Why caveolae are biology's drug-delivery pumps

Most drugs don't fail because they're too weak. They fail because they never reach the tissue where they're needed in a high enough concentration to work — while flooding everywhere else at a dose high enough to cause side effects. That gap between "works in a dish" and "works in a patient" is where most therapeutic programs quietly die.

Thirty years ago, Jan Schnitzer identified one of the few mechanisms nature actually built to solve this problem: caveolae. Caveolae are tiny flask-shaped pits on the surface of endothelial cells — the cells that line every blood vessel in the body. Far from being passive structures, caveolae actively pump material across the vessel wall, from blood into tissue, using an energy-dependent transport process called transcytosis (Schnitzer et al., J Cell Biol, 1994).

That single observation reframed a question the field had been asking for decades. Instead of "how do we make a drug potent enough to work despite poor tissue penetration," the real question became: can we hijack a mechanism the body already uses to actively pump cargo across the vessel wall?

The answer, published in Nature in 2004, was yes. By mapping the surface proteins uniquely exposed on the blood-facing side of the endothelium — a technique now core to Cavatar's AVAPROT engine — the team identified proteins concentrated specifically in caveolae, at specific tissue interfaces (Oh et al., Nature, 2004). Three years later, live dynamic imaging confirmed the mechanism could be exploited directly: an antibody engineered to bind one of these caveolae-enriched proteins was pumped rapidly and specifically across the lung endothelium, reaching target tissue in minutes (Oh et al., Nature Biotechnology, 2007).

This is the mechanism behind every program Cavatar has built since: an antibody engineered against a caveolae-enriched surface protein doesn't wait to leak through gaps in the vessel wall. It gets pumped through, on purpose, by biology that's been doing this job since before drug delivery was a field.

Why this matters for target discovery, not just drug design

Most target-discovery approaches — including AI-driven ones — score a candidate protein on a single axis: is it associated with the disease? That answers "is this protein interesting," not "can a therapeutic physically reach it." Caveolae biology adds a second, harder axis: is this protein sitting on an accessible route across the vessel wall, at a tissue interface a drug can actually use?

That's the axis AVATAR was built to score. Every candidate our models rank is evaluated not just for disease relevance, but for whether it sits on a real, physically validated transport route — the same caveolae-mediated pathway Jan Schnitzer first characterized three decades ago. It's the difference between a target that looks promising on paper and one a therapeutic can actually be delivered to, at a dose safe enough to use.

Read the original research: Schnitzer et al., J Cell Biol (1994) · Oh et al., Nature (2004) · Oh et al., Nature Biotechnology (2007)