MicroRNA Targets Prediction and Analysis

1. MicroRNA Targets Prediction and Analysis

2. Small RNAs play important roles The Nobel Prize in Physiology or Medicine for 2006 Andrew Z. Fire and Craig C. Mello for their discovery of "RNA interference – gene silencing by double-stranded RNA"

3. How microRNA regulates the target mRNA genes • MicroRNAs are ~22nt noncoding small RNAs that influence mRNA stability and translation. • Lin-4 (Lee et al. 1993) • Let-7 (Reihart et al, 2000)

4. There are three problems • How to find microRNA genes? • Given a microRNA gene, how to find its targets? • Target-driven approach: • Xie et al. (2005) analyzed conserved motifs that are overrepresented in 3’ UTRs of genes • Found out they are complementing the seed sequences of known microRNAs. • They predicted 120 new miRNA candidates in human.

5. How to find microRNA genes? • Biological approach • Small-RNA-cloning to identify new small RNAs • Most MicroRNA genes are tissue-specific miR-124a is restricted to the brain and spinal cord in fish and mouse or to the ventral nerve cord in fly miR-1 is restricted to the muscles and the hart in mouse

6. Computational approaches to find microRNA genes • MiRscan (Lim, et al. 2003) • Scan to find conserved hairpin structures in both C. elegans and C. briggsae. • Using known microRNA genes (50) as training set.

7. MiRscan predicted results • Blue: distribution of MiRscan score of 35,697 sequences • Red: training set • Yellow and purple are verified by cloning or other evidence.

8. Other approaches • RNAz (Washietl et al. 2005) can be also used to detect micRNAs.

9. How to find MicroRNA targets? • MicroRNA targets are located in 3’ UTRs, and complementing mature microRNAs • For plants, the targets have a high degree of sequence complementarity. • But for animals, not this case. • It is short (~21 nt) • If we allow G-U pairs, mismatches, gaps (bulges), we will find a lot false positives using standard alignment algorithms. • How to remove the false positives?

10. How to make more accurate predictions? • Incorporating mRNA UTR structure to predict microRNA targets (Robins et al. 2005) • Make sure the predicted target “accessible”. • Not forming basing pairing its self.

11. Other properties of microRNA targets • MicroRNA targets conserve across species. (Stark et al. 2003) • Tends to appear in a cluster. For lins, comparison is between C. elegans and C. briggsae. For hid, comparison is between D. melanogaster and D. pseudoobscura.

12. Another property of microRNA targets • Sequence conservations of target sites • Better complementary to the 5’ ends of the miRNAs.

13. Another property of microRNA targets • Strong binding on miRNA 5’ end. (Doench and Sharp 2004) • MicroRNA and targets form simple helices structure • It is not a complex secondary structure. • It also has relatively good energy.

15. Most target predictions methods try to use these properties • “miRanda” Enright et al. Genome Biology 2003 • Use a weight scheme that rewards complementary pairs at the 5’ ends do the microRNA.

16. Targetscan (Lewis et al. Cell 2003) Given a microRNA that is conserved in multiple species and a set of orthologous 3’ UTR sequences: • Use 7 nt segment of the miRNA as the ‘microRNA seed’ to find the perfect complementary motifs in the UTR regions. • Extend each seeds to find the best energy • Assign a Z score. • Rank Give a rank (Ri) according to that species. • Repeat above process. • Keep those genes for which Zi > Z_c and Ri < R_c.

17. Profile based target search (Stark et al. Plos Biology 2003 • Building profiles for each microRNA family (using HMMer) for first 8 residues, allowing for G:U mismatches. • Only search conserved 3’ UTRs (in two fly genomes) using the profiles. • Sequence matches are extended to miRNA length + 5nt. • Compute the energy using the Mfold and provide the z-scores.

18. Three Classes of microRNA Target Sites(Brennecke et al. Plos Biology 2005)

19. Seed based target search (Stark et al. Cell 2005) • Find all 8 – 4 mers complementary to the 5’ end of miRNA. • For 8mers, allow one gap • For 7mers, allow one mismatch • For each match, we extract the 3’ adjacent sequence from the both genomes (two fly genomes). • Predict the base-pairing and compute the energies. Use the worse one. • 8mers with G-U pair or loop on the target site, energy of 3’ side >= 50% • 8mers with one mismatch or loop the microRNA site 7mers with a G-U pair, 6mers, energy of 3’ side >= 60% • 5mers, energy of 3’ side >= 70% • 4mers, energy of 3’ side >= 80% • Normalize the 5’ and 3’ energy to get the Z scores.

20. Seed based target search (Stark et al. Cell 2005) • Based on the statistical signal, 5’ scores are weighted: • 8mers 2.8× • 7mers 2× • 6mers by 1.2× • 5’ and 3’ scores are added to get final scores for each sites. • The UTR score is the sum of all sites (non-overlapping seeds). • Shuffled microRNA controls. • Use 10 shuffled microRNA for each the 39 cloned 5’ non-redundant miRNAs. • So that the shuffled sequences have an equal number of matches (+/-15%) in the D. melanogaster 3’UTR. • Targets were predicted for shuffled microRNAs using the same methods. Genome-wide occurrences Of conserved 5’ seed matches

21. Summary of Target Validation • 8/9 are validated by reporter assay. • Performance on 133 experimentally text microRNA target pairs: • 72 functional • 62 not • 50/62 predicted by the program. • Compared with 34/62 predicted by Stark et al. 2003

22. Interesting Properties of microRNA targets • Clusters of microRNA targets • Extensive cooccurrence of the sites for different microRNAs in target 3’ UTRs.

23. Presence and absence of target sites correlate with gene function

24. Target site density of Target and Antitarget 3’UTRs • The explanation of significant site avoidance is that antitarget genes are required in the miRNA-expressing cells and miRNA-mediated repression would be detrimental.