A population-based statistical approach identifies parameters characteristic of human microRNA-mRNA interactions
© Smalheiser and Torvik; licensee BioMed Central Ltd. 2004
Received: 12 April 2004
Accepted: 28 September 2004
Published: 28 September 2004
MicroRNAs are ~17–24 nt. noncoding RNAs found in all eukaryotes that degrade messenger RNAs via RNA interference (if they bind in a perfect or near-perfect complementarity to the target mRNA), or arrest translation (if the binding is imperfect). Several microRNA targets have been identified in lower organisms, but only one mammalian microRNA target has yet been validated experimentally.
We carried out a population-wide statistical analysis of how human microRNAs interact complementarily with human mRNAs, looking for characteristics that differ significantly as compared with scrambled control sequences. These characteristics were used to identify a set of 71 outlier mRNAs unlikely to have been hit by chance.
Unlike the case in C. elegans and Drosophila, many human microRNAs exhibited long exact matches (10 or more bases in a row), up to and including perfect target complementarity. Human microRNAs hit outlier mRNAs within the protein coding region about 2/3 of the time. And, the stretches of perfect complementarity within microRNA hits onto outlier mRNAs were not biased near the 5'-end of the microRNA. In several cases, an individual microRNA hit multiple mRNAs that belonged to the same functional class.
The analysis supports the notion that sequence complementarity is the basis by which microRNAs recognize their biological targets, but raises the possibility that human microRNA-mRNA target interactions follow different rules than have been previously characterized in Drosophila and C. elegans.
MicroRNAs (miRNAs) are small, ~18–24 nt. noncoding RNAs that are found in all eukaryotes and are cleaved from larger ~70 nt. precursors via the action of Dicer enzyme [reviews: ref. [1, 2]]. MicroRNAs are thought to degrade messenger RNAs via eliciting mRNA degradation (if they bind in a perfect or near-perfect complementarity to the target mRNA), or to arrest translation of the mRNAs (if the binding complementarity is imperfect). Although a number of microRNA targets have been identified in plants, C. elegans and Drosophila [1, 2], only one mammalian microRNA target has yet been validated [3, 4].
Five different papers have recently appeared that used computational approaches to predict microRNA targets in Drosophila [5–7], and mammals [8, 9]. These studies only considered hits occurring within 3'-UTR regions that were conserved across related species, and favored or required a short region of perfect complementarity towards the 5'-end of microRNAs. However, there is reason to suspect that the rules governing microRNA-target interactions may not be universal. For example, in plants, most of the known microRNAs bind in a perfect or near-perfect manner to mRNA targets located within the protein coding region (cds) [10, 11]. In contrast, in C. elegans  and Drosophila , known microRNAs lack long stretches (>10) of complementarity with their targets and generally interact within the 3'-untranslated region (3'-UTR). Furthermore, whereas the 5'-ends of many Drosophila microRNAs recognize 5–6 nt. common motifs within the target, these motifs are not a general feature of mammalian microRNAs . Thus, it is conceivable that human microRNA targets do not follow the same constraints as observed in C. elegans and Drosophila.
In the present paper, we have performed an unbiased statistical analysis of the manner in which human microRNAs interact complementarily with human mRNAs present in the NCBI human RefSeq database, looking for characteristics that differ significantly as compared with scrambled versions of the same microRNA sequences. The results demonstrate several novel features of human microRNA-mRNA interactions that differ from C. elegans and Drosophila, and identify a short-list of promising candidate microRNA-mRNA target pairs that are unlikely to have arisen by chance.
Population-wide statistical analyses were first carried out by examining the types of complementary interactions that occur between the set of microRNAs listed in Lagos-Quintana et al , and the set of human RefSeq mRNAs downloaded in August 2003. To obtain a fuller list of outlier mRNAs, analyses were repeated using all human microRNAs listed on the Sanger microRNA repository  and the set of human RefSeq mRNAs listed as of December 2003 .
To define the types of interactions that can occur by chance, ten independent sets ("replications") of scrambled microRNA counterpart sequences were generated and examined for complementarity with the mRNA population. Our underlying assumption is that scrambled sequences will hit mRNA at random and define the "noise" level in any given situation, whereas microRNA sequences will hit the same number of "noise" interactions plus any true targets. Unless otherwise noted, the scrambled sequences were random permutations of the microRNA sequences, keeping constant the overall nucleotide composition. Because microRNAs have a distinctive nonrandom di-nucleotide composition, we also confirmed that key findings were obtained when using scrambled sequences that had similar di-nucleotide composition to the microRNAs.
1. Human microRNAs tend to have longer exact hits upon mRNAs than do their scrambled counterparts
2. Constructing an outlier set of microRNAs based on cut-offs of exact hit length, gapped BLAST score and presence of multiple hits
As shown in figure 1, there are a total of 101, 279 microRNA hits upon RefSeq sequences hitting exactly ≥10 bases in a row, compared to 75,031 hits produced by scrambled microRNA sequences. The difference (26, 248 hits, distributed among 8258 mRNA sequences) is highly significant (p = 3 × 10-9) and suggests that about 1/4 of the total hits in this "10+ set" occur upon "true" biological mRNA targets. Our approach is to identify further the mRNAs that represent statistical outliers (i.e. that are unlikely to be hit by chance) within this larger "10+ set" by comparing properties of hits made by the set of microRNA sequences vs. the set of scrambled sequences. At any given parametric value, the number of hits observed in the microRNA set, minus the number of hits in the scrambled set, provides an estimate of the number of true microRNA targets that satisfy that parametric value. We examined three different hit properties – a) exact hit length, b) gapped BLAST score and c) presence of multiple hits – both alone and combined with each other. Starting from the "10+ set" estimated to contain only 26% true targets (see above), we added additional criteria to compile a list of candidates estimated to contain over 80% true targets.
a) Exact hit length
The most important single parameter for discriminating hits produced by microRNAs vs. scrambled sequences appears to be exact hit length. At a cut-off of 17 exact hit length, there were 14 mRNAs hit by the microRNA set that satisfied this criterion, vs. an average of 1.9 mRNAs hit by each of the scrambled sequence sets (fig. 1). Stated another way, this criterion gives a discrimination ratio of 7.4 to 1. A similar discrimination ratio was observed when comparing scrambled sequences maintaining the same di-nucleotide composition as the microRNAs.
b) Gapped-BLAST score
c) Multiple hits
d) Combining parameters
When combined, all three parameters (exact hit length, gapped-BLAST scores and multiple hits) gave better discrimination power than using any single feature, supporting the idea that they are relevant to identifying biologically relevant mRNA targets. We examined three different combinations of parameter cut-off values: 1) One combination consisted of targets with multiple hits from distinct microRNAs less than 25 bases apart, with at least one exact hit ≥13 bases and with at least one gapped BLAST score ≥185 (not counting G:U). For the next two lists, we scored only exact hits ≥10 bases long and that occurred ≤50 times within the entire mRNA population; this minimized "noise" arising from common or low-complexity target sequences, albeit at the cost of removing some target sequences that are shared within protein families. 2) Criteria required two or more hits from distinct microRNAs ≤100 bases apart, at least one exact hit ≥14 bases and one gapped-BLAST score of ≥190 (not counting G:U). 3) This required hits ≤500 bases apart, at least one exact hit ≥14 bases, and at least one gapped-BLAST score > 89% of the best-possible score including G:U matches (this takes into account the fact that longer microRNAs have greater possible absolute scores than shorter microRNAs).
The Outlier mRNA Set (Candidate Target List)
Transcription factors and other nucleic-acid binding proteins (15)
homeo box B8 (HOXB8)
E2F transcription factor 6 (E2F6)
transcription factor 20 (AR1) (TCF20)
DEAD (Asp-Glu-Ala-Asp) box polypeptide 51 (DDX51)
similar to ATP-dependent RNA helicase DDX24 (DEAD-box protein 24) (LOC221311)
myeloid/lymphoid or mixed-lineage leukemia (trithorax homolog); translocated to, 1 (MLLT1)
high mobility group AT-hook 2 (HMGA2)
polymerase (DNA directed), theta (POLQ)
strand-exchange protein 1 (SEP1)
hypothetical protein FLJ12994 – RFX DNA-binding domain
similar to LINE-1 reverse transcriptase homolog (LOC285907)
similar to hypothetical protein (L1H 3 region) – related to reverse transcriptase
similar to putative p150 (LOC282945) – related to reverse transcriptase
similar to reverse transcriptase related protein (LOC222252)
similar to RTl1 (LOC376283) – related to reverse transcriptase
Kinases, receptors and other signaling proteins (13)
fyn-related kinase (FRK)
WNK kinase, lysine deficient 3 (PRKWNK3)
protein phosphatase 2, regulatory subunit B (B56), epsilon isoform (PPP2R5E)
EphA5 receptor (EPHA5)
killer cell lectin-like receptor subfamily A, member 1 (KLRA1)
polycystin and REJ (sperm receptor for egg jelly homolog, sea urchin)-like (PKDREJ)
integrin, alpha X (antigen CD11C (p150), alpha polypeptide) (ITGAX)
inositol 1,4,5-triphosphate receptor, type 1 (ITPR1)
hypothetical protein FLJ32810 – RhoGAP domain, SH3 domain
hypothetical protein FLJ00058 – G protein gamma 7
Rho GTPase-activating protein (RICS)
hypothetical protein FLJ30899 – probable ras GAP
similar to ADP-ribosylation factor-like membrane-associated protein (LOC132946) – ARF-like small GTPase domain, Sar1p-like member of the Ras-family
Membrane and extracellular proteins (11)
Laminin, beta 4 (LAMB4)
laminin, gamma 2 (LAMC2)
fibronectin 1 (FN1)
collagen, type IV, alpha 5 (Alport syndrome) (COL4A5)
collagen, type XIX, alpha 1 (COL19A1)
similar to Voltage-dependent anion-selective channel protein 1 (VDAC-1)
ATPase, Na+/K+ transporting, alpha 2 (+) polypeptide (ATP1A2)
complement component 1, q subcomponent, beta polypeptide (C1QB)
hypothetical protein FLJ20506 – transmembrane protein
MAM domain containing glycosylphosphatidylinositol anchor 1 (MDGA1) – Ig, MAM domains
similar to TCAM-1 (LOC284171)
Cytoskeletal domain-containing proteins (7)
myosin heavy chain Myr 8 (MYR8)
ankyrin repeat domain 17 (ANKRD17)
KIAA1817 protein – intermediate filament, ATPase, PDZ, Band 4.1, FERM domains
chromosome 10 open reading frame 39 (C10orf39) – homologous to myosin, plectin
oxysterol binding protein 2 (OSBP2) – pleckstrin homology domain
KIAA1202 protein – PDZ, ATPase domains
hypothetical protein FLJ23529 – homolgous to dynein heavy chain
Miscellaneous or unknown function (26)
cell cycle progression 2 protein (CPR2)
olfactomedin 3 (OLFM3)
histidine rich calcium binding protein (HRC)
interferon-related developmental regulator 1 (IFRD1)
KIAA1301 protein – NEDD4-related E3 ubiquitin ligase NEDL2
KIAA1203 protein – ubiquitin C-terminal hydrolase
hydroxyprostaglandin dehydrogenase 15-(NAD) (HPGD)
UDP-GlcNAc:betaGal beta-1,3-N-acetylglucosaminyltransferase 1 (B3GNT1)
KIAA1854 protein – leucine rich repeat C-terminal domains
testis specific, 14 (TSGA14)
chromosome 4 open reading frame 1 (C4orf1)-membrane AND nuclear protein
hypothetical protein FLJ33069
hypothetical protein FLJ38464
hypothetical protein LOC285431
hypothetical protein LOC284107
similar to agCP1362 [Anopheles gambiae str. PEST] (LOC344751)
similar to hypothetical protein D11Ertd497e (LOC343360)
4. Characterizing the mRNA outlier set
The outlier mRNA set contained very similar types of targets as predicted by previous computational studies [5–8], including members of the same gene families. For example, Lewis et al.  described E2F1 as a candidate target whereas we found E2F6 (Table 1). Transcription factors (including homeobox genes) and nucleic acid-binding proteins are among the top predicted microRNA targets. As well, many other functional categories are represented including kinases, receptors and other signal transduction proteins, membrane and cytokeletal proteins, and effectors of differentiation (Table 1). However, surprisingly, we found that the human candidate microRNA target list also had some features that differed significantly from the known targets in C. elegans and Drosophila. For example, there was no preference for microRNA hits to be located within 3'-untranslated regions: 5% of hits were located in the 5'-UTR, 1% at the 5'-UTR/coding junction, 67% in the protein coding region, 1% at the coding/3'-UTR junction, and only 26% in the 3'-UTR. This distribution was not significantly different from hits produced by the scrambled sequences. As well, the best microRNA hits upon candidate mRNA targets did not have relatively better target complementarity near their 5'-end: Only 13% of hits had ≥ 7 exact hit length starting at position 1 or 2 relative to the 5'end of the microRNA (vs. 17.5% of hits produced by scrambled sequences).
MicroRNA 145 is particularly interesting because it hits 17 distinct targets on the candidate list, of which a disproportionate number (6) are in the signal transduction category and three of these are related to GTPase activation (Rho GTPase-activating protein (RICS), G protein gamma 7, and hypothetical protein FLJ32810 – containing RhoGAP and SH3 domains; Table 1). A recent study showing that miR-143 and miR-145 are both underexpressed in colorectal neoplasia  had previously proposed the first two of these candidates as potential targets. Interestingly, the third target found here is not only novel (XM_350859, RhoGAP-like) but is hit by both miR-143 and miR-145 in close proximity (see additional data file 2 in http://genomebiology.com/2004/5/2/P4), further suggesting that this is likely to be a true biological target for microRNA regulation.
By comparing how the population of microRNAs vs. their scrambled counterparts interact with the population of human RefSeq mRNA sequences, we estimate that the probability of detecting a true microRNA target increases a) as the length of exact complementarity of a "hit" between microRNA and target increases, b) as the overall complementarity of a "hit" increases (allowing for gaps, mismatches and G:U matches), and c) as two or more distinct microRNAs hit the same mRNA in closer proximity. Targets in the outlier mRNA set also received more hits per unit length and more multiple hits from distinct microRNAs than expected by chance. Finally, we found cases in which an individual microRNA hit multiple mRNAs that belonged to the same functional class. The analysis suggests that target complementarity is a major factor in identifying biologically relevant mRNA targets: As values of each parameter increase, the difference between the number of hits in the microRNA set vs. the scrambled set increases steadily, and combining all three parameters gives better discrimination power than using any single feature.
So far, these conclusions agree with five different papers that used computational approaches to predict microRNA targets in Drosophila [5–7], and mammals [8, 9], using different strategies, criteria and filters than employed here. However, three significant differences were observed between human mRNAs in the outlier set and Drosophila targets: 1) Human microRNAs hit mRNAs with exact hit lengths extending much longer than observed in Drosophila, up to and including perfect complementarity. 2) Human microRNAs hit candidate mRNA targets within the protein coding region about 2/3 of the time. (This resembles the manner in which plant microRNAs hit their mRNA targets [10, 11].) 3) The stretches of perfect complementarity within microRNA hits in the outlier mRNA set were not biased to occur near the 5'-end of the microRNA. This is not necessarily at odds with earlier analyses, since our outlier set includes only perfect stretches of 13 bases or more, and the 5' end may be more critical in those cases where only a short perfect stretch of complementarity exists.
One might object that our ability to detect certain trends seen in Drosophila and C. elegans was simply obscured by the fact that we searched the large sequence space represented by all human mRNA sequences – the larger the sequence space, the greater the chance that any given target criterion will be satisfied by scrambled sequences, hence the more difficult it can be to detect true targets above the noise level. We agree that this can be a problem using very large sequence databases, such as the human EST database or the entire human genome. As well, using cut-off levels of parameter distributions to define the candidate list probably excludes many true human mRNA targets. However, human RefSeq was demonstrably not too large for our analysis, since very strong trends were observed in a variety of other parameters (figs. 1,2,3,4).
Based upon sequence complementarity, at least 57 out of the 71 members of the outlier set are predicted to represent true microRNA targets (Table 1). Indeed, since this paper was first submitted for publication, one of the mRNAs on this list, HOXB8, has been experimentally confirmed . Note, however, that accessory factors in the RISC might also help to determine which potential mRNA targets will actually be sites of regulation in vivo. As well, microRNA and target must be expressed in the same times and places in adequate concentrations; secondary structure of the mRNA target region may be important [19, 20]; see also ; and RNA A-to-I editing [25, 26] might operate to prevent certain target sequences from binding microRNAs adequately.
In summary, the population-wide characteristics of microRNA-mRNA sequence complementarity indicate that microRNAs recognize a subset of human mRNA sequences better than expected by chance. This outlier set does obey a number of properties expected for true biological mRNA targets, but does not show a bias for target regions to be located within the 3'-UTR of the mRNA, and stretches of perfect complementarity are not biased towards the 5'-end of the microRNA. If the candidate list is representative of the full set of biologically significant targets, then the total number of mRNA targets in humans may be much greater than previously proposed .
Statistical analyses were first carried out using the set of mouse and human microRNAs listed in Lagos-Quintana et al , and then repeated to obtain individual candidate mRNA targets using all human microRNAs listed on the Sanger microRNA repository  as of December 2003. These sources were combined to create nonredundant microRNA sets (i.e. microRNAs that have 10 or more consecutive nucleotides in common were collected into groups and the longest member of the group was chosen as nonredundant). Almost all mouse microRNAs have exact human counterparts, but hits were annotated with mouse entries in cases of minor corrections and discrepancies between these two sources. One individual microRNA (mir-207) and several scrambled sequences were found to be low-complexity or complementary to abundant repeats (e.g., Alu) and were removed from consideration.
Analyses were first carried out using the set of human RefSeq mRNAs available in August 2003, and then supplemented with additional human RefSeq mRNAs listed as of December 2003. A) Sequences in RefSeq > 20,000 bases long were removed from consideration because they were hit by many, if not all microRNAs, and a few sequences > 15,000 bases long were removed from the final candidate list because they had a relatively high false-positive probability. B) When counting the number of hits over the population of mRNAs, two hits were counted as redundant if the entire region around the hit (plus or minus 25 nucleotides on each side) was identical. C) When counting distinct hits by microRNAs on the same target, two hits were counted as redundant if they shared the same exact hit. This minimized possible artifacts due to overlapping microRNAs, as well as removed cases in which microRNAs hit exactly-repeating sequences within the target. D) In tabulating hits onto mRNA targets, we did not count hits that contained low-complexity sequences as detected by the DUST algorithm encoded by a Perl script provided by Lincoln Stein . E) When assembling the candidate mRNA target list, we chose a single exemplary mRNA and removed other entries that were transcript variants or nearly identical by BLAST searching. In the course of this study, some of the target mRNAs were removed from RefSeq for routine genome annotation processing. If these were subsequently replaced with updated versions of these mRNAs in RefSeq that included the same hits, the latter version is listed here as well. For those entries removed but not replaced in RefSeq at the time of submission of the manuscript, other active entries currently in Genbank are listed if possible.
To decide whether the number of observed microRNA hits were significantly different from chance, 10 replications of scrambled sequences were used to estimate prediction intervals. The prediction interval allows one to say with 95% confidence that any single new replication of the scrambled set will be below the value of the microRNA set. Prediction intervals were chosen as more conservative and more appropriate than confidence intervals.
5'-untranslated region. CDS, protein coding region. 3'-UTR, 3'-untranslated region.
Supported by NIH grants DA15450 and LM07292. This Human Brain Project/Neuroinformatics research is funded jointly by the National Library of Medicine and the National Institute of Mental Health.
- Lai EC: microRNAs: runts of the genome assert themselves. Curr Biol 2003, 13: R925–936. 10.1016/j.cub.2003.11.017View ArticlePubMedGoogle Scholar
- Carrington JC, Ambros V: Role of microRNAs in plant and animal development. Science 2003, 301: 336–338. 10.1126/science.1085242View ArticlePubMedGoogle Scholar
- Nelson PT, Hatzigeorgiou AG, Mourelatos Z: miRNP:mRNA association in polyribosomes in a human neuronal cell line. RNA 2004, 10: 387–394. 10.1261/rna.5181104PubMed CentralView ArticlePubMedGoogle Scholar
- Sempere LF, Freemantle S, Pitha-Rowe I, Moss E, Dmitrovsky E, Ambros V: Expression profiling of mammalian microRNAs uncovers a subset of brain-expressed microRNAs with possible roles in murine and human neuronal differentiation. Genome Biol 2004, 5: R13. 10.1186/gb-2004-5-3-r13PubMed CentralView ArticlePubMedGoogle Scholar
- Enright AJ, John B, Gaul U, Tuschl T, Sander C, Marks DS: MicroRNA targets in Drosophila. Genome Biol 2003, 5: R1. 10.1186/gb-2003-5-1-r1PubMed CentralView ArticlePubMedGoogle Scholar
- Stark A, Brennecke J, Russell RB, Cohen SM: Identification of Drosophila MicroRNA Targets. PLOS Biol 2003, 1: 397–409. 10.1371/journal.pbio.0000060View ArticleGoogle Scholar
- Rajewsky N, Socci ND: Computational identification of microRNA targets. Dev Biol 267: 529–535. 10.1016/j.ydbio.2003.12.003
- Lewis BP, Shih I-h, Jones-Rhoades MW, Bartel DP, Burge CB: Prediction of Mammalian MicroRNA Targets. Cell 2003, 115: 787–798. 10.1016/S0092-8674(03)01018-3View ArticlePubMedGoogle Scholar
- Kiriakidou M, Nelson PT, Kouranov A, Fitziev P, Bouyioukos C, Mourelatos Z, Hatzigeorgiou A: A combined computational-experimental approach predicts human microRNA targets. Genes Dev 2004, 18: 1165–1178. 10.1101/gad.1184704PubMed CentralView ArticlePubMedGoogle Scholar
- Llave C, Xie Z, Kasschau KD, Carrington JC: Cleavage of Scarecrow-like mRNA targets directed by a class of Arabidopsis miRNA. Science 2002, 297: 2053–2056. 10.1126/science.1076311View ArticlePubMedGoogle Scholar
- Rhoades MW, Reinhart BJ, Lim LP, Burge CB, Bartel B, Bartel DP: Prediction of plant microRNA targets. Cell 2002, 110: 513–520. 10.1016/S0092-8674(02)00863-2View ArticlePubMedGoogle Scholar
- Lim LP, Lau NC, Weinstein EG, Abdelhakim A, Yekta S, Rhoades MW, Burge CB, Bartel DP: The microRNAs of Caenorhabditis elegans. Genes Dev 2003, 17: 991–1008. 10.1101/gad.1074403PubMed CentralView ArticlePubMedGoogle Scholar
- Aravin AA, Lagos-Quintana M, Yalcin A, Zavolan M, Marks D, Snyder B, Gaasterland T, Meyer J, Tuschl T: The small RNA profile during Drosophila melanogaster development. Dev Cell 2003, 5: 337–350. 10.1016/S1534-5807(03)00228-4View ArticlePubMedGoogle Scholar
- Lai EC: Micro RNAs are complementary to 3' UTR sequence motifs that mediate negative post-transcriptional regulation. Nat Genet 2002, 30: 363–364. 10.1038/ng865View ArticlePubMedGoogle Scholar
- Lagos-Quintana M, Rauhut R, Yalcin A, Meyer J, Lendeckel W, Tuschl T: Identification of tissue-specific microRNAs from mouse. Curr Biol 2002, 12: 735–739. 10.1016/S0960-9822(02)00809-6View ArticlePubMedGoogle Scholar
- The miRNA Registry[http://www.sanger.ac.uk/Software/Rfam/mirna/]
- Pusch O, Boden D, Silbermann R, Lee F, Tucker L, Ramratnam B: Nucleotide sequence homology requirements of HIV-1-specific short hairpin RNA. Nucleic Acids Res 2003, 31: 6444–6449. 10.1093/nar/gkg876PubMed CentralView ArticlePubMedGoogle Scholar
- Kretschmer-Kazemi Far R, Sczakiel G: The activity of siRNA in mammalian cells is related to structural target accessibility: a comparison with antisense oligonucleotides. Nucleic Acids Res 2003, 31: 4417–4424. 10.1093/nar/gkg649PubMed CentralView ArticlePubMedGoogle Scholar
- Bohula EA, Salisbury AJ, Sohail M, Playford MP, Riedemann J, Southern EM, Macaulay VM: The efficacy of small interfering RNAs targeted to the type 1 insulin-like growth factor receptor (IGF1R) is influenced by secondary structure in the IGF1R transcript. J Biol Chem 2003, 278: 15991–15997. 10.1074/jbc.M300714200View ArticlePubMedGoogle Scholar
- Varani G, McClain WH: The G × U wobble base pair. A fundamental building block of RNA structure crucial to RNA function in diverse biological systems. EMBO Rep 2000, 1: 18–23. 10.1093/embo-reports/kvd001PubMed CentralView ArticlePubMedGoogle Scholar
- Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 1997, 25: 3389–3402. 10.1093/nar/25.17.3389PubMed CentralView ArticlePubMedGoogle Scholar
- Michael MZ, O' Connor SM, van Holst Pellekaan NG, Young GP, James RJ: Reduced accumulation of specific microRNAs in colorectal neoplasia. Mol Cancer Res 2003, 1: 882–891.PubMedGoogle Scholar
- Yekta S, Shih IH, Bartel DP: MicroRNA-directed cleavage of HOXB8 mRNA. Science 2004, 304: 594–596. 10.1126/science.1097434View ArticlePubMedGoogle Scholar
- Scadden AD, Smith CW: RNAi is antagonized by A-->I hyper-editing. EMBO Rep 2001, 2: 1107–1111. 10.1093/embo-reports/kve244PubMed CentralView ArticlePubMedGoogle Scholar
- Tonkin LA, Bass BL: Mutations in RNAi rescue aberrant chemotaxis of ADAR mutants. Science 2003, 302: 1725. 10.1126/science.1091340PubMed CentralView ArticlePubMedGoogle Scholar
- Bioperl: Repetitive DNA[http://bioperl.org/pipermail/bioperl-l/1999-November/003313.html]
This article is published under license to BioMed Central Ltd. This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.