09 - RNA structures

• fRNA: functional

• cRNA: coding

• mRNA: messenger

• rRNA: ribosomal

• tRNA: transfer

• snRNA: small nuclear

• snoRNA: small nucleolar

• miRNA: micro

• siRNA: small interfering

• Structure

  • identification

  • comparison

  • alignments

• Predict secondary structure

• RNA databases

• RNA motifs

• RNA gene finding

  • Specialized

  • Generalized

• codons in an mRNA molecule do not directly recognize the amino acids they specify

• translation into protein depends on adaptors:

  • recognize and bind to

    • mRNA codon and

    • amino acid

  • consist of a set of small RNA molecules known as transfer RNAs (tRNAs)

  • ~ 80 nucleotides long

• watson-crick

  • A-T (A-U)

  • G-C

• wobble pair: G-T (G-U)

  • only 2 hydrogen bonds instead of 3

• secondary = intermediate step to 3D structure

• primarily: own single-strand to double -> loop structures

  • bases must be complementary

• Energy minimization methods:

  • choose most energetically stable structure

  • Pro: experimental + alignment data

  • Con: no tertiary, very computationally intensive

• Evolutional conservation methods:

  • Covariance models (CM)

  • patterns of conserved base-pairings

  • challenge: discover pattern vs random background

  • Pro: simple, 2nd & 3rd structure

  • Con: requires sufficient number of sequences

• repeats —> large complementary structures —> potential stems/hairpins

• basic example: base pair maximization

  • scoring: +1 per base pair, 0 otherwise

  • get best structure for range i to j in complete sequence

  • Key idea: optimal score S(i,J) -> recursively defined for smaller subunits

4 cases:

1. i,j = base pair -> added to i+1, j-1 —> score+1

2. i unpaired -> added to i+1 —> score+0

3. j unpaired -> added to j+1 —> score+0

4. i paired, j paired -> adds two subsequences —> S(i,k) + S(k+1,j)

Score: —> max of all 4 cases

• Initialize

• Recursive fill —> until S(1,N) is reached

• Traceback

—> Result:

Efficiency:

• matrix storing -> mem = N^2 —> OK

• but: search for step 4 -> time = N^3 —> not good

Limitations:

• best base pairings not always energetically best structure

• pseudoknots break algorithm

• energetically most stable = most likely structure

• energy of base pair is only influenced by previous base pair

Basis:

• Thermodynamic stability —> most stable = most likely form

• cant to pseudoknots

• dynamic algorithm

• e.g., ViennaRNA, MFOLD

  
  

Algorithm:

• find complements

• minimal energy matrix values:

  • consider the minimum energy values obtained by all previous complementary base pairs

  • complementary -> decrease by stacking energy

  • noncomplementary -> increase by the destabilizing energy

    • depends on what loop is formed

      • internal, buldge, hairpin

• Repeat for entire matrix

• Principle:

  • covariated bases coevolve —> ensures that base pairing is maintained -> RNA structure conserved

  • Goal: improve RNA structure prediction by giving weight to conserved regions

• Features:

  • can be constructed automatically from alignments

  • iterative training

  • consensus secondary structure prediction

  • optimal algorithm -> based on pairwise covariations in multiple alignments

  • Covariation-> base pairing and structure = conserved

Mutual information (don’t have to know formula)

Database of ncRNA families -> MSA