When I came to US to join a small startup biotech company, my first assignment was to re-synthesize an active literature compound - a repulsive-looking peptidomimetic macrocycle. The compound was to be used in validating a new bioassay; the father of our young CEO needed the compound for his academic research. Our company “initiated a collaboration” with his group and that meant somebody ought to synthesize the stuff. Nobody volunteered but I was new and couldn’t protest - and my boss gushed that I could deliver the compound in six weeks. It took me over six moths of a full-time work – I was repeatedly offered reassignment to another project but I refused since I did not like to admit a failure.
The SynLet. paper describing the synthesis that I was supposed to reproduce had no experimental procedures and the used starting material was not commercial or described in the literature. The only procedure for the starting material was in the PhD thesis of a person who did also the work on this published macrocycle. The procedures in the thesis were rather vague, they never worked in my hands or in the hands of another chemist. I soon noticed that a hydrolytically-unstable silylated amine intermediate in the first step of a multistep sequence was formed in less than 50% NMR yield under conditions described in the thesis and the entire material was then immediately decomposed if the workup specified in the thesis was used. So I obtained this sensitive intermediate by alternate procedure, purified it by Kugelrohr distillation and took it to the next step – which also failed. Eventually I realized that the irreproducible thesis procedures were meant to be a “simplified” version of a known synthetic sequence. This original unmodified sequence worked exactly just as described so I changed the protecting groups on the material in the end and then I could proceed with making the actual macrocycle.
I had a number of unpleasant surprises following the scheme and finally the key condensation reaction did not work at all. The reagents they “used” in the published scheme were different from the lit reference that they gave as a precedent for this step – and neither of these methods gave any trace of the desired product, on many attempts. I was getting desperate. Eventually I found a synthetic chemist whom I could talk to, from the group that published this irreproducible stuff. As I was describing my troubles he was laughing. It turns out that what they published was apparently an unmodified synthetic scheme from some grant proposal - with made-up yields. (They never retracted anything even after I alerted the PI about the problem. And they kept re-publishing this mess, once in PNAS then Synlett and later I think in Chem. Rev.)
My inside friend told me what kind of method they would use in such a case – different from the one published - so I was able to complete the molecule, convince my superiors that the problem was with the published baloney rather than with my freshly – minted MS degree from Eastern Europe. Looking back I think the PI in question knew from the beginning that his published work was irreproducible. But he could not afford to retract because he had to keep hyping the “technology platform” of his troubled small company in order to make the investors to continue pouring money in. (He also tried to plagiarize our methodology – I carelessly talked about my unpublished work and he published our results as his own, Cordova-style.) His company folded in the end.