OT: {{Just to be clear, I'm not suggesting that the 50% value is fraud. I'm speaking only of reproducibility; that can be due to fraud, ignorance, or the fuzziness that clark refers to.}}
poorgradstudent, I understand what you were inferring a bit more. I am sorry if I misinterpreted what you were saying. I think both you and Clark are missing perhaps the biggest issues, inherent genetic variability in model organisms and certainly in cancer cell lines from lab to lab. The pressure that people will be under differs depending on how hot and competitive a field is, and that must be factored in. The tools available to confirm findings, or to support a conclusion aren't the same for every organism either. Put them all together and one has to come to the conclusion that reproducibility will be far easier in some organism, and in some areas of study than in others.
From the limited information jq1234 provided in his links, it is difficult to understand why a failure to reproduce results is arsing in some companies. If it is work based on cancer cell lines, then the companies needs to check cell lines obtained from the labs which published the studies, as well as test other lines that are supposed to be the same. If some work and others don't, they can then sequence the various lines, and use comparative genomic hybridization to see if any differences co-relate with the reproducibility or failure thereof. They have the chance to learn how genetic variation might alter results, which has real world importance. If they throw up their hands and whine rather investigating, then they are pretty foolish, short-sighted and maybe lazy in my view.
One point about manuscripts is that I view the authors conclusions and models as the least important parts of papers. I focus on the data and what is missing from the data. Someone in the lab recently presented a manuscript for journal club, and while there were some interesting experiments and ideas, we all thought the conclusions weren't supported very well by the data. They even misrepresented something I had published 10 years ago by saying my paper showed Sumoylation of a protein peaks in S phase when I found it did so around anaphase. Their own data showed (anaphase peak) the same but they were fixated on S phase and said it peaked in S phase. Pretty bad analysis and bad reviewing. Still, there we some value to he paper, even though they used a sledge-hammer approach. They also put something in the supplemental data but they clearly didn't understand the significance. I thought it was interesting and was consistent with some phenomenon we have been getting hints about.
News 
Market Data 
Discover 
