IIT Home Page CNR Home Page

Ranking of candidate microRNA-target pairs sensitive to disease and cell-line conditions

Motivation
microRNAs (miRNAs) are post-transcriptional regulators that play important roles in cellular development, differentiation and diseases. They are often deregulated in many pathologies (e.g. cancer and cardiovascular diseases). One essential step to understand the regulatory effect of miRNAs is the reliable prediction of their target mRNAs. However, prediction tools by  themselves are not enough, in that they are solely based on sequence information, and they do not take into account the specific cell line features and the gene expression. To better understand a pathology we need to collect the list of genes related to it, the miRNA-mRNA relationships, the specific cell line data and the gene expression. This work presents a general method applied to cancer, to accomplish this aim.
Methods
Our first step is retrieving the list of genes related to a disease by using the Entrez Programming Utilities, developed by the NCBI for searching publications from the MEDLINE. To do this we obtained the list of PubMed IDs containing gene information, then we filtered it by using the gene-to-publication link table provided by Entrez-Gene. We then applied the Hypergeometric Test (HT) on the number of publications, as done in Jourquin et al., to rank the retrieved genes and we applied the –log(HT) to obtain the gene score: the higher the score the more correlated the gene is to the pathology. Then, we compared the gene list with predicted miRNA targets and we selected those belonging to both the lists. Lastly, we incorporated the mRNA and miRNA expression from the Cancer Cell Line Encyclopedia (CCLE) and PhenomiR respectively. CCLE provides a detailed genetic characterization of a large panel of human cancer cell lines. For each gene in each cell line CCLE provides a single normalized expression value. To determine the regulation level of the gene we computed the median among the specific cancer cell line. The gene is declared to be up-regulated if its value is above the median value, and it is down-regulated otherwise. PhenomiR provides the miRNA expression on specific disease and cell line. Since miRNA generally repress their target mRNAs, a straightforward way to validate miRNA targeting mRNAs is detecting whether their expressions are inversely correlated. Based on this hypothesis, we retained only anti-correlated predicted miRNA-target interactions.
Results
We tested our approach on Prostate Cancer (see Figure 1.A). Among about 113000 papers related to this disease, about 7800 contained gene information and 3700 gene IDs (in Human). From this gene list we selected 242 genes with more than 9 supporting publications and a gene score greater than 2, as done by Jourquin et al. By using CCLE and PhenomiR data we selected a subset of miRNA-target interactions showing very high probability of binding based on target prediction tools. We have molecular and cellular data showing that miR-26a and miR-28 are both down-regulated in two prostate cancer tumor cell lines (DU-145 and PC-3) and that their over-expression inhibited cell proliferation. On the basis of these results we focused on these miRNAs and their targets. For DU-145 we found 11 genes while for PC-3 we found 8 genes, as shown in Figure 1.B. In particular we discovered some validated miRNA-target interactions for specific cell line such as miR-26a-PTEN for DU-145 (Tian et al.), miR-26a-EZH2 for both cell lines (Koh et al.). Notably, no association between PTEN and miR-26a was found for PC-3 (it can be explained because generally PTEN is mutated in PC-3 cell line). Furthermore, the importance of considering the specific cell line is shown by the different subset of targets obtained for the two miRNAs. It is also to notice that while no relationship can be found in literature between miR-28 and Prostate Cancer genes,  this approach was able to predict novel miRNA targets important for the disease. For this reason we are experimentally verifying the importance of miR-28 in Prostate Cancer.
References
Jourquin J  et al. GLAD4U: deriving and prioritizing gene lists from PubMed literature. BMC Genomics. 2012;13 Suppl 8:S20.
Tian L et al. Four microRNAs promote prostate cell proliferation with regulation of PTEN and its downstream signals in vitro. PLoS One. 2013 Sep 30;8(9):e75885.
Koh CM et al. Myc enforces overexpression of EZH2 in early prostatic neoplasia via transcriptional and post-transcriptional mechanisms. Oncotarget. 2011 Sep;2(9):66


Bioinformatics Italian Society (BITS) Annual Meeting 2014, Rome, 2014

External authors: Milena Rizzo (Institute of Clinical Physiology, CNR, Pisa), Gabriele Berti (Institute of Clinical Physiology, CNR, Pisa), Monica Evangelista (Institute of Clinical Physiology, CNR, Pisa), Giuseppe Rainaldi (Institute of Clinical Physiology, CNR, Pisa)
IIT authors:

Francesco Russo

Foto di Francesco Russo

Miriam Baglioni

Foto di Miriam Baglioni

Type: Abstract-poster on national conferences
Field of reference: Computer Science & Engineering