3
$\begingroup$

I have a list of regions of the human genome and I want to predict if single-stranded molecules in a buffer would tend to fold and create pin structures by sequence self-complementarity. What's the most precise software/parameter set I can use to such effect?

TRIED SO FAR: Vienna RNALFold (unsure if applying an RNA software to DNA is adequate):

Compute locally stable RNA secondary structure with a maximal base pair span. 
For a sequence of length n and a base pair span of L the algorithm uses only 
O(n+L*L) memory and O(n*L*L) CPU time. Thus it is practical to "scan" very 
large genomes for short RNA structures.
Output consists of a list of secondary structure components of size <= L, one 
entry per line. Each output line contains the predicted local structure its 
energy in kcal/mol and the starting position of the local structure.

Example:

samtools faidx $ref $region | seqtk seq -A - | $HOME/viennarna/bin/RNALfold --noconv
>chr1:100000-100150
.((((.(......).)))). ( -2.70)  127
.(((.((........)).))). ( -5.70)  123
.(((((((.((........)).)))..)))). ( -6.30)  119
.(((.(.(((.((.....))))).)..))). ( -7.00)  117
.((((.((........)).)))). ( -8.30)  109
.((((..((.((..(..(((.....)))..)..)))))))). (-10.10)  103
.((((.((..((..(((...((((.((........)).)))).....)))))..)))))). (-10.50)   90
.((.((.((((....................)))).)).)). ( -4.10)   77
.((((.((.((((....................)))).)).)))). ( -5.70)   75
.((((..((((.((....)).)))).)))). ( -2.40)   72
.((.((..((((..((((.((....)).)))).))))..)))). ( -5.50)   65
.(.((..(..................((.((..((((..((((.((....)).)))).))))..))))..((((.((........)).)))).)..)).). (-16.20)   40
.((((......(.((..(..................((.((..((((..((((.((....)).)))).))))..))))..((((.((........)).)))).)..)).).....)))). (-18.10)   30
.(((....(((((((((...(.........)..)))).)))))((.((..((((..((((.((....)).)))).))))..))))..((((.((........)).))))......))). (-18.50)   23
.((....(((....(((((((((...(.........)..)))).)))))((.((..((((..((((.((....)).)))).))))..))))..((((.((........)).))))......)))....)). (-18.70)   17
.(((((((((......((...(((((..((((...((...))....))))..)))))...))......))))))))). ( -9.60)    7
.((.(((((((((......((...(((((..((((...((...))....))))..)))))...))......))))))))).)). (-10.90)    4
.(((..(((((........)))))..))). ( -3.60)    1
ACTTAAGTTGTAGAAGGGAATAACGCAAGAGTGAATTTAGGGCGGGGCAAAAGGATAAATTTTACGGTACAAAGTTTCTACGGGTTTATATATGTATAACTAAGTCCAAGCGCGGGGGATATGGCCAGTGCACAACGGCGGGCATCATAAT
 (-21.70)
$\endgroup$
3
  • $\begingroup$ I think this is a nice question but a bit open-ended. Are you searching for other tools because this didn't convince you ? Or what requirements are you looking in a other tools? How do you define "best"? Less FDR or more precision ? $\endgroup$
    – llrs
    Jan 17 '18 at 10:23
  • $\begingroup$ I edited my question to "most precise" and added that the reason I am unsure about RNALFold is because it's an RNA software applied here to DNA molecules. $\endgroup$
    – 719016
    Jan 17 '18 at 10:48
  • $\begingroup$ Great edit, I didn't realize you were using DNA as input, I thought you were using RNA from those regions $\endgroup$
    – llrs
    Jan 17 '18 at 10:54
1
$\begingroup$

You could also take a look at Infernal.

I think your best bet is to use RNA-focused programs; I'm not aware of a DNA-specific one, which makes some sense because DNA is typically double-stranded, so the secondary structure community is primarily focused on RNA. I would predict that secondary structure potential of single-stranded DNA would be similar to that of RNA.

$\endgroup$
1
  • $\begingroup$ Thanks for recommending. I learned about Infernal about 5-8 years ago, and after reading the documentation as it is now, it seems to me that there isn't an easy way of asking for the folding of DNA sequences. The commands seem to be based on building alignments. Am I wrong? $\endgroup$
    – 719016
    Jan 18 '18 at 16:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.