On Friday, July 19, 2013 8:02:01 PM UTC+1, Christoph Junghans wrote: > > 2013/7/19 <[email protected] <javascript:>>: > > Thank you for this. > > I recalculated my distributions and make them a value between 0 and 1 at > > large distances. If I submit then my own potential pot.in which goes to > a > > positive value in this case would votca not shift the potential after > each > > iteration to zero? Please, let me know. > VOTCA always shifts the potential, to U(r_max)=0 for non-bonded > interaction and U_min=0 for bonded interactions. > This is why I said making VOTCA believe your interaction is bonded, might > help. > The update rule is the same for bonded and non-bonded. > > Anyhow the shift doesn't change the physics, so why are you so bother > about it? You could just to do IBI and shift the finial outcome once > it is converged. > > Christoph >
Thank you. This what I will do - shift the final potential. what would be the best cut off for IBI - is that the place where distributions have maxima? Steven > > > > Steven > > > > W dniu środa, 17 lipca 2013 17:43:51 UTC+1 użytkownik Christoph Junghans > > napisał: > >> > >> 2013/7/17 <[email protected]>: > >> > Thank you for this. > >> > > >> > Another question - presumably I will specify my own distributions > (not > >> > from > >> > VOTCA) with some simplifications. Then I will create intial > potentials > >> > to > >> > IBI as name.pot.in. Votca will try to fit to my distributions to the > one > >> > submitted in the main directory. However, by calculating > distributions > >> > after > >> > each step these will not be distributions which could possibly match > my > >> > initial with simpifications, am I right? Do I have to calculate RDFs > at > >> > each > >> > step on my own then? > >> 1.) If you give VOTCA a pot.in, it will not do use the potential of > >> mean as initial guess in step_000, but just your potential. > >> 2.) Independently of pot.in, VOTCA will always compare dist.new > >> against dist.tgt every step > >> 3.) dist.new is calculated in every iteration using csg_stat, you > >> could make VOTCA fit only the part 0 to min. > >> 3b.) You could write a custom post-update script to do whatever rdf > >> calculation/ potential modification after you want. This can be used > >> to overwrite rdf calculated by csg_stat. > >> 3c.) All of this seems a bit hacky, why don't you just declare some > >> interaction of type bonded, so csg_stat will calculate > >> p(r)=H(r)/(4pi*r^2) instead of the rdf for those. > >> > >> > Another issue: How Votca normalize RDFs using csg_stat ? I wish to > have > >> > volume and mass normalized distributions, is that possible? > >> No, we don't have an option to calculate the volume and mass > >> normalized distributions instead. Though it is not hard to implement! > >> For IBI that would also make no difference as all normalization > >> factors drop out in the ratio g(r)/g_target(r) used in the update. > >> > >> Christoph > >> > >> > >> > > >> > Steven > >> > > >> > W dniu wtorek, 16 lipca 2013 23:17:05 UTC+1 użytkownik Christoph > >> > Junghans > >> > napisał: > >> >> > >> >> 2013/7/16 <[email protected]>: > >> >> > > >> >> > > >> >> > W dniu wtorek, 16 lipca 2013 08:48:51 UTC+1 użytkownik Tristan > Bereau > >> >> > napisał: > >> >> >> > >> >> >> That sounds like what I was hinting at: from what I understand, > >> >> >> you're > >> >> >> simulating a single protein, not a pure liquid of stuff. So your > RDF > >> >> >> will never go to 1 because there won't be anything at large > >> >> >> distances. > >> >> >> Is that the case? If so, more iterations and/or better initial > >> >> >> guesses > >> >> >> won't cut it. > >> >> > > >> >> > > >> >> > Yes, this is the case. I will aproximate the RDFs so that they go > to > >> >> > sth > >> >> > which is non zero. Thank you. > >> >> non zero will not be enough, it has to be 1 otherwise your potential > >> >> will still accumulated whatever the value, kT*log(P(r_cut), at the > >> >> cutoff is. > >> >> > >> >> For me the distributions looks more like a something, which could be > >> >> modeled with a non-linear spring type potential (r-> +/-inf P->inf), > >> >> where the minimum is a zero. > >> >> VOTCA could do that for you if declare the interaction as bonded. > >> >> (VOTCA's definition of non-bonded and bonded might not be taken too > >> >> strict.) > >> >> Also from the modeling point of view, it might make sense to have a > >> >> spring between some beads, which cannot go infinitely apart due to > >> >> geometry. > >> >> > >> >> Anyhow, these are scientific decisions you have to make yourself. > >> >> > >> >> Christoph > >> >> > >> >> > > >> >> > Steven > >> >> >> > >> >> >> > >> >> >> On Mon, Jul 15, 2013 at 10:09 PM, <[email protected]> wrote: > >> >> >> > Thank you for this. For heterogenous system RDF does not go to > 1 > >> >> >> > but > >> >> >> > to > >> >> >> > 0. > >> >> >> > In this case I guess I need thousands of iterations... The > system > >> >> >> > input > >> >> >> > are > >> >> >> > 15 potentials which makes it so complicated. > >> >> >> > > >> >> >> > Steven > >> >> >> > > >> >> >> > W dniu poniedziałek, 15 lipca 2013 17:44:46 UTC+1 użytkownik > >> >> >> > Christoph > >> >> >> > Junghans napisał: > >> >> >> >> > >> >> >> >> 2013/7/15 <[email protected]>: > >> >> >> >> > Votca is definitely wrong. If you take the example of > maximum > >> >> >> >> > of > >> >> >> >> > my > >> >> >> >> > ACI-ACI.dist.tgt the maximum corresponds to 65.555. The > >> >> >> >> > potential > >> >> >> >> > at > >> >> >> >> > this > >> >> >> >> > point should be: W = -2.49435*ln(65.55) = -10.433 and in my > >> >> >> >> > ACI-ACI.dist.pot > >> >> >> >> > the value corresponds to -16.1 - it is a huge difference and > >> >> >> >> > that > >> >> >> >> > is > >> >> >> >> > why > >> >> >> >> > my > >> >> >> >> > further distributions are so huge.... > >> >> >> >> No, Votca is 100% correct, and does what it is supposed to do. > >> >> >> >> > >> >> >> >> First, have a look at your ACI-ACI.dist.tgt again, this > >> >> >> >> distribution > >> >> >> >> doesn't go to one hence the potential doesn't go to 0 for > large > >> >> >> >> r. > >> >> >> >> And that is mainly the reason why VOTCA cannot handle it, > >> >> >> >> ACI-ACI.dist.tgt is not a common rdf! > >> >> >> >> You will have to provide an initial guess (pot.in) to make it > >> >> >> >> work. > >> >> >> >> (Please also read my email from July 10th again.) > >> >> >> >> > >> >> >> >> Second, VOTCA does exactly what it is supposed to do. Go into > >> >> >> >> gnuplot > >> >> >> >> and > >> >> >> >> run: > >> >> >> >> p [0:3][-20:5] "ACI-ACI.dist.tgt" u 1:(-2.49435*log($2)-5.7) w > l, > >> >> >> >> "ACI-ACI.pot.cur" w l > >> >> >> >> Except for some small deviations, which come from the cubic > >> >> >> >> spline > >> >> >> >> interpolation, there is no difference in the curves. > >> >> >> >> As Victor said before, VOTCA shifts the potential to be zero > at > >> >> >> >> the > >> >> >> >> cutoff -> -10.433 - 5.7 = -16.1. This shift of 5.7 makes no > >> >> >> >> difference > >> >> >> >> for the thermodynamics however. > >> >> >> >> > >> >> >> >> Third, even pot.new is correct. Run > >> >> >> >> $ paste ACI-ACI.dist.new <(sed '/^#/d' ACI-ACI.dist.tgt) <(sed > >> >> >> >> '/^#/d' > >> >> >> >> ACI-ACI.pot.cur) > ACI-ACI.temp > >> >> >> >> to generate a temp file. > >> >> >> >> And go into gnuplot and plot: > >> >> >> >> p [0:3][-20:5] "ACI-ACI.temp" u 1:(2.49435*log($2/$5)+$8-16.1) > w > >> >> >> >> l, > >> >> >> >> "ACI-ACI.pot.new" w l > >> >> >> >> There is basically no difference in the curves. > >> >> >> >> > >> >> >> >> Conclusion: > >> >> >> >> - check your distributions again > >> >> >> >> - provide pot.in for the interaction, which don't have a > "common" > >> >> >> >> rdf > >> >> >> >> (meaning which doesn't go to 1) > >> >> >> >> > >> >> >> >> Christoph > >> >> >> >> > >> >> >> >> > >> >> >> >> > > >> >> >> >> > > >> >> >> >> > > >> >> >> >> > > >> >> >> >> > W dniu poniedziałek, 15 lipca 2013 12:59:40 UTC+1 użytkownik > >> >> >> >> > [email protected] napisał: > >> >> >> >> >> > >> >> >> >> >> > >> >> >> >> >> > >> >> >> >> >> W dniu poniedziałek, 15 lipca 2013 12:42:37 UTC+1 > użytkownik > >> >> >> >> >> Victor > >> >> >> >> >> Rühle > >> >> >> >> >> napisał: > >> >> >> >> >>> > >> >> >> >> >>> Dear Steven, > >> >> >> >> >>> > >> >> >> >> >>> provided the same kBT was used, I can think of two issues > >> >> >> >> >>> which > >> >> >> >> >>> might > >> >> >> >> >>> lead to these differences > >> >> >> >> >>> > >> >> >> >> >>> 1) votca can shift the potential, but the shape should > match. > >> >> >> >> >>> That > >> >> >> >> >>> can > >> >> >> >> >>> in > >> >> >> >> >>> particular happen if you cut the rdf in a region where > there > >> >> >> >> >>> are > >> >> >> >> >>> still > >> >> >> >> >>> modulations. > >> >> >> >> >>> 2) What type of potential are you lookin at? For bonds and > >> >> >> >> >>> angles, > >> >> >> >> >>> there > >> >> >> >> >>> is indeed a normalization necessary, see > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> >>> http://dx.doi.org/10.1002/(SICI)1521-4044(199802)49:2/3 > <61::AID-APOL61>3.0.CO;2-V > >> >> >> >> >> > >> >> >> >> >> > >> >> >> >> >> Thank you. I am looking at the nonbonded interactions only. > >> >> >> >> >> The > >> >> >> >> >> shape > >> >> >> >> >> of > >> >> >> >> >> the potential matches but the minima is lower than from my > >> >> >> >> >> calulation. > >> >> >> >> >> There > >> >> >> >> >> is no normalization for non bonded so this is weird. I cut > it > >> >> >> >> >> at > >> >> >> >> >> the > >> >> >> >> >> begining as there were very small values and Votca was not > >> >> >> >> >> able > >> >> >> >> >> to > >> >> >> >> >> extrapolate it properly. > >> >> >> >> >> > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> >>> Your second point indeed sounds a bit weired. Could you > >> >> >> >> >>> please > >> >> >> >> >>> post > >> >> >> >> >>> these > >> >> >> >> >>> few curves to help debugging (i.e. the <name>.pot.cur, > >> >> >> >> >>> <name>.pot.new > >> >> >> >> >>> <name>.dist.tgt <name>.dist.new of the iteration 1 > folder)? > >> >> >> >> >>> > >> >> >> >> >>> Victor > >> >> >> >> >> > >> >> >> >> >> > >> >> >> >> >> Please, see attached. > >> >> >> >> >> > >> >> >> >> >> > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> >>> 2013/7/15 <[email protected]> > >> >> >> >> >>>> > >> >> >> >> >>>> Dear Votca Users, > >> >> >> >> >>>> > >> >> >> >> >>>> I have to issues with IBI: > >> >> >> >> >>>> > >> >> >> >> >>>> 1) I took one my ditributions and calculated on my own > >> >> >> >> >>>> potential > >> >> >> >> >>>> W= > >> >> >> >> >>>> -kBT > >> >> >> >> >>>> ln(RDF) and I got different potential than Votca provide > me. > >> >> >> >> >>>> For > >> >> >> >> >>>> instance > >> >> >> >> >>>> lets calculate the potential minimum for the distribution > >> >> >> >> >>>> maximum > >> >> >> >> >>>> of > >> >> >> >> >>>> 162. > >> >> >> >> >>>> Pot = -.249435*ln(164) = -12.69. The minimum of Votca > >> >> >> >> >>>> potential > >> >> >> >> >>>> corresponds > >> >> >> >> >>>> to approximately -16 kJ/mol. Where I missed something? is > it > >> >> >> >> >>>> somehow > >> >> >> >> >>>> normalized? > >> >> >> >> >>>> > >> >> >> >> >>>> 2) After 1st iteration my distribution was much higher > than > >> >> >> >> >>>> the > >> >> >> >> >>>> target > >> >> >> >> >>>> one so I guess the potential should decrease but > apparently > >> >> >> >> >>>> the > >> >> >> >> >>>> new > >> >> >> >> >>>> potential has deeper minima so the next distribution has > a > >> >> >> >> >>>> even > >> >> >> >> >>>> higer > >> >> >> >> >>>> distribution. Could anyone please explain me this? > >> >> >> >> >>>> > >> >> >> >> >>>> Steven > >> >> >> >> >>>> > >> >> >> >> >>>> > >> >> >> >> >>>> -- > >> >> >> >> >>>> You received this message because you are subscribed to > the > >> >> >> >> >>>> Google > >> >> >> >> >>>> Groups "votca" group. > >> >> >> >> >>>> To unsubscribe from this group and stop receiving emails > >> >> >> >> >>>> from > >> >> >> >> >>>> it, > >> >> >> >> >>>> send > >> >> >> >> >>>> an email to [email protected]. > >> >> >> >> >>>> To post to this group, send email to > [email protected]. > >> >> >> >> >>>> Visit this group at http://groups.google.com/group/votca. > > >> >> >> >> >>>> For more options, visit > >> >> >> >> >>>> https://groups.google.com/groups/opt_out. > >> >> >> >> >>>> > >> >> >> >> >>>> > >> >> >> >> >>> > >> >> >> >> >>> > >> >> >> >> > -- > >> >> >> >> > You received this message because you are subscribed to the > >> >> >> >> > Google > >> >> >> >> > Groups > >> >> >> >> > "votca" group. > >> >> >> >> > To unsubscribe from this group and stop receiving emails > from > >> >> >> >> > it, > >> >> >> >> > send > >> >> >> >> > an > >> >> >> >> > email to [email protected]. > >> >> >> >> > To post to this group, send email to [email protected]. > > >> >> >> >> > Visit this group at http://groups.google.com/group/votca. > >> >> >> >> > For more options, visit > >> >> >> >> > https://groups.google.com/groups/opt_out. > >> >> >> >> > > >> >> >> >> > > >> >> >> >> > >> >> >> >> > >> >> >> >> > >> >> >> >> -- > >> >> >> >> Christoph Junghans > >> >> >> >> Web: http://www.compphys.de > >> >> >> > > >> >> >> > -- > >> >> >> > You received this message because you are subscribed to the > Google > >> >> >> > Groups > >> >> >> > "votca" group. > >> >> >> > To unsubscribe from this group and stop receiving emails from > it, > >> >> >> > send > >> >> >> > an > >> >> >> > email to [email protected]. > >> >> >> > To post to this group, send email to [email protected]. > >> >> >> > Visit this group at http://groups.google.com/group/votca. > >> >> >> > For more options, visit > https://groups.google.com/groups/opt_out. > >> >> >> > > >> >> >> > > >> >> > > >> >> > -- > >> >> > You received this message because you are subscribed to the Google > >> >> > Groups > >> >> > "votca" group. > >> >> > To unsubscribe from this group and stop receiving emails from it, > >> >> > send > >> >> > an > >> >> > email to [email protected]. > >> >> > > >> >> > To post to this group, send email to [email protected]. > >> >> > Visit this group at http://groups.google.com/group/votca. > >> >> > For more options, visit https://groups.google.com/groups/opt_out. > >> >> > > >> >> > > >> >> > >> >> > >> >> > >> >> -- > >> >> Christoph Junghans > >> >> Web: http://www.compphys.de > >> > > >> > -- > >> > You received this message because you are subscribed to the Google > >> > Groups > >> > "votca" group. > >> > To unsubscribe from this group and stop receiving emails from it, > send > >> > an > >> > email to [email protected]. > >> > > >> > To post to this group, send email to [email protected]. > >> > Visit this group at http://groups.google.com/group/votca. > >> > For more options, visit https://groups.google.com/groups/opt_out. > >> > > >> > > >> > >> > >> > >> -- > >> Christoph Junghans > >> Web: http://www.compphys.de > > > > -- > > You received this message because you are subscribed to the Google > Groups > > "votca" group. > > To unsubscribe from this group and stop receiving emails from it, send > an > > email to [email protected] <javascript:>. > > To post to this group, send email to [email protected]<javascript:>. > > > Visit this group at http://groups.google.com/group/votca. > > For more options, visit https://groups.google.com/groups/opt_out. > > > > > > > > -- > Christoph Junghans > Web: http://www.compphys.de > -- You received this message because you are subscribed to the Google Groups "votca" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/votca. For more options, visit https://groups.google.com/groups/opt_out.
