I have not viewed this video yet, but have attended a similar talk and read a similar research paper by Keith Baggerly. His general message is that large biological and genetic experiments are sometimes designed so poorly as to invalidate the results. You can often discover these design flaws through a careful examination of the data sets themselves and their metadata. This process of uncovering design flaws is sometimes called “Forensic Statistics.”
Keith Baggerly (2010) The Importance of Reproducible Research in High-Throughput Biology. Case Studies in Forensic Bioinformatics. Available at http://videolectures.net/cancerbioinformatics2010_baggerly_irrh/.