Binding site prediction with RNAduplex

The file duplex.seq contains the 3'UTR of NM_024615 and the microRNA mir-145.

	  $ RNAduplex < duplex.seq
	  >NM_024615
	  >hsa-miR-145
	  .(((..............................(((((.(((...((((((((((.&))))))))))))))))))))).   1,57  :   1,22  (-22.20)

Most favorable binding has an interaction energy of -22.2 kcal/mol and pairs up on positions 1-57 of the UTR with positions 1-22 of the miRNA.

RNAduplex can also produce alternative binding sites, e.g. running RNAduplex -e 10 would list all binding sites within 10 kcal/mol of the best one.

Since RNAduplex forms only intermolecular pairs, it neglects the competition between intramolecular folding and hybridization. Thus, it is recommended to use RNAduplex as a pre-filter and analyse good RNAduplex hits additionally with RNAcofold or RNAup. Using the example above, running RNAup will yield:

	  $ RNAup -b < duplex.seq

	  >NM_024615
	  >hsa-miR-145
	  (((((((&)))))))  50,56  :   1,7   (-9.13 = -10.30 + 0.70 + 0.48)
	  GCUGGAU&GUCCAGU
	  RNAup output in file: hsa-miR-145_NM_024615_w25_u1.out

The free energy of the duplex is -10.30 kcal/mol and therefore shows a discrepancy computed by RNAduplex (differences maybe from the fact that RNAup computes partition functions rather than optimal structures). However, the total free energy of binding is much less favorable (-9.13 kcal/mol), since it includes the energetic penalty for opening the binding site on the mRNA (0.70 kcal/mol) and miRNA (0.48 kcal/mol). The -b option includes the probability of unpaired regions in both RNAs.

You can also run RNAcofold on the example to see the complete structure after hybridization (neither RNAduplex nor RNAup produce structure drawings). Note however, that the input format for RNAcofold is different, an input file suitable for RNAcofold is given in duplex2.seq.

As a more difficult example let's look at the interaction of the bacterial smallRNA RybB and its target mRNA ompN. First we'll try predicting the binding site using RNAduplex:

  $ RNAduplex < RybB.seq
>RybB
>ompN
.(((.((((..((((((.(((....((((((((..(((((.((..((.((....((((..(((((((((((..(((((.&.))))).))))))).)))) /
.....))))....)).)).)).))).))..))))........))))..))).)))))).))))......))).   1,79  :  82,173 (-34.40)

Note, that the predicted structure spans the full length of the RybB small RNA. Compare the predicted interaction to the structures predicted for RybB and ompN alone, and ask yourself whether the predicted interaction is indeed plausible.

Below the structure of ompN on the left and RybB on the right side.

GCCAC-----TGCTTTTCTTTGATGTCCCCATTTT-GTGGA-------GC-CCATCAACCCCGCCATTTCGGTT---CAAG-GTTGGTGGGTTTTTT
 |||      ||||  |||||| |||    ||||| ||||        || ||| || ||  ||    ||||     |||| ||  |||  |||||| -40.30
AGGTCAAACAACGGC-AGAAACAATATT--TAAAGTCGCCGCACACGACGCGGTCGTCGGT-CGTCTCGGCCCTACTGTTCACGGTTATGAAAAGAAACC-3'

Compare the RNAduplex prediction with the interaction predicted by RNAcofold, RNAup and the handcrafted prediction you see above.