Evolution of MicroRNA Targets

1
Evolution of MicroRNA Targets

• Kai-How Farh
• 9/21/2005

2
Nonconserved MicroRNA targets

• A typical mammalian microRNA has
• 1000-3000 targets in one genome
• About 15-25 of these sites are conserved across
  all mammals
• Subtracting signal to noise, estimated functional
  targets for each microRNA around 100-300
• What about the other 90?

3
MicroRNA targeting specificity

• MicroRNAs target a huge number of genes
• Only 10 of targets are conserved
• There is no difference in conserved
  nonconserved target sites
• Additional determinants of targeting?
• Location and timing of expression?

4
MicroRNAs are tissue specific

• Many microRNAs have highly tissue specific
  profiles
• Expression analysis underestimates specificity of
  miRNAs due to combining tissues

5
Can tissue specificity account for microRNA
targeting specificity?
Profile of MicroRNA miR-1s targets, ranked by
their expression in muscle
Black histogram of all genes on the chip, ranked
by relative expression Blue histogram of
microRNA targets, ranked by relative expression
6
Kolmogorov-Smirnoff Test

• Measures the difference between two cumuluative
  distributions
• P-values calculated by permutation

7
Kolmogorov-Smirnoff Test Demonstration
8
Control for length dinucleotide frequency

• Two factors contribute to the likelihood of
  matching a target site
• Length of UTR (affects the likelihood of matching
  a site in the UTR)
• Dinucleotide frequency (affects the likelihood of
  matching any given 7mer within the UTR)
• Probability of getting a match
• 1-((1-prob_7mer)length)

9
Incorporating information from other genomes

• Instead of looking for conservation of a motif
  across different UTRs, look for avoidance of the
  motif across different UTRs
• Combine UTRs from orthologous mammalian UTRs
• Need to control for long conserved regions

10
Building a tissue-specificity map
Create a matrix of tissues vs microRNAs. Each
entry in the matrix represents the p-value of the
microRNAs targets in the tissue, according to
the KS statistic. In this case, we are using
miR-7, which gives a strong signal for pituitary,
as well as weaker signals for other brain
tissues. Next, fill out the rest of the matrix
with the p-values for different microRNAs.
11
Tissue-specificity of microRNAs as revealed by
target avoidance