dna_denovo

[Anonymous]

Hi all,
is there a fundamental reason that there’s no dna_denovo algorithm?
Best regards,
Frits.

[Anonymous]

Not a fundamental one, no.

There’s an RNA denovo algorithm because people in the Rosetta community like Rhiju Das are interested in RNA folding and predicting RNA structure. There as of yet hasn’t been someone who is as interested in DNA structure prediction, so comparatively work has been made toward designing algorithms for it.

That said, the basic mechanism for RNA structure prediction could likely be re-purposed for DNA structure prediction. It would take a bit of work to track down and fix any residue type assumptions in the rna_denovo code, and more importantly some scientific work to determine how to (slightly) alter the sampling and scoring schemes to account for the differences in RNA versus DNA biases, but there’s no fundamental scientific reason or program architectural reason why a dna_denovo protocol couldn’t be developed.

[Anonymous]

Frits,

Phil Bradley (pbradley@fhcrc.org) has done a lot of work on DNA score functions, terms, and modeling with Rosetta. Unfortunately, much of his work is in his lab’s branch. I have emailed him in the past for help adding DNA to DNA in order to create larger DNA structures such as the extended nucleosome. His main DNA denovo type method is to paste nucleotides together to form double helical DNA, and do some modeling of the connecting bonds and a few residues around the connection. If this is something your interested in, you could try emailing him for help. It worked very well. A lot of his Rosetta DNA papers have used this type of algorithm, so I would imagine he wouldn’t mind helping.

Also FYI,
The DNA and general scorefunctions seem to have some mixed results depending on what your doing with the DNA. There has been success in DNA/protein design, but we didn’t find the same level of success for relaxing DNA/protein structures. We did, however, have great success using a new orbitals scorefunction from a different Rosetta lab (which is still being worked on in the developers version – look for it and the resulting paper from the Meiler lab (http://www.meilerlab.org) sometime within the next year). Any other scorefunction we tried would rip the double helix apart, and result in extended strands of nucleotides.

[Anonymous]

Thanks jadolfbr,
we’re actually interested to model short (30 Regards,
Frits.

[Author]

Posts