Difference between revisions of "Top Ten Tips"

From Phaserwiki
(remove #: formatting (unstable) with <br/> line breaks)
(Update tips, add some for SAD and both)
Line 1: Line 1:
 +
'''Molecular replacement'''<br/>
 
#'''Use the automated search'''<br/>Don't have a quick read of the instructions and feel the need to change the defaults. If Phaser can solve your problem, it can probably solve it with the automated search using default inputs.
 
#'''Use the automated search'''<br/>Don't have a quick read of the instructions and feel the need to change the defaults. If Phaser can solve your problem, it can probably solve it with the automated search using default inputs.
#'''Search for components concurrently'''<br/>You should search for all components of the asymetric unit at the same time, and not in independent jobs. Phaser is not the same as other MR programs in this respect. Prior knowledge is used to build up a tree search. The correct solution for the first component may not be obvious until the last component is placed.
+
#'''Search for components concurrently'''<br/>You should search for all components of the asymmetric unit at the same time, and not in independent jobs. Phaser is not the same as other MR programs in this respect. Prior knowledge is used to build up a tree search. The correct solution for the first component may not be obvious until the last component is placed.
#'''Search order is important'''<br/>The order in which you search for multiple solutions can affect your success in finding a solution. Components that account for the highest proportion of the scattering (highest completeness, highest sequence identity, and lowest B-factors) should be searched for first. If you are not sure which order is best, you can use the PERMUTATIONS ON keyword, which will search with all permutations of your input search components, but if you have a large number of different components be aware that this search will take a long time, especially if the signal from one of the components is not clear.
+
#'''Search separately for domains of flexible proteins'''<br/>If there are domain movements, the individual domains will be better models giving a clearer signal than the whole protein.
#'''You may need to change the packing criteria'''<br/>Sometimes valid solutions are excluded because the number clashes is higher than the allowed default number. Look in the log and summary files to see how many high Z-score solutions Phaser is rejecting because of clashes, and how many clashes there are in each case. In versions 1.x the default packing criterion is very strict - *no* clashes between C-alpha atoms are allowed. In versions 2.x the default packing criterion is relaxed so that the solutions with the lowest number of clashes are allowed, provided that number is 10 or less.
+
#'''Use sculptor'''<br/>For low sequence identity (<30%) or other difficult cases for which a default Phaser run fails, use sculptor to trim off loops that don't align and to prune back non-identical side chains. Test models derived with all the protocols.  The best results are obtained with accurate sequence alignments, which can be obtained for instance from HHpred, PROMALS3D or FFAS.
#'''You may need to change the peak selection criteria'''<br/>By default, Phaser selects peaks that are between 75% and 100% of the difference between the top peak and the mean of the search for inclusion in the next round of the search (rotation or translation function). If you have many components in the asymmetric unit, the signal from the first rotation search may be very low, and you may need to decrease the cut-off for the rotation search to 70% or 65% using FINAL ROT SELECT PERCENT 65.
+
#'''Use ensembler'''<br/>Molecular replacement can be difficult when the sequence identity is low, but then you probably have several choices of model with similarly low sequence identity.  Construct an ensemble by using the multiple structure alignment in ensembler, and try the option to trim poorly-conserved regions to produce a conserved core.
#'''Beware the automated space group determination'''<br/>The space group is determined from the signal in the translation function search for the *first* model. If you have many components in the asymmetric unit, the signal from this search may be poor, and you should perform and exhaustive search in all the alternative space groups.
+
#'''Homology models do not have 100% identity'''<br/>If you mutate the side chains of your model so that the sequence is the same as the target structure, do not enter an identity of 100% into Phaser. The sequence identity is used to calculate the rms deviation in atomic positions between the search and target structure, and this will only be improved slightly by the most sophisticated modelling algorithms. The identity you enter should thus be the same as that of the original sequence of your model.
#'''Homology models do not have 100% identity'''<br/>If you mutate the side chains of your model so that the sequence is the same as the target structure, do not enter an identity of 100% into Phaser. The sequence identity is used to calculate the rms deviation in atomic positions between the search and target structure, and this has *not* been improved by the mutation (in fact, it has probably been made worse!). The identity you enter should thus be the same as (or less!) than that of the original sequence of your model.
+
#'''You may need to change the packing criteria'''<br/>Sometimes valid solutions are excluded because the number clashes is higher than the allowed default number. Look in the log and summary files to see how many high Z-score solutions Phaser is rejecting because of clashes, and how many clashes there are in each case. In recent versions of Phaser this is less of a problem than it used to be, but it is still worth being aware of.
#'''Reduce the resolution if you can't find second and subsequent components'''<br/>The current algorithms have a limitation in that they do not handle the case where components of the asymmetric unit have greatly varying B-factors. This problem can be worked around by cutting the resolution to about 4 Angstroms, where the B-factor difference is less significant.
+
<br/>'''SAD phasing'''<br/>
#'''Look at the FAQ list'''<br/>If Phaser produces a strange error, or you don't understand some of the output, chances are that you are not the first person to want know what it all means. If we get asked a question three times it usually goes on the FAQ list.
+
#'''Search for different anomalous scatterers concurrently'''<br/>If there are multiple types of anomalous scatterer present in your crystal (such as Fe and S), Phaser will get better results by distinguishing between them. If there is a good signal, the correct hand can be identified as the one with the highest LLG score.
 +
<br/>'''Both'''<br/>
 +
#'''Use the MR-SAD mode'''<br/>After molecular replacement, if there is significant anomalous signal in your data, use the MR-SAD mode to identify the anomalous scatterers and improve the phase information.
 
#'''Install the latest version of Phaser'''<br/>We work hard so you don't have to!
 
#'''Install the latest version of Phaser'''<br/>We work hard so you don't have to!

Revision as of 10:38, 24 April 2012

Molecular replacement

  1. Use the automated search
    Don't have a quick read of the instructions and feel the need to change the defaults. If Phaser can solve your problem, it can probably solve it with the automated search using default inputs.
  2. Search for components concurrently
    You should search for all components of the asymmetric unit at the same time, and not in independent jobs. Phaser is not the same as other MR programs in this respect. Prior knowledge is used to build up a tree search. The correct solution for the first component may not be obvious until the last component is placed.
  3. Search separately for domains of flexible proteins
    If there are domain movements, the individual domains will be better models giving a clearer signal than the whole protein.
  4. Use sculptor
    For low sequence identity (<30%) or other difficult cases for which a default Phaser run fails, use sculptor to trim off loops that don't align and to prune back non-identical side chains. Test models derived with all the protocols. The best results are obtained with accurate sequence alignments, which can be obtained for instance from HHpred, PROMALS3D or FFAS.
  5. Use ensembler
    Molecular replacement can be difficult when the sequence identity is low, but then you probably have several choices of model with similarly low sequence identity. Construct an ensemble by using the multiple structure alignment in ensembler, and try the option to trim poorly-conserved regions to produce a conserved core.
  6. Homology models do not have 100% identity
    If you mutate the side chains of your model so that the sequence is the same as the target structure, do not enter an identity of 100% into Phaser. The sequence identity is used to calculate the rms deviation in atomic positions between the search and target structure, and this will only be improved slightly by the most sophisticated modelling algorithms. The identity you enter should thus be the same as that of the original sequence of your model.
  7. You may need to change the packing criteria
    Sometimes valid solutions are excluded because the number clashes is higher than the allowed default number. Look in the log and summary files to see how many high Z-score solutions Phaser is rejecting because of clashes, and how many clashes there are in each case. In recent versions of Phaser this is less of a problem than it used to be, but it is still worth being aware of.


SAD phasing

  1. Search for different anomalous scatterers concurrently
    If there are multiple types of anomalous scatterer present in your crystal (such as Fe and S), Phaser will get better results by distinguishing between them. If there is a good signal, the correct hand can be identified as the one with the highest LLG score.


Both

  1. Use the MR-SAD mode
    After molecular replacement, if there is significant anomalous signal in your data, use the MR-SAD mode to identify the anomalous scatterers and improve the phase information.
  2. Install the latest version of Phaser
    We work hard so you don't have to!