DOPE score based model selection
Dear Modeller_usage members,
Is it necessary to select the best protein model based on DOPE scores? I have made 100 models of a ~260 residue protein (enzyme) using Modeller. Then I chose the best model based on the DOPE Score and did loop refinement on a 16-residue long loop (50 refined models created) located at active site cleft. In loop-refined model the loop occludes the active site cleft, but in another loop-refined model (with intermediate DOPE score) it is alright. Can I use this loop-refined model? Also, is it essential to refine the loop even when some initial Models have good conformations of the loop? Any suggestions would be very helpful.
Thanks in advance, Vipul
the DOPE score is not a correct indication from my experience. Ive modelled proteins of already known structures, then calculated their RMSD by aligning my computed models through modeller with the structure found in PDB. it has shown that not always the model with the lowest DOPE or nomralized DOPE for that matter has the best RMSD results with the PDB structure.
________________________________________ From: modeller_usage-bounces@salilab.org modeller_usage-bounces@salilab.org on behalf of vipul@nccs.res.in vipul@nccs.res.in Sent: Thursday, April 16, 2015 4:31 PM To: modeller_usage@salilab.org Subject: [modeller_usage] DOPE score based model selection
Dear Modeller_usage members,
Is it necessary to select the best protein model based on DOPE scores? I have made 100 models of a ~260 residue protein (enzyme) using Modeller. Then I chose the best model based on the DOPE Score and did loop refinement on a 16-residue long loop (50 refined models created) located at active site cleft. In loop-refined model the loop occludes the active site cleft, but in another loop-refined model (with intermediate DOPE score) it is alright. Can I use this loop-refined model? Also, is it essential to refine the loop even when some initial Models have good conformations of the loop? Any suggestions would be very helpful.
Thanks in advance, Vipul _______________________________________________ modeller_usage mailing list modeller_usage@salilab.org https://salilab.org/mailman/listinfo/modeller_usage
Sorry to intrude
Dear Mouses, May I ask what method do you use to select the best (or bests) models if you do not relay on DOPE score?.
My experience is that DOPE, like any other score for any other "bioinformatics" prediction method is not perfect, but it has always given me good structures, perhaps not the best, but more than good enough (except when there are "knots" in the structure, but for that I have eyes :) ).
thank you in advance
Oscar C.S.
Oscar Conchillo Solé Group of Computational Biology and Proteomics IBB Data Center Manager and Linux Sysadmin Institut de Biotecnologia I Biomedicina (UAB) mail: txino@bioinf.uab.es telf: 0034 93581 4431; 0034 93586 8939
On 04/16/2015 03:29 PM, Mouses Stamboulian wrote: > the DOPE score is not a correct indication from my experience. Ive modelled proteins of already known structures, then calculated their RMSD by aligning my computed models through modeller with the structure found in PDB. it has shown that not always the model with the lowest DOPE or nomralized DOPE for that matter has the best RMSD results with the PDB structure. > > ________________________________________ > From: modeller_usage-bounces@salilab.orgmodeller_usage-bounces@salilab.org on behalf of vipul@nccs.res.invipul@nccs.res.in > Sent: Thursday, April 16, 2015 4:31 PM > To: modeller_usage@salilab.org > Subject: [modeller_usage] DOPE score based model selection > > Dear Modeller_usage members, > > Is it necessary to select the best protein model based on DOPE scores? > I have made 100 models of a ~260 residue protein (enzyme) using Modeller. > Then I chose the best model based on the DOPE Score and did loop > refinement on a 16-residue long loop (50 refined models created) located > at active site cleft. In loop-refined model the loop occludes the active > site cleft, but in another loop-refined model (with intermediate DOPE > score) it is alright. Can I use this loop-refined model? Also, is it > essential to refine the loop even when some initial Models have good > conformations of the loop? Any suggestions would be very helpful. > > Thanks in advance, > Vipul > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage >
> Sorry to intrude > > Dear Mouses, > May I ask what method do you use to select the best (or bests) models if > you do not relay on DOPE score?. > > My experience is that DOPE, like any other score for any other > "bioinformatics" prediction method is not perfect, but it has always > given me good structures, perhaps not the best, but more than good > enough (except when there are "knots" in the structure, but for that I > have eyes :) ). > > thank you in advance > > Oscar C.S. > > > Oscar Conchillo Solé > Group of Computational Biology and Proteomics > IBB Data Center Manager and Linux Sysadmin > Institut de Biotecnologia I Biomedicina (UAB) > mail: txino@bioinf.uab.es > telf: 0034 93581 4431; 0034 93586 8939 > > > On 04/16/2015 03:29 PM, Mouses Stamboulian wrote: >> the DOPE score is not a correct indication from my experience. Ive >> modelled proteins of already known structures, then calculated their >> RMSD by aligning my computed models through modeller with the structure >> found in PDB. it has shown that not always the model with the lowest >> DOPE or nomralized DOPE for that matter has the best RMSD results with >> the PDB structure. >> >> ________________________________________ >> From: >> modeller_usage-bounces@salilab.orgmodeller_usage-bounces@salilab.org >> on behalf of vipul@nccs.res.invipul@nccs.res.in >> Sent: Thursday, April 16, 2015 4:31 PM >> To: modeller_usage@salilab.org >> Subject: [modeller_usage] DOPE score based model selection >> >> Dear Modeller_usage members, >> >> Is it necessary to select the best protein model based on DOPE scores? >> I have made 100 models of a ~260 residue protein (enzyme) using >> Modeller. >> Then I chose the best model based on the DOPE Score and did loop >> refinement on a 16-residue long loop (50 refined models created) located >> at active site cleft. In loop-refined model the loop occludes the active >> site cleft, but in another loop-refined model (with intermediate DOPE >> score) it is alright. Can I use this loop-refined model? Also, is it >> essential to refine the loop even when some initial Models have good >> conformations of the loop? Any suggestions would be very helpful. >> >> Thanks in advance, >> Vipul >> _______________________________________________ >> modeller_usage mailing list >> modeller_usage@salilab.org >> https://salilab.org/mailman/listinfo/modeller_usage >> _______________________________________________ >> modeller_usage mailing list >> modeller_usage@salilab.org >> https://salilab.org/mailman/listinfo/modeller_usage >> > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage > Dear Oscar, I too found knots in loop-refined models which had best DOPE scores. So is it ok to select a model with intermediate DOPE score over the one which has lowest (best) DOPE score? Thanks, Vipul
Dear Vipul, Totally ok! knots are artificial and you should totally ignore those models with them (once I had to visualize almost 400 models until I found one without them)
And answering your original question: If I don't remember wrong, automodel also includes the function loopmodel for those regions that align with nothing (gaps) so if you are generating enough models in your first run, probably you do not need loopmodel.
cheers OCS
https://salilab.org/modeller/FAQ.html#14
Oscar Conchillo Solé Group of Computational Biology and Proteomics IBB Data Center Manager and Linux Sysadmin Institut de Biotecnologia I Biomedicina (UAB) mail: txino@bioinf.uab.es telf: 0034 93581 4431; 0034 93586 8939
On 04/16/2015 09:09 PM, vipul@nccs.res.in wrote: >> Sorry to intrude >> >> Dear Mouses, >> May I ask what method do you use to select the best (or bests) models if >> you do not relay on DOPE score?. >> >> My experience is that DOPE, like any other score for any other >> "bioinformatics" prediction method is not perfect, but it has always >> given me good structures, perhaps not the best, but more than good >> enough (except when there are "knots" in the structure, but for that I >> have eyes :) ). >> >> thank you in advance >> >> Oscar C.S. >> >> >> Oscar Conchillo Solé >> Group of Computational Biology and Proteomics >> IBB Data Center Manager and Linux Sysadmin >> Institut de Biotecnologia I Biomedicina (UAB) >> mail: txino@bioinf.uab.es >> telf: 0034 93581 4431; 0034 93586 8939 >> >> >> On 04/16/2015 03:29 PM, Mouses Stamboulian wrote: >> >>> the DOPE score is not a correct indication from my experience. Ive >>> modelled proteins of already known structures, then calculated their >>> RMSD by aligning my computed models through modeller with the structure >>> found in PDB. it has shown that not always the model with the lowest >>> DOPE or nomralized DOPE for that matter has the best RMSD results with >>> the PDB structure. >>> >>> ________________________________________ >>> From: >>> modeller_usage-bounces@salilab.orgmodeller_usage-bounces@salilab.org >>> on behalf of vipul@nccs.res.invipul@nccs.res.in >>> Sent: Thursday, April 16, 2015 4:31 PM >>> To: modeller_usage@salilab.org >>> Subject: [modeller_usage] DOPE score based model selection >>> >>> Dear Modeller_usage members, >>> >>> Is it necessary to select the best protein model based on DOPE scores? >>> I have made 100 models of a ~260 residue protein (enzyme) using >>> Modeller. >>> Then I chose the best model based on the DOPE Score and did loop >>> refinement on a 16-residue long loop (50 refined models created) located >>> at active site cleft. In loop-refined model the loop occludes the active >>> site cleft, but in another loop-refined model (with intermediate DOPE >>> score) it is alright. Can I use this loop-refined model? Also, is it >>> essential to refine the loop even when some initial Models have good >>> conformations of the loop? Any suggestions would be very helpful. >>> >>> Thanks in advance, >>> Vipul >>> _______________________________________________ >>> modeller_usage mailing list >>> modeller_usage@salilab.org >>> https://salilab.org/mailman/listinfo/modeller_usage >>> _______________________________________________ >>> modeller_usage mailing list >>> modeller_usage@salilab.org >>> https://salilab.org/mailman/listinfo/modeller_usage >>> >>> >> _______________________________________________ >> modeller_usage mailing list >> modeller_usage@salilab.org >> https://salilab.org/mailman/listinfo/modeller_usage >> >> > Dear Oscar, > I too found knots in loop-refined models which had best DOPE > scores. So is it ok to select a model with intermediate DOPE > score over the one which has lowest (best) DOPE score? > Thanks, > Vipul > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage >
Dear Vipul,
What you are referring to is actually a whole field in the computational structural modeling. Its called "Model Quality Assessment". You can perhaps run something like TMScore or other methods like ProQ2 etc (downloadable methods) or upload them to some of the best MQAP servers. A good place to search for the best methods is the CASP result page ( http://predictioncenter.org/casp11/qa_analysis.cgi).
You can run your models in these methods and see if they give you an appropriate ranking or not. All of them will give you a score just like DOPE score.
Cheers
Arjun
On Fri, Apr 17, 2015 at 1:47 PM, Oscar Conchillo Solé txino@bioinf.uab.es wrote:
> Dear Vipul, > Totally ok! knots are artificial and you should totally ignore those > models with them (once I had to visualize almost 400 models until I found > one without them) > > And answering your original question: > If I don't remember wrong, automodel also includes the function loopmodel > for those regions that align with nothing (gaps) so if you are generating > enough models in your first run, probably you do not need loopmodel. > > cheers > OCS > > > https://salilab.org/modeller/FAQ.html#14 > > Oscar Conchillo Solé > Group of Computational Biology and Proteomics > IBB Data Center Manager and Linux Sysadmin > Institut de Biotecnologia I Biomedicina (UAB) > mail: txino@bioinf.uab.es > telf: 0034 93581 4431; 0034 93586 8939 > > > On 04/16/2015 09:09 PM, vipul@nccs.res.in wrote: > >> Sorry to intrude >>> >>> Dear Mouses, >>> May I ask what method do you use to select the best (or bests) models if >>> you do not relay on DOPE score?. >>> >>> My experience is that DOPE, like any other score for any other >>> "bioinformatics" prediction method is not perfect, but it has always >>> given me good structures, perhaps not the best, but more than good >>> enough (except when there are "knots" in the structure, but for that I >>> have eyes :) ). >>> >>> thank you in advance >>> >>> Oscar C.S. >>> >>> >>> Oscar Conchillo Solé >>> Group of Computational Biology and Proteomics >>> IBB Data Center Manager and Linux Sysadmin >>> Institut de Biotecnologia I Biomedicina (UAB) >>> mail: txino@bioinf.uab.es >>> telf: 0034 93581 4431; 0034 93586 8939 >>> >>> >>> On 04/16/2015 03:29 PM, Mouses Stamboulian wrote: >>> >>> >>>> the DOPE score is not a correct indication from my experience. Ive >>>> modelled proteins of already known structures, then calculated their >>>> RMSD by aligning my computed models through modeller with the structure >>>> found in PDB. it has shown that not always the model with the lowest >>>> DOPE or nomralized DOPE for that matter has the best RMSD results with >>>> the PDB structure. >>>> >>>> ________________________________________ >>>> From: >>>> modeller_usage-bounces@salilab.orgmodeller_usage-bounces@salilab.org >>>> on behalf of vipul@nccs.res.invipul@nccs.res.in >>>> Sent: Thursday, April 16, 2015 4:31 PM >>>> To: modeller_usage@salilab.org >>>> Subject: [modeller_usage] DOPE score based model selection >>>> >>>> Dear Modeller_usage members, >>>> >>>> Is it necessary to select the best protein model based on DOPE scores? >>>> I have made 100 models of a ~260 residue protein (enzyme) using >>>> Modeller. >>>> Then I chose the best model based on the DOPE Score and did loop >>>> refinement on a 16-residue long loop (50 refined models created) located >>>> at active site cleft. In loop-refined model the loop occludes the active >>>> site cleft, but in another loop-refined model (with intermediate DOPE >>>> score) it is alright. Can I use this loop-refined model? Also, is it >>>> essential to refine the loop even when some initial Models have good >>>> conformations of the loop? Any suggestions would be very helpful. >>>> >>>> Thanks in advance, >>>> Vipul >>>> _______________________________________________ >>>> modeller_usage mailing list >>>> modeller_usage@salilab.org >>>> https://salilab.org/mailman/listinfo/modeller_usage >>>> _______________________________________________ >>>> modeller_usage mailing list >>>> modeller_usage@salilab.org >>>> https://salilab.org/mailman/listinfo/modeller_usage >>>> >>>> >>>> >>> _______________________________________________ >>> modeller_usage mailing list >>> modeller_usage@salilab.org >>> https://salilab.org/mailman/listinfo/modeller_usage >>> >>> >>> >> Dear Oscar, >> I too found knots in loop-refined models which had best DOPE >> scores. So is it ok to select a model with intermediate DOPE >> score over the one which has lowest (best) DOPE score? >> Thanks, >> Vipul >> _______________________________________________ >> modeller_usage mailing list >> modeller_usage@salilab.org >> https://salilab.org/mailman/listinfo/modeller_usage >> >> > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage >
>Thank you all. Your suggestions were very helpful in clarifying my doubts. Vipul
Dear Vipul, > > What you are referring to is actually a whole field in the computational > structural modeling. Its called "Model Quality Assessment". You can > perhaps > run something like TMScore or other methods like ProQ2 etc (downloadable > methods) or upload them to some of the best MQAP servers. A good place to > search for the best methods is the CASP result page ( > http://predictioncenter.org/casp11/qa_analysis.cgi). > > You can run your models in these methods and see if they give you an > appropriate ranking or not. All of them will give you a score just like > DOPE score. > > Cheers > > Arjun > > On Fri, Apr 17, 2015 at 1:47 PM, Oscar Conchillo Solé > txino@bioinf.uab.es > wrote: > >> Dear Vipul, >> Totally ok! knots are artificial and you should totally ignore those >> models with them (once I had to visualize almost 400 models until I >> found >> one without them) >> >> And answering your original question: >> If I don't remember wrong, automodel also includes the function >> loopmodel >> for those regions that align with nothing (gaps) so if you are >> generating >> enough models in your first run, probably you do not need loopmodel. >> >> cheers >> OCS >> >> >> https://salilab.org/modeller/FAQ.html#14 >> >> Oscar Conchillo Solé >> Group of Computational Biology and Proteomics >> IBB Data Center Manager and Linux Sysadmin >> Institut de Biotecnologia I Biomedicina (UAB) >> mail: txino@bioinf.uab.es >> telf: 0034 93581 4431; 0034 93586 8939 >> >> >> On 04/16/2015 09:09 PM, vipul@nccs.res.in wrote: >> >>> Sorry to intrude >>>> >>>> Dear Mouses, >>>> May I ask what method do you use to select the best (or bests) models >>>> if >>>> you do not relay on DOPE score?. >>>> >>>> My experience is that DOPE, like any other score for any other >>>> "bioinformatics" prediction method is not perfect, but it has always >>>> given me good structures, perhaps not the best, but more than good >>>> enough (except when there are "knots" in the structure, but for that I >>>> have eyes :) ). >>>> >>>> thank you in advance >>>> >>>> Oscar C.S. >>>> >>>> >>>> Oscar Conchillo Solé >>>> Group of Computational Biology and Proteomics >>>> IBB Data Center Manager and Linux Sysadmin >>>> Institut de Biotecnologia I Biomedicina (UAB) >>>> mail: txino@bioinf.uab.es >>>> telf: 0034 93581 4431; 0034 93586 8939 >>>> >>>> >>>> On 04/16/2015 03:29 PM, Mouses Stamboulian wrote: >>>> >>>> >>>>> the DOPE score is not a correct indication from my experience. Ive >>>>> modelled proteins of already known structures, then calculated their >>>>> RMSD by aligning my computed models through modeller with the >>>>> structure >>>>> found in PDB. it has shown that not always the model with the lowest >>>>> DOPE or nomralized DOPE for that matter has the best RMSD results >>>>> with >>>>> the PDB structure. >>>>> >>>>> ________________________________________ >>>>> From: >>>>> modeller_usage-bounces@salilab.orgmodeller_usage-bounces@salilab.org >>>>> on behalf of vipul@nccs.res.invipul@nccs.res.in >>>>> Sent: Thursday, April 16, 2015 4:31 PM >>>>> To: modeller_usage@salilab.org >>>>> Subject: [modeller_usage] DOPE score based model selection >>>>> >>>>> Dear Modeller_usage members, >>>>> >>>>> Is it necessary to select the best protein model based on DOPE >>>>> scores? >>>>> I have made 100 models of a ~260 residue protein (enzyme) using >>>>> Modeller. >>>>> Then I chose the best model based on the DOPE Score and did loop >>>>> refinement on a 16-residue long loop (50 refined models created) >>>>> located >>>>> at active site cleft. In loop-refined model the loop occludes the >>>>> active >>>>> site cleft, but in another loop-refined model (with intermediate DOPE >>>>> score) it is alright. Can I use this loop-refined model? Also, is it >>>>> essential to refine the loop even when some initial Models have good >>>>> conformations of the loop? Any suggestions would be very helpful. >>>>> >>>>> Thanks in advance, >>>>> Vipul >>>>> _______________________________________________ >>>>> modeller_usage mailing list >>>>> modeller_usage@salilab.org >>>>> https://salilab.org/mailman/listinfo/modeller_usage >>>>> _______________________________________________ >>>>> modeller_usage mailing list >>>>> modeller_usage@salilab.org >>>>> https://salilab.org/mailman/listinfo/modeller_usage >>>>> >>>>> >>>>> >>>> _______________________________________________ >>>> modeller_usage mailing list >>>> modeller_usage@salilab.org >>>> https://salilab.org/mailman/listinfo/modeller_usage >>>> >>>> >>>> >>> Dear Oscar, >>> I too found knots in loop-refined models which had best >>> DOPE >>> scores. So is it ok to select a model with intermediate DOPE >>> score over the one which has lowest (best) DOPE score? >>> Thanks, >>> Vipul >>> _______________________________________________ >>> modeller_usage mailing list >>> modeller_usage@salilab.org >>> https://salilab.org/mailman/listinfo/modeller_usage >>> >>> >> _______________________________________________ >> modeller_usage mailing list >> modeller_usage@salilab.org >> https://salilab.org/mailman/listinfo/modeller_usage >> > > > > -- > **************************************************** > > The dream is not that you see in sleep , dream is which does not let you > sleep. ~Dr. Abdul Kalam (Former President of Republic of India) > ***************************************************** > Arjun Ray > > Mobile : +919582948808 > Email : arjun@ebi.ac.uk / arjun@cbr.su.se arjun.ray@cbr.su.se > Lab Page: http://www.igib.res.in/ssg/index.html > Personal Page: http://www.igib.res.in/ssg/Arjun.html > *********************************************** > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage >
participants (4)
-
Arjun Ray -
Mouses Stamboulian -
Oscar Conchillo Solé -
vipul＠nccs.res.in