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/Atoms.h"
32 : #include "core/PlumedMain.h"
33 : #include "core/ActionSet.h"
34 : #include "core/GenericMolInfo.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<AtomNumber> 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 :
192 :
193 : public:
194 : explicit FitToTemplate(const ActionOptions&ao);
195 : static void registerKeywords( Keywords& keys );
196 : void calculate() override;
197 : void apply() override;
198 0 : unsigned getNumberOfDerivatives() override {plumed_merror("You should not call this function");};
199 : };
200 :
201 10437 : PLUMED_REGISTER_ACTION(FitToTemplate,"FIT_TO_TEMPLATE")
202 :
203 10 : void FitToTemplate::registerKeywords( Keywords& keys ) {
204 10 : Action::registerKeywords( keys );
205 10 : ActionAtomistic::registerKeywords( keys );
206 20 : 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!");
207 20 : keys.add("compulsory","REFERENCE","a file in pdb format containing the reference structure and the atoms involved in the CV.");
208 20 : keys.add("compulsory","TYPE","SIMPLE","the manner in which RMSD alignment is performed. Should be OPTIMAL or SIMPLE.");
209 20 : keys.addFlag("NOPBC",false,"ignore the periodic boundary conditions when calculating distances");
210 10 : }
211 :
212 9 : FitToTemplate::FitToTemplate(const ActionOptions&ao):
213 : Action(ao),
214 : ActionPilot(ao),
215 : ActionAtomistic(ao),
216 : ActionWithValue(ao),
217 18 : nopbc(false)
218 : {
219 : std::string reference;
220 9 : parse("REFERENCE",reference);
221 9 : type.assign("SIMPLE");
222 9 : parse("TYPE",type);
223 :
224 9 : parseFlag("NOPBC",nopbc);
225 : // if(type!="SIMPLE") error("Only TYPE=SIMPLE is implemented in FIT_TO_TEMPLATE");
226 :
227 9 : checkRead();
228 :
229 9 : PDB pdb;
230 :
231 : // read everything in ang and transform to nm if we are not in natural units
232 18 : if( !pdb.read(reference,plumed.getAtoms().usingNaturalUnits(),0.1/atoms.getUnits().getLength()) )
233 0 : error("missing input file " + reference );
234 :
235 9 : requestAtoms(pdb.getAtomNumbers());
236 9 : log.printf(" found %zu atoms in input \n",pdb.getAtomNumbers().size());
237 9 : log.printf(" with indices : ");
238 42 : for(unsigned i=0; i<pdb.getAtomNumbers().size(); ++i) {
239 33 : if(i%25==0) log<<"\n";
240 33 : log.printf("%d ",pdb.getAtomNumbers()[i].serial());
241 : }
242 9 : log.printf("\n");
243 :
244 9 : std::vector<Vector> positions=pdb.getPositions();
245 9 : weights=pdb.getOccupancy();
246 9 : aligned=pdb.getAtomNumbers();
247 :
248 :
249 : // normalize weights
250 42 : double n=0.0; for(unsigned i=0; i<weights.size(); ++i) n+=weights[i];
251 9 : if(n==0.0) {
252 0 : error("PDB file " + reference + " has zero weights. Please check the occupancy column.");
253 : }
254 9 : n=1.0/n;
255 42 : for(unsigned i=0; i<weights.size(); ++i) weights[i]*=n;
256 :
257 : // normalize weights for rmsd calculation
258 9 : std::vector<double> weights_measure=pdb.getBeta();
259 42 : n=0.0; for(unsigned i=0; i<weights_measure.size(); ++i) n+=weights_measure[i]; n=1.0/n;
260 42 : for(unsigned i=0; i<weights_measure.size(); ++i) weights_measure[i]*=n;
261 :
262 : // subtract the center
263 42 : for(unsigned i=0; i<weights.size(); ++i) center+=positions[i]*weights[i];
264 42 : for(unsigned i=0; i<weights.size(); ++i) positions[i]-=center;
265 :
266 13 : if(type=="OPTIMAL" or type=="OPTIMAL-FAST" ) {
267 10 : rmsd=Tools::make_unique<RMSD>();
268 5 : rmsd->set(weights,weights_measure,positions,type,false,false);// note: the reference is shifted now with center in the origin
269 10 : log<<" Method chosen for fitting: "<<rmsd->getMethod()<<" \n";
270 : }
271 9 : if(nopbc) {
272 1 : log<<" Ignoring PBCs when doing alignment, make sure your molecule is whole!<n";
273 : }
274 : // register the value of rmsd (might be useful sometimes)
275 9 : addValue(); setNotPeriodic();
276 :
277 : // I remove this optimization now in order to use makeWhole()
278 : // Notice that for FIT_TO_TEMPLATE TYPE=OPTIMAL a copy was made anyway
279 : // (due to the need to store position to propagate forces on rotational matrix later)
280 : // For FIT_TO_TEMPLATE TYPE=SIMPLE in principle we could use it and write an ad hoc
281 : // version of makeWhole that only computes the center. Too lazy to do it now.
282 : // In case we do it later, remember that uncommenting this line means that
283 : // getPositions will not work anymore! GB
284 : // doNotRetrieve();
285 :
286 : // this is required so as to allow modifyGlobalForce() to return correct
287 : // also for forces that are not owned (and thus not zeored) by all processors.
288 : allowToAccessGlobalForces();
289 18 : }
290 :
291 :
292 108 : void FitToTemplate::calculate() {
293 :
294 108 : if(!nopbc) makeWhole();
295 :
296 108 : if (type=="SIMPLE") {
297 48 : Vector cc;
298 :
299 144 : for(unsigned i=0; i<aligned.size(); ++i) {
300 96 : cc+=weights[i]*getPosition(i);
301 : }
302 :
303 48 : shift=center-cc;
304 48 : setValue(shift.modulo());
305 6420 : for(unsigned i=0; i<getTotAtoms(); i++) {
306 : Vector & ato (modifyGlobalPosition(AtomNumber::index(i)));
307 6372 : ato+=shift;
308 : }
309 : }
310 60 : else if( type=="OPTIMAL" or type=="OPTIMAL-FAST") {
311 : // specific stuff that provides all that is needed
312 60 : double r=rmsd->calc_FitElements( getPositions(), rotation, drotdpos, centeredpositions, center_positions);
313 60 : setValue(r);
314 8004 : for(unsigned i=0; i<getTotAtoms(); i++) {
315 : Vector & ato (modifyGlobalPosition(AtomNumber::index(i)));
316 7944 : ato=matmul(rotation,ato-center_positions)+center;
317 : }
318 : // rotate box
319 : Pbc & pbc(modifyGlobalPbc());
320 60 : pbc.setBox(matmul(pbc.getBox(),transpose(rotation)));
321 : }
322 :
323 108 : }
324 :
325 108 : void FitToTemplate::apply() {
326 108 : if (type=="SIMPLE") {
327 48 : Vector totForce;
328 6420 : for(unsigned i=0; i<getTotAtoms(); i++) {
329 6372 : totForce+=modifyGlobalForce(AtomNumber::index(i));
330 : }
331 : Tensor & vv(modifyGlobalVirial());
332 48 : vv+=Tensor(center,totForce);
333 144 : for(unsigned i=0; i<aligned.size(); ++i) {
334 : Vector & ff(modifyGlobalForce(aligned[i]));
335 96 : ff-=totForce*weights[i];
336 : }
337 60 : } else if ( type=="OPTIMAL" or type=="OPTIMAL-FAST") {
338 60 : Vector totForce;
339 8004 : for(unsigned i=0; i<getTotAtoms(); i++) {
340 : Vector & f(modifyGlobalForce(AtomNumber::index(i)));
341 : // rotate back forces
342 7944 : f=matmul(transpose(rotation),f);
343 : // accumulate rotated c.o.m. forces - this is already in the non rotated reference frame
344 7944 : totForce+=f;
345 : }
346 : Tensor& virial(modifyGlobalVirial());
347 : // notice that an extra Tensor(center,matmul(rotation,totForce)) is required to
348 : // compute the derivatives of the rotation with respect to center
349 60 : Tensor ww=matmul(transpose(rotation),virial+Tensor(center,matmul(rotation,totForce)));
350 : // rotate back virial
351 60 : virial=matmul(transpose(rotation),matmul(virial,rotation));
352 :
353 : // now we compute the force due to alignment
354 360 : for(unsigned i=0; i<aligned.size(); i++) {
355 300 : Vector g;
356 1200 : for(unsigned k=0; k<3; k++) {
357 : // this could be made faster computing only the diagonal of d
358 900 : Tensor d=matmul(ww,RMSD::getMatrixFromDRot(drotdpos,i,k));
359 900 : g[k]=(d(0,0)+d(1,1)+d(2,2));
360 : }
361 : // here is the extra contribution
362 300 : modifyGlobalForce(aligned[i])+=-g-weights[i]*totForce;
363 : // here it the contribution to the virial
364 : // notice that here we can use absolute positions since, for the alignment to be defined,
365 : // positions should be in one well defined periodic image
366 300 : virial+=extProduct(getPosition(i),g);
367 : }
368 : // finally, correction to the virial
369 60 : virial+=extProduct(matmul(transpose(rotation),center),totForce);
370 : }
371 108 : }
372 :
373 : }
374 : }
|
