‘It’s all in the technique,” researchers are finding at Baylor College of Medicine in Houston, Texas. Dr. Matthew L. Anderson, assistant professor of obstetrics and gynecology at Baylor, led a study that used either Agilent microarrays or miRNA Next Generation Sequencing (miRNA-Seq) and data from the Cancer Genome Atlas (TCGA) on the same miRNA specimens from ovarian cancer patients and found two very different results. The microarrays identified 61 miRNAs associated with survival in 469 of the specimens, but miRNA-Seq found only 12. The overlap of the two methods was a single miRNA sequence associated with survival.
“The choice of tools in genomic profiling can make a difference in the answers at which you arrive,” said Dr. Anderson. The researchers conducting the study were careful to correct for potential issues inherent to signal detection algorithms, but corrections for false discovery and miRNA abundance made no difference in interpreting the results.
These findings might seem alarming. After all, “if you have a reliable tool for profiling, the picture should look the same or at least similar” when using the same samples, stated Dr. Anderson. “For microRNAs in ovarian cancer, it doesn’t.” But this does not discredit gene sequencing shown to be successful in making predictions about cancer patients. Said Dr. Anderson, “This discrepancy appears to be something specific to microRNAs, as other data dealing with genes and other genetic material are more consistent.”
Dr. Zhandong Liu, another co-author and professor at Baylor and a member of the Jan and Dan Duncan Neurological Research Institute at Texas Children’s Hospital, hypothesized that the discrepancy occurs because miRNAs are short sequences, and tools (microarrays) that are designed to work on long sequences (DNA) are less accurate on short sequences. “The impact of this could be important,” said Liu. “MicroRNAs are believed to be key drivers for cancer. This could have an important impact on similar studies. For now, I would urge caution in interpreting the microRNA data.”
Funding for this work came from the Partnership for Baylor College of Medicine; the Collaborative Advances in Biomedical Computing Seed Funding Program at the Ken Kennedy Institute for Information Technology at Rice University supported by the John and Ann Doerr Fund for Computational Biomedicine and through the Center for Computational and Integrative Biomedical Research Seed Funding Program at Baylor College of Medicine; the National Science Foundation (Grant DMS-1209017 and DMS-1263932) and the Houston Bioinformatics Endowment.