Chemspider and Structure of Digitonin

Antony Williams of Chemspider posted an interesting challenge for the community to work out, and decide upon, the structure of digitonin. This molecule is a “saponin” meaning a steroid with sugars attached. The details are on the Chemspider blog and there is a discussion on Friendfeed.

The structure was reported (in German) in Tetrahedron in 1963, with an NMR spectroscopy study being published some years later. This latter paper gives the structure as this:

Published digitoninstructure

I have not gone through the NMR reasoning to check the conclusions. It sounds like they analysed a pure compound, not a mixture: “The seed of Digitalis purpurea L. contain ca. 2% of a saponin mixture whose main component is digitonin. With preparative distribution chromatography on formamide-saturated cellulose, it was possible to obtain 1 in pure form.” The paper’s been cited 12 times, and from the titles it doesn’t look as though there has been a structure revision. Chemspider lists three structures at the moment, which all differ in some aspect of the stereochemistry:

Chemspider structures

Circled are a few anomalies. These structures are actually rotated through 180 degrees to depict the steroid skeleton in (approximately) the usual sense. It’s nice to be proper about these things.

To make the comparison between CS and the literature  easier, we can number the structure and redraw the literature structure in this orientation. It turns out that the Chemspider structures have the steroid ring flipped, so we can flip that too:

Lit structure (top), Chemspider orientation (bottom left) and lit structure redrawn (bottom right)

So, which is right? It looks like none of them. Structure 4976216 is close, but the stereochemistry at position 20 (red methyl) looks to be inverted (in all structures). The literature paper is unequivocal about the stereochemistry there.

Digitonin fragment from paper - stereochem at position 20

Interestingly the varied depictions of the spiro moiety can make a comparison confusing. I included two representations in the above figure, but the Chemspider structures have three different depictions of this. They’re all the same, in fact.

Same stereocentre, different spiro drawing

So I’d suggest “digitonin” means a pure compound, with the literature structure as shown above.

The question that’s interesting is why do the CS structures (which are derived from other sources, clearly, and is no fault of CS) differ? Where do the anomalies creep in? If you look at the Wikipedia structure, that stereocentre is also wrong.

Digitonin Structure from Wikipedia

This picture was drawn using “http://www.commonchemistry.org” as source, but I haven’t gone back to check for sources for that. I’m guessing it’s a chemical case of what we used to call ‘Chinese Whispers’ where what starts out as one thing becomes something subtly different after being passed between people. I asked my wife about this, and she thinks it’s a virus (Epimerizer) that’s sweeping the web, randomly inverting stereocentres in chemical databases. (She’s an anthropologist, so doesn’t understand what a nightmare this idea is.)