MiRNA

• leader: Molly
  

  Readings

  If miRNA's seem like a new topic, then first read this article from the NYTimes placed on the web at this link (probably illegally); be sure to click on the "MICRO-RNA HAIRPIN" picture to enlarge it- this gives a very clear picture of miRNA's are formed: "RNA trades bit part for starring role in the cell" http://www.sickofdoctors.addr.com/articles/rna_messenger2.htm If miRNA's are familiar or after reading the NYTimes article, read "The microRNA world: small is mighty." by Nelson P, Kiriakidou M, Sharma A, Maniataki E, Mourelatos Z. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=14559182&dopt=Abstract and then read "Prediction of mammalian microRNA targets." by Lewis BP, Shih IH, Jones-Rhoades MW, Bartel DP, Burge CB. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=14697198&dopt=Abstract if there's time, read "Vertebrate microRNA genes." by Lim LP, Glasner ME, Yekta S, Burge CB, Bartel DP. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12624257&dopt=Abstract&itool=iconnoabstr

  summary

  • Biology:
    miRNA's are small (~22 nuc) pieces of RNA which can bind with or without perfect complementarity to the mRNA of genes and silence or reduce translation. The exact mechanism by which miRNA's cause the silencing is not well understood. To form a miRNA, a "miRNA gene" is transcribed, a hairpin loop is formed, the Dicer enzyme cuts from one side of the hairpin a small piece of RNA (the miRNA) which is carried by the Argonaut complex to it's target. Since miRNA's have never been observed independently of Argonaut, it may be that Argonaut attaches to the miRNA even before the miRNA is cut from the hairpin loop.
  • Relevance:
    miRNA's are potentially a medical breakthrough because they could be used to target undesirable RNA transcripts (such as transcripts for gene products involved in cancer, viral gene products, etc.). Since miRNA's require the Argonaut complex to carry them to their targets, one idea would be to create desired miRNA's which are similar to known miRNA's so that Argonaut would still carry them to their targets. This underscores the importance of miRNA gene and target prediction, because these predictions facilitate the biological discovery/confirmation of previously unknown miRNA's.

  Questions and Potential Projects

  • 1. Compare the location of miRNA genes to the location of their target genes; can one see a relationship that would be useful in predicting miRNA targets from miRNA genes (for example, could the relationship be used to increase/reduce the score of a potential target based on the distance between the potential target gene and the miRNA gene)?
  • 2. Come up with a method for combining a miRNA target prediction model with miRNA gene prediction model to improve results of both.
  • 3. The Burge target prediction paper uses positions 2-8 of the miRNA as a "seed" which is required to have perfect target complementarity; why not use the most conserved part of the miRNA as the "seed"? In this way, could potentially reduce both false positives and false negatives.
  • 4. Examine the Burge paper's model for the shuffled miRNA's used for the "noise" in the paper's measure of signal:noise ratio; is it appropriate? Can it be improved? (To it's credit, the Burge paper's biological confirmation of 11 out of 15 predicted targets does roughly correspond to the ~30% false positives predicted by their shuffling method.)
  • 5. Can one come up with a better predictive measure of a target model's success then a permutation test (i.e., their signal:noise ratio)? One can think of ways in which this signal:noise ratio could give a false measure of a model's success; for example: suppose there existed less actual miRNA targets than those predicted from "shuffled" miRNA's, not more (perhaps because the model's complementarity requirement was wrong, or because other biological mechanisms unaccounted for in the model such as secondary structure play a role); then larger values (larger then 1) of the paper's signal:noise ratio simply indicate a greater likelihood of false positives, not a greater likelihood of correct predictions. Is there another predictive measure which doesn't have the same vulnerabilities as Burge's measure of signal:noise?
  • 6. miRNA gene discovery is still an open problem. Could investigate relevant papers/software and think about which models or improvements to models may be useful.
  Hanno suggested this reference, generally useful for understanding of background biology for all of our reading topics: Chapter 1 "Microbiology for Computer Scientists", from Lawrence Hunter's freely available online book "Artificial Intelligence and Molecular Biology": http://www.aaai.org//Library/Books/Hunter/hunter.html