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 [RMSD](RMSD.md)).
52 : Make sure your PDB file is correctly formatted as explained in the documentation for [PDB2CONSTANT](PDB2CONSTANT.md).
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 [DUMPATOMS](DUMPATOMS.md) 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 : - [POSITION](POSITION.md) CV since it is affected by a rigid shift of the system.
62 : - [DISTANCE](DISTANCE.md) 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 : - [CELL](CELL.md) components for a similar reason.
65 : - [DISTANCE](DISTANCE.md) from a [FIXEDATOM](FIXEDATOM.md), provided the fixed atom is introduced _after_ the FIT_TO_TEMPLATE action.
66 :
67 : > [!CAUTION]
68 : > The implementation of TYPE=OPTIMAL is available but should be considered in testing phase. Please report any
69 : > strange behavior.
70 :
71 : > [!CAUTION]
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 [WHOLEMOLECULES](WHOLEMOLECULES.md) . Notice that
83 : rebuilding is local to this action. This is different from [WHOLEMOLECULES](WHOLEMOLECULES.md)
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 : ## 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 :
97 : ```plumed
98 : #SETTINGS INPUTFILES=regtest/basic/rt63/align.pdb
99 : # dump coordinates before fitting, to see the difference:
100 : DUMPATOMS FILE=dump-before.xyz ATOMS=1-20
101 :
102 : # fit coordinates to ref.pdb template
103 : # this is a "TYPE=SIMPLE" fit, so that only translations are used.
104 : FIT_TO_TEMPLATE STRIDE=1 REFERENCE=regtest/basic/rt63/align.pdb TYPE=SIMPLE
105 :
106 : # dump coordinates after fitting, to see the difference:
107 : DUMPATOMS FILE=dump-after.xyz ATOMS=1-20
108 : ```
109 :
110 : The following example instead performs a rototranslational fit.
111 :
112 : ```plumed
113 : #SETTINGS INPUTFILES=regtest/basic/rt63/align.pdb
114 : # dump coordinates before fitting, to see the difference:
115 : DUMPATOMS FILE=dump-before.xyz ATOMS=1-20
116 :
117 : # fit coordinates to ref.pdb template
118 : # this is a "TYPE=OPTIMAL" fit, so that rototranslations are used.
119 : FIT_TO_TEMPLATE STRIDE=1 REFERENCE=regtest/basic/rt63/align.pdb TYPE=OPTIMAL
120 :
121 : # dump coordinates after fitting, to see the difference:
122 : DUMPATOMS FILE=dump-after.xyz ATOMS=1-20
123 : ```
124 :
125 : In the following example you see two completely equivalent way
126 : to restrain an atom close to a position that is defined in the reference
127 : frame of an aligned molecule. You could for instance use this command to calculate the
128 : position of the center of mass of a ligand after having aligned the atoms to the reference
129 : frame of the protein that is determined by aligning the atoms in the protein to the coordinates
130 : provided in the file ref.pdb
131 :
132 : ```plumed
133 : #SETTINGS INPUTFILES=regtest/basic/rt63/align.pdb
134 : # center of the ligand:
135 : center: CENTER ATOMS=100-110
136 :
137 : FIT_TO_TEMPLATE REFERENCE=regtest/basic/rt63/align.pdb TYPE=OPTIMAL
138 :
139 : # place a fixed atom in the protein reference coordinates:
140 : fix: FIXEDATOM AT=1.0,1.1,1.0
141 :
142 : # take the distance between the fixed atom and the center of the ligand
143 : d: DISTANCE ATOMS=center,fix
144 :
145 : # apply a restraint
146 : RESTRAINT ARG=d AT=0.0 KAPPA=100.0
147 : ```
148 :
149 : Notice that you could have obtained an (almost) identical result by adding a fictitious
150 : atom to `ref.pdb` with the serial number corresponding to the atom labelled `center` (there is no automatic way
151 : to get it, but in this example it should be the number of atoms of the system plus one),
152 : and properly setting the weights for alignment and displacement in [RMSD](RMSD.md).
153 : There are two differences to be expected:
154 : (ab) FIT_TO_TEMPLATE might be slower since it has to rototranslate all the available atoms and
155 : (b) variables employing periodic boundary conditions (such as [DISTANCE](DISTANCE.md) without `NOPBC`, as in the example above)
156 : are allowed after FIT_TO_TEMPLATE, whereas [RMSD](RMSD.md) expects the issues related to the periodic boundary conditions to be already solved.
157 : The latter means that before the [RMSD](RMSD.md) statement one should use [WRAPAROUND](WRAPAROUND.md) or [WHOLEMOLECULES](WHOLEMOLECULES.md) to properly place
158 : the ligand.
159 :
160 :
161 : */
162 : //+ENDPLUMEDOC
163 :
164 :
165 : class FitToTemplate:
166 : public ActionPilot,
167 : public ActionAtomistic,
168 : public ActionWithValue {
169 : std::string type;
170 : bool nopbc;
171 : std::vector<double> weights;
172 : std::vector<std::pair<std::size_t,std::size_t> > p_aligned;
173 : Vector center;
174 : Vector shift;
175 : // optimal alignment related stuff
176 : std::unique_ptr<PLMD::RMSD> rmsd;
177 : Tensor rotation;
178 : Matrix< std::vector<Vector> > drotdpos;
179 : // not used anymore (see notes below at doNotRetrieve())
180 : // std::vector<Vector> positions;
181 : std::vector<Vector> DDistDRef;
182 : std::vector<Vector> ddistdpos;
183 : std::vector<Vector> centeredpositions;
184 : Vector center_positions;
185 : // Copy of the box value
186 : Value* boxValue;
187 : PbcAction* pbc_action;
188 : public:
189 : explicit FitToTemplate(const ActionOptions&ao);
190 : static void registerKeywords( Keywords& keys );
191 55 : bool actionHasForces() override {
192 55 : return true;
193 : }
194 : void calculate() override;
195 : void apply() override;
196 0 : unsigned getNumberOfDerivatives() override {
197 0 : plumed_merror("You should not call this function");
198 : };
199 : };
200 :
201 : PLUMED_REGISTER_ACTION(FitToTemplate,"FIT_TO_TEMPLATE")
202 :
203 11 : void FitToTemplate::registerKeywords( Keywords& keys ) {
204 11 : Action::registerKeywords( keys );
205 11 : ActionAtomistic::registerKeywords( keys );
206 11 : 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 11 : keys.add("compulsory","REFERENCE","a file in pdb format containing the reference structure and the atoms involved in the CV.");
208 11 : keys.add("compulsory","TYPE","SIMPLE","the manner in which RMSD alignment is performed. Should be OPTIMAL or SIMPLE.");
209 11 : keys.addFlag("NOPBC",false,"ignore the periodic boundary conditions when calculating distances");
210 22 : keys.setValueDescription("scalar","the RMSD distance");
211 11 : }
212 :
213 9 : FitToTemplate::FitToTemplate(const ActionOptions&ao):
214 : Action(ao),
215 : ActionPilot(ao),
216 : ActionAtomistic(ao),
217 : ActionWithValue(ao),
218 18 : nopbc(false) {
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 9 : if( !pdb.read(reference,usingNaturalUnits(),0.1/getUnits().getLength()) ) {
233 0 : error("missing input file " + reference );
234 : }
235 :
236 9 : requestAtoms(pdb.getAtomNumbers());
237 9 : log.printf(" found %zu atoms in input \n",pdb.getAtomNumbers().size());
238 9 : log.printf(" with indices : ");
239 42 : for(unsigned i=0; i<pdb.getAtomNumbers().size(); ++i) {
240 33 : if(i%25==0) {
241 9 : log<<"\n";
242 : }
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();
250 9 : p_aligned.resize( aligned.size() );
251 42 : for(unsigned i=0; i<aligned.size(); ++i) {
252 33 : p_aligned[i] = getValueIndices( aligned[i] );
253 : }
254 :
255 :
256 : // normalize weights
257 : double n=0.0;
258 42 : for(unsigned i=0; i<weights.size(); ++i) {
259 33 : n+=weights[i];
260 : }
261 9 : if(n==0.0) {
262 0 : error("PDB file " + reference + " has zero weights. Please check the occupancy column.");
263 : }
264 9 : n=1.0/n;
265 42 : for(unsigned i=0; i<weights.size(); ++i) {
266 33 : weights[i]*=n;
267 : }
268 :
269 : // normalize weights for rmsd calculation
270 9 : std::vector<double> weights_measure=pdb.getBeta();
271 : n=0.0;
272 42 : for(unsigned i=0; i<weights_measure.size(); ++i) {
273 33 : n+=weights_measure[i];
274 : }
275 9 : n=1.0/n;
276 42 : for(unsigned i=0; i<weights_measure.size(); ++i) {
277 33 : weights_measure[i]*=n;
278 : }
279 :
280 : // subtract the center
281 42 : for(unsigned i=0; i<weights.size(); ++i) {
282 33 : center+=positions[i]*weights[i];
283 : }
284 42 : for(unsigned i=0; i<weights.size(); ++i) {
285 33 : positions[i]-=center;
286 : }
287 :
288 13 : if(type=="OPTIMAL" or type=="OPTIMAL-FAST" ) {
289 5 : rmsd=Tools::make_unique<RMSD>();
290 5 : rmsd->set(weights,weights_measure,positions,type,false,false);// note: the reference is shifted now with center in the origin
291 10 : log<<" Method chosen for fitting: "<<rmsd->getMethod()<<" \n";
292 : }
293 9 : if(nopbc) {
294 1 : log<<" Ignoring PBCs when doing alignment, make sure your molecule is whole!<n";
295 : }
296 : // register the value of rmsd (might be useful sometimes)
297 9 : addValue();
298 9 : setNotPeriodic();
299 :
300 : // I remove this optimization now in order to use makeWhole()
301 : // Notice that for FIT_TO_TEMPLATE TYPE=OPTIMAL a copy was made anyway
302 : // (due to the need to store position to propagate forces on rotational matrix later)
303 : // For FIT_TO_TEMPLATE TYPE=SIMPLE in principle we could use it and write an ad hoc
304 : // version of makeWhole that only computes the center. Too lazy to do it now.
305 : // In case we do it later, remember that uncommenting this line means that
306 : // getPositions will not work anymore! GB
307 : // doNotRetrieve();
308 :
309 : // this is required so as to allow modifyGlobalForce() to return correct
310 : // also for forces that are not owned (and thus not zeored) by all processors.
311 9 : pbc_action=plumed.getActionSet().selectWithLabel<PbcAction*>("Box");
312 9 : if( !pbc_action ) {
313 0 : error("cannot align box has not been set");
314 : }
315 9 : boxValue=pbc_action->copyOutput(0);
316 18 : }
317 :
318 :
319 108 : void FitToTemplate::calculate() {
320 :
321 108 : if(!nopbc) {
322 96 : makeWhole();
323 : }
324 :
325 108 : if (type=="SIMPLE") {
326 48 : Vector cc;
327 :
328 144 : for(unsigned i=0; i<p_aligned.size(); ++i) {
329 96 : cc+=weights[i]*getPosition(i);
330 : }
331 :
332 48 : shift=center-cc;
333 48 : setValue(shift.modulo());
334 48 : unsigned nat = getTotAtoms();
335 6384 : for(unsigned i=0; i<nat; i++) {
336 6336 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
337 6336 : Vector ato=getGlobalPosition(a);
338 6336 : setGlobalPosition(a,ato+shift);
339 : }
340 60 : } else if( type=="OPTIMAL" or type=="OPTIMAL-FAST") {
341 : // specific stuff that provides all that is needed
342 60 : double r=rmsd->calc_FitElements( getPositions(), rotation, drotdpos, centeredpositions, center_positions);
343 60 : setValue(r);
344 60 : unsigned nat = getTotAtoms();
345 8004 : for(unsigned i=0; i<nat; i++) {
346 7944 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
347 7944 : Vector ato=getGlobalPosition(a);
348 7944 : setGlobalPosition(a,matmul(rotation,ato-center_positions)+center);
349 : }
350 : // rotate box
351 60 : Pbc& pbc(pbc_action->getPbc());
352 60 : pbc.setBox(matmul(pbc_action->getPbc().getBox(),transpose(rotation)));
353 : }
354 108 : }
355 :
356 108 : void FitToTemplate::apply() {
357 108 : auto nat=getTotAtoms();
358 108 : if (type=="SIMPLE") {
359 48 : Vector totForce;
360 6384 : for(unsigned i=0; i<nat; i++) {
361 6336 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
362 6336 : totForce+=getForce(a);
363 : }
364 48 : Tensor vv=Tensor(center,totForce);
365 192 : for(unsigned i=0; i<3; ++i)
366 576 : for(unsigned j=0; j<3; ++j) {
367 432 : boxValue->addForce( 3*i+j, vv(i,j) );
368 : }
369 144 : for(unsigned i=0; i<p_aligned.size(); ++i) {
370 96 : addForce( p_aligned[i], -totForce*weights[i]);
371 : }
372 60 : } else if ( type=="OPTIMAL" or type=="OPTIMAL-FAST") {
373 60 : Vector totForce;
374 8004 : for(unsigned i=0; i<nat; i++) {
375 7944 : std::pair<std::size_t,std::size_t> a = getValueIndices( AtomNumber::index(i));
376 7944 : Vector f=getForce(a);
377 : // rotate back forces
378 7944 : Vector nf=matmul(transpose(rotation),f);
379 7944 : addForce(a, nf-f);
380 : // accumulate rotated c.o.m. forces - this is already in the non rotated reference frame
381 7944 : totForce+=nf;
382 : }
383 60 : Tensor virial;
384 240 : for(unsigned i=0; i<3; ++i)
385 720 : for(unsigned j=0; j<3; ++j) {
386 540 : virial[i][j] = boxValue->getForce( 3*i+j );
387 : }
388 : // notice that an extra Tensor(center,matmul(rotation,totForce)) is required to
389 : // compute the derivatives of the rotation with respect to center
390 60 : Tensor ww=matmul(transpose(rotation),virial+Tensor(center,matmul(rotation,totForce)));
391 : // rotate back virial
392 60 : virial=matmul(transpose(rotation),matmul(virial,rotation));
393 :
394 : // now we compute the force due to alignment
395 360 : for(unsigned i=0; i<p_aligned.size(); i++) {
396 300 : Vector g;
397 1200 : for(unsigned k=0; k<3; k++) {
398 : // this could be made faster computing only the diagonal of d
399 900 : Tensor d=matmul(ww,RMSD::getMatrixFromDRot(drotdpos,i,k));
400 900 : g[k]=(d(0,0)+d(1,1)+d(2,2));
401 : }
402 : // here is the extra contribution
403 300 : addForce( p_aligned[i], -g-weights[i]*totForce );
404 : // here it the contribution to the virial
405 : // notice that here we can use absolute positions since, for the alignment to be defined,
406 : // positions should be in one well defined periodic image
407 300 : virial+=extProduct(getPosition(i),g);
408 : }
409 : // finally, correction to the virial
410 60 : boxValue->clearInputForce();
411 60 : virial+=extProduct(matmul(transpose(rotation),center),totForce);
412 240 : for(unsigned i=0; i<3; ++i)
413 720 : for(unsigned j=0; j<3; ++j) {
414 540 : boxValue->addForce( 3*i+j, virial(i,j) );
415 : }
416 : }
417 108 : }
418 :
419 : }
420 : }
|
