Peer review filters out the stuff that is obvious crap, stuff that doesn't even fit the form of a proper scientific article. The purpose is not to say that articles are true, but rather to get rid of articles that are obviously wrong. If the scientists are lying about their data, it's hard for peer review to catch that. That's why reproducibility is important. If it's a result you care about, you can reproduce it.

For the Cell submission, there were concerns about methodology and the lack of supporting evidence for the extraordinary claims, says [stem cell scientist Hans] Schöler, who reviewed the paper and, as is standard practice at Cell, saw the comments of other reviewers for the journal. At Science, according to the 8 May RIKEN investigative committee’s report, one reviewer spotted the problem with lanes being improperly spliced into gel images. “This figure has been reconstructed,” the RIKEN report quotes from the feedback provided by a Science reviewer. The committee writes that the “lane 3” mentioned by the Science reviewer is probably the lane 3 shown in Figure 1i in the Nature article. The investigative committee report says [co-author Haruko] Obokata told the committee that she did not carefully consider the comments of the Science reviewer.

All three Nature reviewers concluded that the data presented in the submitted manuscripts were not enough to support such radical claims. “I would recommend the authors to be extremely cautious in their claims . The authors should look into the actual effect that the treatment elicits in the genome and they should assess genomic instability,” one writes. “There are several issues that I consider should be clarified beyond doubt because of the potential revolutionary nature of the observations,” writes another.