Line data Source code
1 : /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
2 : Copyright (c) 2014-2023 The plumed team
3 : (see the PEOPLE file at the root of the distribution for a list of names)
4 :
5 : See http://www.plumed.org for more information.
6 :
7 : This file is part of plumed, version 2.
8 :
9 : plumed is free software: you can redistribute it and/or modify
10 : it under the terms of the GNU Lesser General Public License as published by
11 : the Free Software Foundation, either version 3 of the License, or
12 : (at your option) any later version.
13 :
14 : plumed is distributed in the hope that it will be useful,
15 : but WITHOUT ANY WARRANTY; without even the implied warranty of
16 : MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 : GNU Lesser General Public License for more details.
18 :
19 : You should have received a copy of the GNU Lesser General Public License
20 : along with plumed. If not, see <http://www.gnu.org/licenses/>.
21 : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
22 : #include "core/ActionAtomistic.h"
23 : #include "core/ActionPilot.h"
24 : #include "core/ActionRegister.h"
25 : #include "core/ActionWithValue.h"
26 : #include "tools/Vector.h"
27 : #include "tools/Matrix.h"
28 : #include "tools/AtomNumber.h"
29 : #include "tools/Tools.h"
30 : #include "tools/RMSD.h"
31 : #include "core/PlumedMain.h"
32 : #include "core/ActionSet.h"
33 : #include "core/GenericMolInfo.h"
34 : #include "core/PbcAction.h"
35 : #include "tools/PDB.h"
36 : #include "tools/Pbc.h"
37 :
38 : #include <vector>
39 : #include <string>
40 : #include <memory>
41 :
42 : namespace PLMD {
43 : namespace generic {
44 :
45 : //+PLUMEDOC GENERIC FIT_TO_TEMPLATE
46 : /*
47 : This action is used to align a molecule to a template.
48 :
49 : This can be used to move the coordinates stored in plumed
50 : so as to be aligned with a provided template in PDB format. Pdb should contain
51 : also weights for alignment (see the format of PDB files used e.g. for \ref RMSD).
52 : Make sure your PDB file is correctly formatted as explained \ref pdbreader "in this page".
53 : Weights for displacement are ignored, since no displacement is computed here.
54 : Notice that all atoms (not only those in the template) are aligned.
55 : To see what effect try
56 : the \ref DUMPATOMS directive to output the atomic positions.
57 :
58 : Also notice that PLUMED propagate forces correctly so that you can add a bias on a CV computed
59 : after alignment. For many CVs this has no effect, but in some case the alignment can
60 : change the result. Examples are:
61 : - \ref POSITION CV since it is affected by a rigid shift of the system.
62 : - \ref DISTANCE CV with COMPONENTS. Since the alignment could involve a rotation (with TYPE=OPTIMAL) the actual components could be different
63 : from the original ones.
64 : - \ref CELL components for a similar reason.
65 : - \ref DISTANCE from a \ref FIXEDATOM, provided the fixed atom is introduced _after_ the \ref FIT_TO_TEMPLATE action.
66 :
67 : \attention
68 : The implementation of TYPE=OPTIMAL is available but should be considered in testing phase. Please report any
69 : strange behavior.
70 :
71 : \attention
72 : This directive modifies the stored position at the precise moment
73 : it is executed. This means that only collective variables
74 : which are below it in the input script will see the corrected positions.
75 : As a general rule, put it at the top of the input file. Also, unless you
76 : know exactly what you are doing, leave the default stride (1), so that
77 : this action is performed at every MD step.
78 :
79 : When running with periodic boundary conditions, the atoms should be
80 : in the proper periodic image. This is done automatically since PLUMED 2.5,
81 : by considering the ordered list of atoms and rebuilding the molecules using a procedure
82 : that is equivalent to that done in \ref WHOLEMOLECULES . Notice that
83 : rebuilding is local to this action. This is different from \ref WHOLEMOLECULES
84 : which actually modifies the coordinates stored in PLUMED.
85 :
86 : In case you want to recover the old behavior you should use the NOPBC flag.
87 : In that case you need to take care that atoms are in the correct
88 : periodic image.
89 :
90 : \par Examples
91 :
92 : Align the atomic position to a template then print them.
93 : The following example is only translating the system so as
94 : to align the center of mass of a molecule to the one in the reference
95 : structure `ref.pdb`:
96 : \plumedfile
97 : # dump coordinates before fitting, to see the difference:
98 : DUMPATOMS FILE=dump-before.xyz ATOMS=1-20
99 :
100 : # fit coordinates to ref.pdb template
101 : # this is a "TYPE=SIMPLE" fit, so that only translations are used.
102 : FIT_TO_TEMPLATE STRIDE=1 REFERENCE=ref.pdb TYPE=SIMPLE
103 :
104 : # dump coordinates after fitting, to see the difference:
105 : DUMPATOMS FILE=dump-after.xyz ATOMS=1-20
106 : \endplumedfile
107 :
108 : The following example instead performs a rototranslational fit.
109 : \plumedfile
110 : # dump coordinates before fitting, to see the difference:
111 : DUMPATOMS FILE=dump-before.xyz ATOMS=1-20
112 :
113 : # fit coordinates to ref.pdb template
114 : # this is a "TYPE=OPTIMAL" fit, so that rototranslations are used.
115 : FIT_TO_TEMPLATE STRIDE=1 REFERENCE=ref.pdb TYPE=OPTIMAL
116 :
117 : # dump coordinates after fitting, to see the difference:
118 : DUMPATOMS FILE=dump-after.xyz ATOMS=1-20
119 : \endplumedfile
120 :
121 : In both these cases the reference structure should be provided in a reference pdb file such as the one below:
122 :
123 : \auxfile{ref.pdb}
124 : ATOM 8 HT3 ALA 2 -1.480 -1.560 1.212 1.00 1.00 DIA H
125 : ATOM 9 CAY ALA 2 -0.096 2.144 -0.669 1.00 1.00 DIA C
126 : ATOM 10 HY1 ALA 2 0.871 2.385 -0.588 1.00 1.00 DIA H
127 : ATOM 12 HY3 ALA 2 -0.520 2.679 -1.400 1.00 1.00 DIA H
128 : ATOM 14 OY ALA 2 -1.139 0.931 -0.973 1.00 1.00 DIA O
129 : END
130 : \endauxfile
131 :
132 : In the following example you see two completely equivalent way
133 : to restrain an atom close to a position that is defined in the reference
134 : frame of an aligned molecule. You could for instance use this command to calculate the
135 : position of the center of mass of a ligand after having aligned the atoms to the reference
136 : frame of the protein that is determined by aligning the atoms in the protein to the coordinates
137 : provided in the file ref.pdb
138 : \plumedfile
139 : # center of the ligand:
140 : center: CENTER ATOMS=100-110
141 :
142 : FIT_TO_TEMPLATE REFERENCE=ref.pdb TYPE=OPTIMAL
143 :
144 : # place a fixed atom in the protein reference coordinates:
145 : fix: FIXEDATOM AT=1.0,1.1,1.0
146 :
147 : # take the distance between the fixed atom and the center of the ligand
148 : d: DISTANCE ATOMS=center,fix
149 :
150 : # apply a restraint
151 : RESTRAINT ARG=d AT=0.0 KAPPA=100.0
152 : \endplumedfile
153 :
154 : Notice that you could have obtained an (almost) identical result by adding a fictitious
155 : atom to `ref.pdb` with the serial number corresponding to the atom labelled `center` (there is no automatic way
156 : to get it, but in this example it should be the number of atoms of the system plus one),
157 : and properly setting the weights for alignment and displacement in \ref RMSD.
158 : There are two differences to be expected:
159 : (ab) \ref FIT_TO_TEMPLATE might be slower since it has to rototranslate all the available atoms and
160 : (b) variables employing periodic boundary conditions (such as \ref DISTANCE without `NOPBC`, as in the example above)
161 : are allowed after \ref FIT_TO_TEMPLATE, whereas \ref RMSD expects the issues related to the periodic boundary conditions to be already solved.
162 : The latter means that before the \ref RMSD statement one should use \ref WRAPAROUND or \ref WHOLEMOLECULES to properly place
163 : the ligand.
164 :
165 :
166 : */
167 : //+ENDPLUMEDOC
168 :
169 :
170 : class FitToTemplate:
171 : public ActionPilot,
172 : public ActionAtomistic,
173 : public ActionWithValue
174 : {
175 : std::string type;
176 : bool nopbc;
177 : std::vector<double> weights;
178 : std::vector<std::pair<std::size_t,std::size_t> > p_aligned;
179 : Vector center;
180 : Vector shift;
181 : // optimal alignment related stuff
182 : std::unique_ptr<PLMD::RMSD> rmsd;
183 : Tensor rotation;
184 : Matrix< std::vector<Vector> > drotdpos;
185 : // not used anymore (see notes below at doNotRetrieve())
186 : // std::vector<Vector> positions;
187 : std::vector<Vector> DDistDRef;
188 : std::vector<Vector> ddistdpos;
189 : std::vector<Vector> centeredpositions;
190 : Vector center_positions;
191 : // Copy of the box value
192 : Value* boxValue;
193 : PbcAction* pbc_action;
194 : public:
195 : explicit FitToTemplate(const ActionOptions&ao);
196 : static void registerKeywords( Keywords& keys );
197 55 : bool actionHasForces() override { return true; }
198 : void calculate() override;
199 : void apply() override;
200 0 : unsigned getNumberOfDerivatives() override {plumed_merror("You should not call this function");};
201 : };
202 :
203 : PLUMED_REGISTER_ACTION(FitToTemplate,"FIT_TO_TEMPLATE")
204 :
205 11 : void FitToTemplate::registerKeywords( Keywords& keys ) {
206 11 : Action::registerKeywords( keys );
207 11 : ActionAtomistic::registerKeywords( keys );
208 22 : keys.add("compulsory","STRIDE","1","the frequency with which molecules are reassembled. Unless you are completely certain about what you are doing leave this set equal to 1!");
209 22 : keys.add("compulsory","REFERENCE","a file in pdb format containing the reference structure and the atoms involved in the CV.");
210 22 : keys.add("compulsory","TYPE","SIMPLE","the manner in which RMSD alignment is performed. Should be OPTIMAL or SIMPLE.");
211 22 : keys.addFlag("NOPBC",false,"ignore the periodic boundary conditions when calculating distances");
212 22 : keys.setValueDescription("scalar","the RMSD distance");
213 11 : }
214 :
215 9 : FitToTemplate::FitToTemplate(const ActionOptions&ao):
216 : Action(ao),
217 : ActionPilot(ao),
218 : ActionAtomistic(ao),
219 : ActionWithValue(ao),
220 18 : nopbc(false)
221 : {
222 : std::string reference;
223 9 : parse("REFERENCE",reference);
224 9 : type.assign("SIMPLE");
225 9 : parse("TYPE",type);
226 :
227 9 : parseFlag("NOPBC",nopbc);
228 : // if(type!="SIMPLE") error("Only TYPE=SIMPLE is implemented in FIT_TO_TEMPLATE");
229 :
230 9 : checkRead();
231 :
232 9 : PDB pdb;
233 :
234 : // read everything in ang and transform to nm if we are not in natural units
235 9 : if( !pdb.read(reference,usingNaturalUnits(),0.1/getUnits().getLength()) )
236 0 : error("missing input file " + reference );
237 :
238 9 : requestAtoms(pdb.getAtomNumbers());
239 9 : log.printf(" found %zu atoms in input \n",pdb.getAtomNumbers().size());
240 9 : log.printf(" with indices : ");
241 42 : for(unsigned i=0; i<pdb.getAtomNumbers().size(); ++i) {
242 33 : if(i%25==0) log<<"\n";
243 33 : log.printf("%d ",pdb.getAtomNumbers()[i].serial());
244 : }
245 9 : log.printf("\n");
246 :
247 9 : std::vector<Vector> positions=pdb.getPositions();
248 9 : weights=pdb.getOccupancy();
249 9 : std::vector<AtomNumber> aligned=pdb.getAtomNumbers(); p_aligned.resize( aligned.size() );
250 42 : for(unsigned i=0; i<aligned.size(); ++i) p_aligned[i] = getValueIndices( aligned[i] );
251 :
252 :
253 : // normalize weights
254 42 : double n=0.0; for(unsigned i=0; i<weights.size(); ++i) n+=weights[i];
255 9 : if(n==0.0) {
256 0 : error("PDB file " + reference + " has zero weights. Please check the occupancy column.");
257 : }
258 9 : n=1.0/n;
259 42 : for(unsigned i=0; i<weights.size(); ++i) weights[i]*=n;
260 :
261 : // normalize weights for rmsd calculation
262 9 : std::vector<double> weights_measure=pdb.getBeta();
263 42 : n=0.0; for(unsigned i=0; i<weights_measure.size(); ++i) n+=weights_measure[i]; n=1.0/n;
264 42 : for(unsigned i=0; i<weights_measure.size(); ++i) weights_measure[i]*=n;
265 :
266 : // subtract the center
267 42 : for(unsigned i=0; i<weights.size(); ++i) center+=positions[i]*weights[i];
268 42 : for(unsigned i=0; i<weights.size(); ++i) positions[i]-=center;
269 :
270 13 : if(type=="OPTIMAL" or type=="OPTIMAL-FAST" ) {
271 5 : rmsd=Tools::make_unique<RMSD>();
272 5 : rmsd->set(weights,weights_measure,positions,type,false,false);// note: the reference is shifted now with center in the origin
273 10 : log<<" Method chosen for fitting: "<<rmsd->getMethod()<<" \n";
274 : }
275 9 : if(nopbc) {
276 1 : log<<" Ignoring PBCs when doing alignment, make sure your molecule is whole!<n";
277 : }
278 : // register the value of rmsd (might be useful sometimes)
279 18 : addValue(); setNotPeriodic();
280 :
281 : // I remove this optimization now in order to use makeWhole()
282 : // Notice that for FIT_TO_TEMPLATE TYPE=OPTIMAL a copy was made anyway
283 : // (due to the need to store position to propagate forces on rotational matrix later)
284 : // For FIT_TO_TEMPLATE TYPE=SIMPLE in principle we could use it and write an ad hoc
285 : // version of makeWhole that only computes the center. Too lazy to do it now.
286 : // In case we do it later, remember that uncommenting this line means that
287 : // getPositions will not work anymore! GB
288 : // doNotRetrieve();
289 :
290 : // this is required so as to allow modifyGlobalForce() to return correct
291 : // also for forces that are not owned (and thus not zeored) by all processors.
292 9 : pbc_action=plumed.getActionSet().selectWithLabel<PbcAction*>("Box");
293 9 : if( !pbc_action ) error("cannot align box has not been set");
294 9 : boxValue=pbc_action->copyOutput(0);
295 18 : }
296 :
297 :
298 108 : void FitToTemplate::calculate() {
299 :
300 108 : if(!nopbc) makeWhole();
301 :
302 108 : if (type=="SIMPLE") {
303 48 : Vector cc;
304 :
305 144 : for(unsigned i=0; i<p_aligned.size(); ++i) {
306 96 : cc+=weights[i]*getPosition(i);
307 : }
308 :
309 48 : shift=center-cc;
310 48 : setValue(shift.modulo());
311 48 : unsigned nat = getTotAtoms();
312 6384 : for(unsigned i=0; i<nat; i++) {
313 6336 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
314 6336 : Vector ato=getGlobalPosition(a);
315 6336 : setGlobalPosition(a,ato+shift);
316 : }
317 : }
318 60 : else if( type=="OPTIMAL" or type=="OPTIMAL-FAST") {
319 : // specific stuff that provides all that is needed
320 60 : double r=rmsd->calc_FitElements( getPositions(), rotation, drotdpos, centeredpositions, center_positions);
321 60 : setValue(r); unsigned nat = getTotAtoms();
322 8004 : for(unsigned i=0; i<nat; i++) {
323 7944 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
324 7944 : Vector ato=getGlobalPosition(a);
325 7944 : setGlobalPosition(a,matmul(rotation,ato-center_positions)+center);
326 : }
327 : // rotate box
328 60 : Pbc& pbc(pbc_action->getPbc());
329 60 : pbc.setBox(matmul(pbc_action->getPbc().getBox(),transpose(rotation)));
330 : }
331 108 : }
332 :
333 108 : void FitToTemplate::apply() {
334 108 : auto nat=getTotAtoms();
335 108 : if (type=="SIMPLE") {
336 48 : Vector totForce;
337 6384 : for(unsigned i=0; i<nat; i++) {
338 6336 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
339 6336 : totForce+=getForce(a);
340 : }
341 48 : Tensor vv=Tensor(center,totForce);
342 624 : for(unsigned i=0; i<3; ++i) for(unsigned j=0; j<3; ++j) boxValue->addForce( 3*i+j, vv(i,j) );
343 144 : for(unsigned i=0; i<p_aligned.size(); ++i) { addForce( p_aligned[i], -totForce*weights[i]); }
344 60 : } else if ( type=="OPTIMAL" or type=="OPTIMAL-FAST") {
345 60 : Vector totForce;
346 8004 : for(unsigned i=0; i<nat; i++) {
347 7944 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
348 7944 : Vector f=getForce(a);
349 : // rotate back forces
350 7944 : Vector nf=matmul(transpose(rotation),f);
351 7944 : addForce(a, nf-f);
352 : // accumulate rotated c.o.m. forces - this is already in the non rotated reference frame
353 7944 : totForce+=nf;
354 : }
355 60 : Tensor virial;
356 780 : for(unsigned i=0; i<3; ++i) for(unsigned j=0; j<3; ++j) virial[i][j] = boxValue->getForce( 3*i+j );
357 : // notice that an extra Tensor(center,matmul(rotation,totForce)) is required to
358 : // compute the derivatives of the rotation with respect to center
359 60 : Tensor ww=matmul(transpose(rotation),virial+Tensor(center,matmul(rotation,totForce)));
360 : // rotate back virial
361 60 : virial=matmul(transpose(rotation),matmul(virial,rotation));
362 :
363 : // now we compute the force due to alignment
364 360 : for(unsigned i=0; i<p_aligned.size(); i++) {
365 300 : Vector g;
366 1200 : for(unsigned k=0; k<3; k++) {
367 : // this could be made faster computing only the diagonal of d
368 900 : Tensor d=matmul(ww,RMSD::getMatrixFromDRot(drotdpos,i,k));
369 900 : g[k]=(d(0,0)+d(1,1)+d(2,2));
370 : }
371 : // here is the extra contribution
372 300 : addForce( p_aligned[i], -g-weights[i]*totForce );
373 : // here it the contribution to the virial
374 : // notice that here we can use absolute positions since, for the alignment to be defined,
375 : // positions should be in one well defined periodic image
376 300 : virial+=extProduct(getPosition(i),g);
377 : }
378 : // finally, correction to the virial
379 120 : boxValue->clearInputForce(); virial+=extProduct(matmul(transpose(rotation),center),totForce);
380 780 : for(unsigned i=0; i<3; ++i) for(unsigned j=0; j<3; ++j) boxValue->addForce( 3*i+j, virial(i,j) );
381 : }
382 108 : }
383 :
384 : }
385 : }
|
