Case 2 The CD47-signal regulatory protein alpha (SIRPα) interaction is a therapeutic target for human solid tumors \cite{Willingham_2012}.
CD47 is a ubiquitously expressed membrane protein with many functions, one of which is to signal “do not eat me” to macrophages and dendritic cells by binding SIRPα, a protein present on their surface \cite{Matlung_2017}. Willingham et al. have reported that antibodies to CD47 that disrupt the CD47-SIRPα interaction inhibit the growth and metastasis of solid human tumors explanted into mice, and thus claimed that this interaction is a therapeutic target for human solid tumors.
The replication report \cite{Horrigan_2017} refuted the claim by finding no statistically significant effect of a CD47 antibody on tumor growth in an animal model, with the stated caveats that some of the tumors regressed spontaneously in the control group, that only one of the CD47 antibodies used in the original study was used in the replication study, and that only one out of several models used in the original study was used to test its claim.
By reviewing 224 reports citing the study by Willingham et al., we have identified 16 independent studies (Data) that confirmed the claim. These studies used several CD47 antibodies, SIRPα mimetics, and other approaches. Two citing studies, including the replication report, refuted the claim \cite{Horrigan_2017}; R-factor of the claim is 16/(16+2) = 0.8818 (Fig. \ref{900585}, middle), which is discordant with the conclusion of the replication study but concordant with the current standing of the claim in the field, the debate about the complexity of the underlying mechanisms notwithstanding \cite{Matlung_2017}.
Case 3 Coadministration of a tumor-penetrating peptide enhances the efficacy of cancer drugs \cite{Sugahara_2010}.
The authors of this report previously found that conjugating the peptide iRGD to various cancer drugs makes these drugs more potent by helping them penetrate into target tissues \cite{Sugahara_2009}. In this study, Sugahara et al. claim that the same effect can be achieved by injecting free iRGD together with a cancer drug, thus bypassing the need to link the peptide chemically.
The replication study found no evidence supporting the claim \cite{Mantis_2017}.
By reviewing 405 citing publications, we have identified nine studies that supported the claim and five that refuted it (Data). Two of the studies \cite{Akashi_2014}; both categories, as some of the reported experiments supported the claim while others did not, apparently because the effect depended on the experimental system used.
We then excluded two of supporting reports because although they tested different drugs, the graph of tumor growth presented in the first report \cite{Zhang_2015}  was superimposable, except of the labeling, to the graph in the second \cite{Zhang_2016}. In two other supporting studies the graphs representing the survival of treated mice implied that the animals died surprisingly regularly, one mouse per each observation point in each of the thirteen cohorts (six groups of six mice each \cite{Gu_2013}, and seven groups of 10 mice \cite{Wang_2014}. Because we considered such orderly demise improbable and an expert in animal experiments concurred with our opinion, we excluded these studies from analysis.
The remaining five confirming and five refuting studies resulted in the R-factor = 5/(5+5) = 0.5010 (Fig. \ref{900585}, right).