There is no reason the underlying idea of nanotainers being used for blood analysis is fundamentally impossible but it is a hard problem because the volume of blood is so small there can easily be statistical fluctuations that affect the classifier. That said, it's strictly correct that, if engineered properly, it could have provided additional actionable medical data.

> There is no reason the underlying idea of nanotainers being used for blood analysis is fundamentally impossible but it is a hard problem because the volume of blood is so small there can easily be statistical fluctuations that affect the classifier.

Do you not see how the second half of this statement strictly contradicts the first?

Hey, um, I've got a PhD in Biophysics, I've worked with biology, and I've got 20+ years experience in the field. The VC I advise declined to fund theranos for this very reason. That said, everything I know about biophysics and building classifiers (something I do for a large company known for machine learning) says that it's still technically possible to provide actionable medical data from tiny volumes of blood.

It does not "strictly contradict".

Difficult doesn't mean impossible.

What they promised is theoretically possible, but they ran afoul of the huge practical problems that made people with actual biology and engineering backgrounds (like myself) very, very skeptical.

it's not theoretically possible, due to the statistics of small numbers. You can resolve the nanotainer problem by collecting many of them, but by the time you can measure a statistical difference, you might as well have collected a large volume of blood in the first place.

I agree that analyzing small volumes of blood is theoretically possible , but afaik it's yet unclear that you can get good samples from blood drawn from fingers, because it's mixed with interstitial fluid.

(Oh, and massive financial shenanigans.)