NEW YORK (GenomeWeb) – A new study suggests that researchers can get a refined look at the microRNAs contributing to cancer by incorporating information on related messenger RNA (mRNA) and protein expression data.
"Compared with the miRNA rankings and modules constructed using only mRNA expression data, the rankings and modules constructed using mRNA and protein expression data were shown to have better performance," senior author Hyunju Lee, an electrical engineering and computer science researcher at Gwangju Institute of Science and Technology, and her co-authors wrote.
Lee and her colleagues from Korea considered differences in the ways that miRNA influences were ranked in almost 200 tumor samples from the Cancer Genome Atlas project using mRNA-miRNA and/or protein-miRNA interaction profiles. Their findings, published online yesterday in PLOS One, indicated that the combination of gene and protein expression profiles can help in constructing miRNA interaction modules — an approach that the investigators subsequently applied to uncover dozens of miRNA modules that appear to be relevant to the brain cancer glioblastoma.
The team turned to two different computational strategies to bring together miRNA, mRNA, and protein profiles for 192 tumor samples and 10 unmatched samples from TCGA, producing miRNA interaction modules through tweaks and expansions to available algorithms.
Although the researchers saw broad correlation between gene and protein expression levels, some differences did arise, prompting them to continue taking both mRNA levels and protein patterns into account when ranking miRNAs and producing miRNA interaction modules.
The team then focused on the glioblastoma tumors, comparing them with the normal samples. From the nearly 5,900 genes that were differentially expressed in glioblastoma, it picked a subset of genes with available protein expression data to do miRNA interaction analyses.
In addition to establishing rankings of the miRNAs that appear to be important in glioblastoma, the researchers used differential gene expression data, together with protein profiles and miRNA data, to narrow in on 52 glioblastoma-related miRNA modules. Each of the modules was made up of some 69 mRNA transcripts, 10 proteins, and half a dozen miRNAs, on average, they reported.
Through a series of analyses with existing functional data, the team validated some of these modules. It also did follow-up transfection, miRNA inhibitor, and expression experiments to take a closer look at miR-504, one of the most highly ranked miRNAs in glioblastoma.
In particular, the study's authors saw signs that miR-504 may act to suppress glioblastoma development. On the other hand, the available patient data hinted that enhanced miR-504 expression appeared to coincide with better outcomes in individuals with glioblastoma.
"In [the] future, our approach can be expanded to other cancer types in order to identify cancer-related miRNAs or genes," they wrote. "These analyses can be used for the determination of candidate markers for cancer therapy, following additional validation by other methods."