- IMP-dev - salilab.org

Re: [IMP-dev] patches to embed
by Keren Lasker 16 Jul '08

16 Jul '08

Frido - can you point me to the EMDep unittests? On Jul 16, 2008, at 2:34 PM, Friedrich Foerster wrote: > keren will be the appropriate addressee. > > anyway, some comments from my side: > - thanks for propping up the error handling > - for the endian: best provide a unit test and make sure your files > (mrc?) are correct! for example, if your magic number indicates big > endian, but the data are actually stored as little endian, your > output will be messed up - but the reader works perfectly as it > should do. so best take an example from the EMDep for the unit test. > might well be there is a problem, but make sure it is not due to > messy files. > - in DensityHeader there is a confusing comment: > ... > float Objectpixelsize; //this is the actual pixelsize > float Microscope; //Microscope > + //! Use this to set the size of the voxel > float Pixelsize; //Pixelsize - used for the microscope CCD camera > ... > the comment should probably apply to Objectpixelsize. > > best > > frido > > > > On Jul 16, 2008, at 1:47 AM, Daniel Russel wrote: > >> <patches> > > > > -- > > Friedrich Foerster > Max-Planck Institut fuer Biochemie > Am Klopferspitz 18 > D-82152 Martinsried > > Tel: +49 89 8578 2651 > > foerster(a)biochem.mpg.de > > www.tomotronic.org > > >

1 0

number_of_->get_number_of_
by Daniel Russel 15 Jul '08

15 Jul '08

Just to make things consistent (and make it so I can submit more useful changes), here is a patch which replaces the scattered uses of number_of_foo with get_number_of_foo as we had some of each before.

1 0

Transformed pair score
by Daniel Russel 08 Jul '08

08 Jul '08

Since I needed symmetry restraints I am adding a TransformedDistancePairScore which applies a linear transform to the coordinates of one of the particles before scoring the pair based on the distance. This, coupled with the PairListRestraint can be used to easily set up general symmetry restraints on particles.

1 0

Moving Pointer
by Daniel Russel 07 Jul '08

07 Jul '08

Some quick changes moving IMP::internal::ObjectPointer to IMP::Pointer. All pointers to IMP objects kept in C++ code should already use this type as otherwise reference counting won't happen. Note that this patch uses svn move, which caused some issues before, I think, so there may be problems.

1 0

Proposal: TypedPairScore
by Ben Webb 02 Jul '08

02 Jul '08

For statistical potentials, TypedPairScore would delegate to another PairScore depending on an integer attribute of the particles: TypedPairScore(IntKey typekey) { create empty map(pair(atype,btype), PairScore*) } void set_pair_score(PairScore *ps, Int atype, Int btype) { swap if needed so atype <= btype set map[pair(atype,btype)] = ps } virtual void set_particle_type(Particle *p) {} virtual Float evaluate(a, b, da) { if a does not have typekey, call set_particle_type to set it atype = a.get_value(typekey) same for b, btype swap if needed so atype <= btype return map[pair(atype,btype)].s(a,b,da) } For something like DOPE, a subclass would a) provide an additional map, from the DOPE residue name:atom name pairs to DOPE types, and b) override set_particle_type to set the DOPE type of a particle by querying the atom and residue decorators. Comments? Ben -- ben(a)salilab.org http://salilab.org/~ben/ "It is a capital mistake to theorize before one has data." - Sir Arthur Conan Doyle

2 1

various changes
by Daniel Russel 02 Jul '08

02 Jul '08

First, the dna reading patch seems to have gotten lost somewhere, so here it is again (perhaps with a few changes) Second, some cleanup of the comments in the grid add set methods for the linear parameters fix header include order on opt fix a bad bug in sqrt with units change the constant internal::T to internal::DEFAULT_TEMPERATURE since T is a bit short of a name :-) Add the tunnel restraint which keeps particles out of an axis aligned slap with a hole in it. Kind of specific, but...

1 0

Various proposed changes
by Daniel Russel 30 Jun '08

30 Jun '08

- For the Domino optimizer, the right addition on restraints may be to have them return a list of pairs of interacting particles. This way we don't have to worry about the granularity of the restraint and nonbonded lists can just return their current list of nonbonded pairs. - The ChainPair restraint could be simplified to only contain one chain now that we have reference counting on the PairScores (so you can share pairscores via shared objects rather than having one restraint). This simplifies the restraint code a bit. - I think ObjectPointer should be made part of the API now that we have reference counting and renamed to IMP::Pointer and it is pretty stable.

3 2

Re: [IMP-dev] restraint specific derivatives
by Ben Webb 26 Jun '08

26 Jun '08

Friedrich Foerster wrote: > i spread imp throughout the world. even in exotic bavaria imp is > functional now. > here, i am trying ever more challenging things: i am using restraints > from different sources and want to weight these terms differently. for > that purpose, i want to employ a common strategy in x-ray: the different > restraints are weighted according to the inverse magnitude of the > corresponding derivatives. > Do you know a way how to get the derivatives of a particle with respect > to a specific restraint set in IMP. Particle.get_derivative uses the > full scoring function as i understand it... It's not something that's immediately possible, as I see it, in IMP currently. But I suppose you could go through every *other* restraint and call set_is_active(false). Otherwise, if what you want to do is precisely that - weight the restraints by the derivatives - then perhaps we could add extra functionality to the DerivativeAccumulator to allow for that. Ben -- ben(a)salilab.org http://salilab.org/~ben/ "It is a capital mistake to theorize before one has data." - Sir Arthur Conan Doyle

3 2

reading dna
by Daniel Russel 23 Jun '08

23 Jun '08

the IMP pdb reader now reads (at least some) dna. Ben should probably look through the code to make sure I am not abusing modeller somehow. Also IMP.read_pdb now returns a MolecularHierarchyDecorator rather than a Particle. This seems more reasonable to me as you need the former in order to do anything useful with it. Hopefully now files are missing.

1 0

Adding refineOncePairScore
by Daniel Russel 23 Jun '08

23 Jun '08

I think I forgot to add the files before sending the last patch. This is a pair score which uses a particle refiner to refine each of the particles once before calling a passed pair score on each of the generated pairs. Add a comment to restraint to point out that it can now be added to multiple sets: Clean up ref counted Check the arguments for the wlc constructor since it has two float arguments and the order is not obvious a priori.

1 0