On Nov 23, 2010, at 12:57 AM, Wee Kiang Yeo wrote:
> Hi, Dr Flores,
>
> I am a graduate student at the National University of Singapore.
> Ermm... perhaps that's kinda far from your current location?
>
Ha yes, quite far. However we are still on the same planet. If you think a sufficient number of people would be interested it might be possible to put together the funding for a mini-workshop. Sometime around the time of the RNA society meeting in Japan next summer, for instance. Anyway, plenty of time to think about such things.
> Yes, please post the question and answer on the mailing list.
>
Ok, this will be my default behavior. Please send any further questions directly to the list and I will reply on there.
> Is there a preferred way to treat insertion/deletions and mutations
> during threading? My current approach = Only use the Superimpose
> command on the nucleotides that match exactly and no such commands for
> any others.
I would go ahead and use the Superimpose forces on substitution mutations as well. Based on what I see you are making a good model and are nearly finished. There is no real need for a major strategy shift.
The ribosome is highly conserved across organisms as you know. However if you had insertions (e.g. an entire stem-loop that exists in the model but not in the template) then you could solve the inserted part by applying WatsonCrick interactions to form any helices, and other interactions as needed to recover any tertiary contacts. Modeling in regions of low sequence identity is an art, but a good account of how we deal with this is given in our 2010 PSB paper, "Predicting RNA structure by multiple template homology modeling." Bear in mind that RNABuilder studiously avoids thinking for you. It is not a physics-based simulation code, it is more like a molecular "erector set" which only does what you tell it to do.
>
> I also tried to thread the original sequence from the PDB template
> structure onto the template structure. But I noticed that, despite
> enforcing Superimpose for every nucleotide residue, there are some
> recalcitrant ones which refuse to superimpose. The main problem, I
> observed from the VMD animations, is that some RNA segments get
> 'entangled' with other segments. Consequently, they are unable to
> fold/superimpose correctly. To help you visualise what I mean, try
> place placing your right arm over your left elbow and pull them
> towards your chest. I am still rather puzzled with why this happens
> since the threading should proceed from one end to the other. The same
> entanglements also occur with the homologous (non-template) RNA
> sequence.
Ah yes. The phenomenon you are referring to is a kind of kinetic trapping, and is a recurring issue in modeling. There are multiple ways of dealing with this.
1. Turn off the sterics until the Superimpose contacts have been fully or mostly satisfied. Then turn the sterics back on at a subsequent stage.
2. Apply sterics to the entire threaded chain EXCEPT those residues where you have a tangle, e.g. where the leftmost arrow is pointing in your graphic.
3. Use the "scrubber" algorithm.
The Scrubber is my standard means of resolving kinetic traps. It is a combination of simulated annealing and potential rescaling. This means that I start with a high temperature and decrease it stage by stage. At the same time, I periodically turn all forces (including sterics) on and off. In a final stage I stop cycling the forces and let the molecule converge at low temperature. It sounds complicated but really it's not hard and is the most reliable way of getting out of traps. Here is a snippet that would do this:
# I imagine the model you sent was built in stage 2. So start at stage 3 so you don't have to re-fold:
firstStage 3
# and end at stage 5
lastStage 5
# start at some relatively higher temperature. If need be you can try 150 or even 1000. However 10 might be enough:
temperature 10 3
# then lower the temperature:
temperature 1 4
# stage 5 is when you are converged or close to it:
temperature 1 5
# a dutyCycle of .95 means the forces are turned off 5% of the time.
dutyCycle .95 3
dutyCycle .95 4
# a dutyCycle of 1 means forces are on all the time, as they are in a more conventional simulation:
dutyCycle 1 5
# this means that you want the baseInteraction as well as the steric forces to be cycled (turned on and off periodically).
setForceAndStericScrubber true
# This is the duration, in picoseconds, of one scrubber cycle. I don't know if this will be long enough, you might make it as much as 10 times longer.
scrubberPeriod 16.0
Every system is a little different so you might need to adjust the temperature's and dutyCycle's, as well as the reportingInterval and numReportingIntervals.
Let me know how you do. If need be, you can send me your contacts.dat and last.1.pdb and I can debug. But please try this on your own first and see how things go.
Sam
>
> This is the first time I am modeling RNA structure, hence I may be
> rather unfamiliar with some of above-mentioned problems.
>
> Regards,
> Wee Kiang YEO (Mr)
>
>
> On 22 November 2010 18:28, Samuel Flores <samuelfloresc(a)gmail.com> wrote:
>> can i post your question and answer to the RNABuilder-help list, so others can benefit?
>>
>> On Nov 22, 2010, at 3:20 AM, Wee Kiang Yeo wrote:
>>
>>> Dear Dr Flores,
>>>
>>> Two days ago, I have used RNABuilder to thread an RNA sequence through
>>> a X-ray structure of 23S RNA (obtained from PDB). This X-ray structure
>>> contains a ligand in a binding pocket. The original intention of
>>> threading the RNA sequence was to ascertain the nucleotide residues
>>> located near the ligand. However, during threading, the template X-ray
>>> structure actually translocated in 3D space, leaving the
>>> previously-bound ligand behind.
>>>
>>> My question: Is there a way to fix the position of the RNA-ligand
>>> complex template during the threading process? If so, I would
>>> appreciate it if you can give me an example of how to do so.
>>>
>>> Thanks.
>>>
>>> Regards,
>>> Wee Kiang YEO (Mr)
>>
>>
