours to the CXCL12/CXCR4 signalling axis. Each step shows where its evidence comes from, and the last link is a hypothesis, not a measurement.

  1. Fibroblasts in tumours secrete CXCL12

    From published literature

    Carcinoma-associated fibroblasts secrete elevated CXCL12, which acts through CXCR4 on carcinoma cells.

    Shown in breast carcinoma, not glioma.

  2. CXCL12 acts on CXCR4 on glioma cells

    From published literature

    CXCR4 is the main chemokine receptor on 13 of 16 human glioma cell lines, and CXCL12 makes those cells move toward it.

    Glioma cell lines and resected tumours, in vitro.

  3. CXCR4 drives glioma invasion

    From published literature

    Invasive glioma cells overexpress CXCR4, and blocking it impairs invasion. CXCL12 and CXCR4 rise with tumour grade.

  4. This cell migrates

    Observed in the image

    Glioblastoma cell U373-02-T7, tracked for 28.5 hours: 175.9 µm net displacement, 0.193 µm per minute, directionality 0.53.

Flagship figure: a phase-contrast frame of U373 glioblastoma cells with the tracked cell outlined, its migration path, measured phenotype, and the CXCL12/CXCR4 literature links.
The image, cell, phenotype and literature chain for cell U373-02-T7, produced by the BioLayers pipeline.
Flagship figure: a phase-contrast frame of U373 glioblastoma cells with the tracked cell outlined, its migration path, measured phenotype, and the CXCL12/CXCR4 literature links.

The image, cell, phenotype and literature chain for cell U373-02-T7, produced by the BioLayers pipeline.

What this does not show

We did not measure CXCR4 in this cell. Many cell stocks labelled U-373 are actually the line U-251 (Torsvik et al., 2014). The link from this cell to CXCR4 is a hypothesis the literature makes plausible.

What you can do here

Mind map

Upload a PDF or DOCX and get a structured map of its mechanisms, each linked to the sentence it came from. Uses your own Gemini API key.

Map a paper

Cell atlas

634 catalogued microscopy records from the Broad Bioimage Benchmark Collection and the Cell Tracking Challenge, each with its source and licence.

Browse the atlas

Cipher

Cause-and-effect maps of cancer pathways such as KRAS G12D resistance, readable in plain language or with full terminology.

Open Cipher

Library

Mind maps that other researchers have saved, newest first, with a page for each author.

Read the library

New to the biology? Take the 3D journey from a paper to a mechanism.

Three kinds of statement, never mixed

Observed in the image
Measured from segmented pixels: area, shape, intensity, speed.
From published literature
Checked against a PubMed abstract, with the system the finding was made in.
Inferred by a rule or model
A label from a transparent rule or a model score. Useful, never an experimental fact.

The cell atlas holds no AI-generated microscopy. Every record names its public dataset, accession and licence, and records we cannot verify are marked for review.

Built with partners

BioLayers SI started in Tashkent and is guided by oncologists and computational biologists. Featured by HundrED and AI Business.

All partnersTeam and mentors

Start with one paper

Upload a PDF or DOCX. You get a map of its mechanisms, with the sentence each one came from.

Map a paperSee published maps