About a year ago, I came across a really interesting Science paper entitled "Rapid and inefficient isolation of single cells from soil". In it, the authors -- from a well-known lab at UC Davis -- described how they used low-percentage agarose gels to extract thousands of individual cells from a soil sample, split them into a microtitre plate, apply single-cell amplification and massively parallel barcoding, and then sequence them in aggregate. It didn't need any specialized hardware, and could be done with off the shelf hardware. What was even cooler is that the bioinformatics was very straightforward and used completely open tools. Awesome!
My postdoc set out to try it on our own sample, and immediately ran into problems. The percentage of the agarose gel wasn't specified, and neither was the brand; we couldn't get similar results on the separation. The brand of the microtitre plate also wasn't specified in the paper, and we were worried that there was something different with our sample.
We contacted the authors and spent several frustrating e-mail exchanges trying to get all the information. They were forthcoming about many things, but wouldn't tell us their procedure for mixing the agar, which was key -- it turns out they were thinking of commercializing the process, and their lawyers had advised them not to let it out. Separately, there were several protocols in the lab for doing the barcoding, and it emerged that they weren't quite sure which one had been used for the data published in the paper.
Undeterred, I shipped my postdoc out to the Davis lab to collaborate with them. She came back having been unable to replicate their experiments, but with a much greater appreciation for the complexity of the process.
We finally got things working in our lab, and went to publish our own protocol and results. At that point, a new wrinkle emerged: the details of the original protocol were copyrighted and had only been licensed for the purpose of use, not for adaptation and reuse. So we couldn't publish a derivative protocol without first buying a license from the UC system! That took another 6 months to arrange with lawyers.
In the end, we submitted the paper for review, and it just got published. We've been asked multiple times for our protocol, but unfortunately we can only point them towards the authors of the first paper, because we don't have a license to distribute the original protocol. But at least we got our publication!
p.s. This is a fictional story, illustrating how most papers and labs deal with source code. Replace 'protocol' with 'source code', and reread.
Posted by Ian on 2012-06-08 at 19:56.
Posted by Jonathan Eisen on 2012-06-08 at 21:20.
Did you forget the part about how you had to ay 30$ to get the article when you were at home b/c Science likes to screw people?
Posted by Ed Rybicki on 2012-06-09 at 09:08.
To be honest, I think that ANYONE who publishes a non-replicable protocol ought to have their publication retracted - because the essence of publication is to share, not only results, but also the damn METHODS! Thinking about patenting something is also ridiculous, if the thing was supposedly published, as only in the US could this be used as an excuse - and not for very much longer, if US patent law gets in line with the rest in the world, where prior publication voids patentability. And Jonathan, I don't know about you, but our developing country University makes it possible for us to log in from anywhere, and access their subscriptions...B-)