BBBatlas

Which drugs actually cross the blood–brain barrier?

Permeability predictions built from human single-cell RNA-seq of the transporters that guard the barrier — efflux pumps, solute carriers and RMT receptors — plus how each one shifts in Alzheimer’s and Parkinson’s disease.

Research prototype — hypothesis-generating, not clinical or diagnostic. Data updated Jul 28, 2026. Open the tool ↓

How a verdictis actually made.

No black box. Every prediction is three transparent steps over published human data, and each step is inspectable in the tool.

  1. 01

    Map the drug to its transporters

    Each of the 38 drugs is linked to the transporter genes it is a documented substrate or inhibitor of, from a literature-curated list. Drugs with no well-established transporter are shown as a gap, not guessed at.

  2. 02

    Read those genes in real barrier cells

    Expression is taken from brain capillary endothelial cells in three independent human single-cell datasets, then averaged and converted to a percentile rank across the 26-gene panel.

  3. 03

    Weigh influx against efflux

    A passive-diffusion estimate from the molecule's own properties is combined with influx and efflux terms weighted by measured transporter expression. A clear lean gives permeable or restricted; a close call is inconclusive rather than forced, and cross-dataset disagreement is flagged.

Disease shifts in Alzheimer’s and Parkinson’s are reported separately, with the fold change and adjusted p-value shown rather than hidden behind a label.

Shared and used by

Researchers who have shared BBBatlas with their labs or used it themselves.

12
researchers have shared or used BBBatlas
11
universities and research institutions
7
sent substantive feedback or corrections

Counted from direct correspondence. Each entry describes what that researcher actually did, in their own account: shared it with a lab, ran queries against it, or sent feedback. Several entries exist because the feedback was critical, which is the point. Anyone listed who would prefer not to be can email me and I will remove them the same day.

Institutions
MIT Stanford University Duke University Brown University UNC Chapel Hill University of Washington University of Wisconsin–Madison Mount Sinai Memorial Sloan Kettering Mayo Clinic University of Kentucky
Researchers

If you have used BBBatlas and would be willing to be listed here, or if you have found something wrong with it, I would be glad to hear from you.

Ready when you are.Open the atlas.

Search any of 38 drugs or 26 transporter genes, browse expression across seven vascular cell types, or compare shifts in Alzheimer’s and Parkinson’s disease.

Transporter atlas

Search a drug or transporter gene, browse expression across seven vascular cell types, or compare disease shifts. Every verdict traces back to measured expression in human brain capillary endothelial cells.

All Transporter Genes
Class:
Gene Class Cap EC % Expressing Avg Expression Consistent Disease Hits Drugs Actions
Bars in the Avg Expression column show per-dataset capillary-EC expression.
Not yet queried in these datasets. This entry corrects an invalid original panel symbol, SLCO1A4 — a rodent-only (mouse/rat) gene symbol, not a human gene. The correct human ortholog is SLCO1A2 (OATP1A2), a documented human BBB statin transporter. Candidate addition for a future data pull. See “Methods & Known Limitations” below for detail.
All Curated Drugs
Verdict:
Global Expression Heatmap
26 genes × 7 cell types × 3 datasets — Capillary EC column is the basis for permeability scoring. Hover cells for exact values.
Dataset: Sort by:
Low
High — mean log-normalized expression (per-dataset scale)
Disease Differential Expression (AD / PD)
Disease:
⚠ Interpretation column is inferred Log₂FC and Padj are measured. CNS exposure predictions are model-derived heuristics, not validated PK.
GeneClassDiseaseDatasetCell Type DirectionLog₂FCPadj Predicted CNS Impact INFERRED

Showing all genes tested; ★ = statistically significant (Padj < 0.05 and |Log₂FC| > 0.25). Full dataset includes both significant and non-significant results.

Side-by-Side Comparison
Compare two genes or two drugs across their key data — expression, verdict, and disease shifts — without switching back and forth.
vs