Skip to main content

Archived Comments for: Probabilistic prediction and ranking of human protein-protein interactions

Back to article

  1. 90% false positives?? Doubtful.

    Frederick Roth, Harvard Medical School

    22 November 2007

    The abstract states:

    "...it has recently been estimated that the overall average false positive rate of available computational and high-throughput experimental interaction datasets is as high as 90%."

    This doesn't seem correct, especially given that Rual et al (Nature 2005; one of the two available high-throughput Y2H human interaction datasets) showed false positive rates closer to 20%.

    Could this statement be a typo? Apparently not, since it is repeated in the main text.

    Where did this number come from? If you follow the citation to an Opinion piece by Hart et al (Genome Biology 2006), you find:

    "With false-positive rates exceeding 50%, and false-negative rates (the proportion of true interactions missed) for two-hybrid assays in particular approaching 90%..."

    Oops. Looks like the citation said 50% false positive rate, not 90%. Whether the 50% estimate is accurate is another question for another time.

    Competing interests

    I co-authored Rual et al, Nature 2005

  2. 90% average false positive rate estimate

    Michelle Scott, University of Dundee

    22 January 2008

    In response to the comment posted on November 22 2007:

    We wrote

    "...it has recently been estimated that the overall average false positive rate of available computational and high-throughput experimental interaction datasets is as high as 90%."

    This is not a typo. As cited in our paper, this estimate of 90% false positive rate was reported in Hart GT, Ramani AK, Marcotte EM (2006) Genome Biol 7(11):120. Table 3 of Hart et al, (bottom right-hand value) clearly states that the average false positive rate over all datasets examined is estimated to be 90%.

    Competing interests

    None declared

Advertisement