Mapping of conserved RNA secondary structures predicts thousands of functional noncoding RNAs in the human genome

Abstract

In contrast to the fairly reliable and complete annotation of the protein coding genes in the human genome, comparable information is lacking for non-coding RNAs. We present a comparative screen of vertebrate genomes for structural non-coding RNAs, which evaluates sequence conservation, secondary structure conservation, and thermodynamic stability of putative RNA structures. We predict more than 30 000 structured RNA elements in the human genome, almost 1000 of which are conserved across all vertebrates. Roughly a third is found in introns of known genes, a sixth are potential regulatory elements in untranslated regions, about half are located far away of any known gene. Only a small fraction of these sequences has been described previously. EST data demonstrate, however, that the majority of them is at least transcribed. The widespread conservation of secondary structure points to a large number of functional ncRNAs in the human genome, which we estimate to be comparable to the number of protein-coding genes. The recent finishing of the human genome sequence emphasizes the "need for reliable experimental and computational methods for comprehensive identification of non-coding RNAs" 1 . A variety of experimental techniques have been used to uncover the human and mouse transcriptomes, in particular tiling arrays 2-4 , cDNA sequencing 5,6 , and unbiased mapping of transcription factor binding sites 7 . All these studies agree that a substantial fraction of the genome is transcribed and that a large fraction of the transcriptome consists of non-coding RNAs. It is unclear, however, which fraction are functional non-coding RNAs (ncRNAs), and which constitutes "transcriptional noise" 8 .Genome-wide computational surveys of ncRNAs, on the other hand, have been impossible until recently, because ncRNAs do not share common signals that could be detected at the sequence level. A large class of ncRNAs, however, has characteristic structures that are functional and hence are well conserved over evolutionary timescales: most of the "classical" ncRNAs, including rRNAs, tRNAs, snRNAs, snoRNAs, as well as the RNA components of RNAse P and the signal recognition particle, are of this type. The stabilizing selection acting on the secondary structure causes characteristic substitution patterns in the underlying sequences: Consistent and compensatory mutations replace one type of base-pair by another one in the paired regions (helices) of the molecule. In addition, loop regions are more variable than helices. These patterns can be ex-1 ploited in comparative computational approaches 9-12 to discriminate functional RNAs from other types of conserved sequence. Recently, high levels of sequence conservation of non-coding DNA regions have been reported 13, 14 . Here we screen the complete collection of conserved non-coding DNA sequences from mammalian genomes and provide a first annotation of the complement of structurally conserved RNAs in the human genome. ResultsSelection of conserved sequences and screening for structural RNAs We s...

Genome-wide discovery and verification of novel structured RNAs inPlasmodium falciparum

“…These randomized data suggested a false discovery rate of 35.6% (RNAz P-value of 0.5) and 23.3% (RNAz P-value of 0.9). This rate is higher than that reported for mammalian RNA genes (28.8% and 19.2%, respectively) (Washietl et al 2005a), as is the number of predictions per kilo base pair of shuffled sequence (mammals [P > 0.5]: 0.32 predictions/kbp [P > 0.9]: 0.08. Plasmodium [P > 0.5]:0.66 [P > 0.9]: 0.27).…”

“…To estimate the specificity of our predictions, we adopted the approach of Washietl et al (2005a) and shuffled the alignments (Washietl and Hofacker 2004). These randomized data suggested a false discovery rate of 35.6% (RNAz P-value of 0.5) and 23.3% (RNAz P-value of 0.9).…”

“…Recent studies indicate that a large number of these noncoding structured RNAs play critical roles in regulating gene expression at multiple levels in diverse organisms (Eddy 2001;Washietl et al 2005a;Mattick and Makunin 2006;Pedersen et al 2006;Backofen et al 2007;Prasanth and Spector 2007). In recent years, more than 800 ncRNA genes have been described in mammals, a subset of which are alternatively spliced and show tissue-specific expression or developmental regulation (Cheng et al 2005;Jongeneel et al 2005;Washietl et al 2005a;Pedersen et al 2006;Ravasi et al 2006), and a few have been implicated in disease phenotypes or developmental disorders (Szymanski and Barciszewski 2006). More recently, the ENCODE pilot project has identified a large number of non-protein-coding transcripts, with many overlapping protein-coding loci and a large number of mappings to genomic regions that were previously thought to be transcriptionally inactive (The ENCODE Project Consortium 2007).…”

Deep sequencing of the mouse lung transcriptome reveals distinct long non-coding RNAs expression associated with the high virulence of H5N1 avian influenza virus in mice

“…Through University of California Santa Cruz (UCSC) Genome Browser (http://genome-asia.ucsc.edu/index.html), the genomic location of NONMMUT011061 was predicted to be on chromosome 11qB5 and was assumed to be unable to encode genes (predicated through PhyloCSF tracks: https://data.broadinstitute.org/compbio1/PhyloCSFtracks/trackHub/hub.txt) (Figure 9(A)). In silico prediction of the secondary structure of lncRNA is another useful method to define putative functions of non-coding transcripts, based on the widely-held assumption that highly-folded structures affect functionality through binding interactions with proteins/nucleotides[48]. Using RNAfold minimum free energy estimations based on RNAfold webserver (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi), a highly-folded secondary structure with several hairpin loops of NONMMUT011061 was identified (Figure 9(B)).…”

Fast RNA Structure Alignment for Crossing Input Structures

“…A recent screen on ncRNAs has detected more than 30000 putative ncRNAs in human genome [3], most of them with unknown function. Since the structure of RNA is evolutionarily more conserved than its sequence, predicting the RNA's secondary structure is the most important step towards its functional analysis [4].…”