Selective discussion and transparency in microarray research findings for cancer outcomes

We examined the interpretation of research findings and public availability of transparent information on data and processing for 46 articles of microarray studies that had addressed major cancer outcomes. Unsupervised and supervised methods selected molecular signatures with a median of 675 and 50 genes, respectively, but only a median of eight genes or groups thereof were further discussed. Across 479 genes or groups thereof discussed in all 46 studies, 65% reflected specific comments (reflecting external relevant data from other studies or other lines of reasoning relevant to the gene of interest), and 59% of the comments were referenced. Among specific comments, supportive ones outnumbered comments against the research findings by nine to one (270 versus 29). Discussion was similarly selective in early studies and in studies published in 2006. Even in 2006 only 10 of 15 studies had publicly deposited data. Only three studies had scanned images, raw and processed data available. Processing details varied. Public transparency and unbiased interpretation of findings can be improved in microarray research.

1. 1
   Why Most Published Research Findings Are False

   PLoS Medicine200510,409 citationsDataRank 1.4
2. 2
   Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling

   Nature20009,923 citationsDataRank 1.4
3. 3
   Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): Explanation and Elaboration

   Annals of Internal Medicine20155,093 citationsDataRank 1.3
4. 4
   Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses

   Proceedings of the National Academy of Sciences20012,583 citationsDataRank 1.2
5. 5
   Why Most Discovered True Associations Are Inflated

   Epidemiology20081,529 citationsDataRank 1.1