Dear Reza
Using unconventional search models - ensembles that have been automatically truncated (often brutally) - AMPLE regularly solves structures with search models containing as little as 5-10% of the target chain (although somewhat larger search models tend to work better).
eg Fig 4 here http://journals.iucr.org/d/issues/2012/12/00/tz5014/index.html
or Supp Fig S3b here http://journals.iucr.org/m/issues/2015/02/00/lz5005/
Best wishes
Daniel
________________________________________
From: CCP4 bulletin board [[log in to unmask]] on behalf of Randy Read [[log in to unmask]]
Sent: 30 August 2015 09:43
To: [log in to unmask]
Subject: Re: [ccp4bb] Minimum model size for successful MR
Dear Reza,
It turns out that the way you have to look at this is in terms of how accurate the model is and how many reflections you have to a resolution where the model is accurate enough to help explain the data, in addition to what fraction of the structure the model accounts for. We’ve had a feature in Phaser for a couple of years now (which we’re finally writing up!) to compute the eLLG, i.e. the expected value of the log-likelihood-gain score that should be found from a model, based on the fraction scattering accounted for, the expected RMS error of the model, the total number of reflections and the resolution limit. The same eLLG can be obtained in a variety of ways (complete poor model, very incomplete but highly accurate model with high resolution data, etc.), but the actual value of the eLLG is an indication of how hard it will be to solve the structure. On the basis that finding a one-component solution with an LLG of 60 generally means that it’s right, we classify something with an eLLG < 60 as very hard, between 60 and 120 as hard, and above 120 as straightforward. Of course, this all depends on how good the assumptions are: e.g., if there’s an unexpected conformational change, then the RMS error of the model will be higher than expected.
One nice thing is that this calculation helps to predict when Isabel Uson’s Arcimboldo procedure (placing single helices or other fragments to solve a structure ab initio) is likely to work.
Best wishes,
Randy Read
-----
Randy J. Read
Department of Haematology, University of Cambridge
Cambridge Institute for Medical Research Tel: +44 1223 336500
Wellcome Trust/MRC Building Fax: +44 1223 336827
Hills Road E-mail: [log in to unmask]
Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk
> On 29 Aug 2015, at 19:29, Reza Khayat <[log in to unmask]> wrote:
>
>
> Hi,
>
> Does anyone know, or can estimate, what the smallest percentage of a structure a model must to be in order to attain a successful molecular replacement solution? For example, if a protein is composed of domains A and B, what is the smallest fraction of the A+B structure can B be in order to solve for the structure? Lets assume that B is rigid in both the model and the solution. Thanks.
>
> Best wishes,
> Reza
>
> Reza Khayat, PhD
> Assistant Professor
> City College of New York
> Department of Chemistry
> New York, NY 10031
>
|