A systems biology approach to non-coding RNAs: the networks of cancer

A non-coding RNA is a functional RNA molecule that is not translated into a protein. This class of molecules is involved in many cellular processes and includes highly abundant and functionally important RNAs such as transfer RNA (tRNA), ribosomal RNA (rRNA), as well as small interfering RNAs (siRNAs), microRNAs (miRNAs), transcribed ultraconserved regions (T-UCRs) and others. First of all, we investigate the specificity for normal tissues of two selected non-coding RNAs: Transcribed UltraConserved Region and microRNAs. Second, we want to find whether these non-coding RNAs can be candidates as features for the selection of specific cancers, using statistical algorithms and bioinformatics tools. Third, we generate miRNA gene networks in normal and different cancer and leukemia. The term “ultraconserved” refer to genomic regions longer than 200 base pairs that are absolutely conserved (100% homology with no insertions or deletions) in human, mouse, and rat genomes. There are 481 T-UCRs. The reason for this extreme conservation remains a mystery; T-UCRs may play a functional role in the ontogeny and phylogeny of mammals and other vertebrates. Genome-wide profiling revealed that UCRs are frequently located on overlapping exons in genes involved in RNA processing and can be found in introns or at fragile sites and in cancer-associated genomic regions. We investigate the expression of T-UCRs in 374 normal samples from 46 different tissues, grouped by 16 systems. Moreover, we analyzed the specificity of T-UCRs in cancers. Tissue specific T-UCRs can differentiate cell types. We then examine the expression of T-UCRs in human embryonic stem cells, induced pluripotent stem cells, and a series of differentiated cell types (trophoblast, embryoid bodies at 7 and 14 days of differentiation, definitive endoderm, and spontaneous differentiated monolayers). One T-UCR in particular, uc.283 plus, is highly specific for embryonic and induced pluripotent stem cells, as confirmed by real time PCR (RT-PCR). MiRNAs are global regulators of protein output. Each miRNA has been studied for its single contribution to differential expression or to a compact predictive signature. Thus, we propose a study of miRNAs in cancer by applying a systems biology approach. We study miRNA profiles in 4419 human samples (3312 neoplastic, 1107 non-malignant), corresponding to 50 normal tissues (grouped by 17 systems) and 51 cancer types. We calculate tissue specificity and cancer type specificity, a small set of miRNAs were tissue-specific while many others were broadly expressed. Then we find whether non-coding RNAs can be candidates as features for the selection of specific cancers, using statistical algorithms and bioinformatics tools, as decision trees. Afterwards, we build miRNA gene networks by using our very large expression miRNA database. The complexity of our expression database enables us to perform a detailed analysis of coordinated miRNA activities. We also build specialized miRNA networks for different solid tumors and leukemias. Combining differential expression, genetic networks, DNA copy number alterations and other systems biology approaches we confirm or discovered miRNAs with comprehensive roles in cancer. We find that normal tissues are represented by single complete miRNA networks. Cancers instead show separate and unlinked miRNA sub-networks. miRNAs independent from the general transcriptional program were often known as cancer-related. We validate our results by in silico, in vitro and in vivo analysis. We demonstrate that the target genes of these uncoordinated miRNA involve in specific cancer-related pathways.