July 2011 - modeller_usage

Re: [modeller_usage] modeller_usage Digest, Vol 10, Issue 108
by Ashish Runthala 30 Jul '11

30 Jul '11

Dear Mustafa, As such energy can't discriminate between close decoys, Normalized_dope gives a better idea. You may see the effect of each of the Modeller assessment score through your modeling script this way. a = automodel(env, alnfile = '------', knowns = ('---', '---'), sequence = 'target', assess_methods=(assess.DOPE, assess.GA341,assess.normalized_dope)) Use a redirection operator > to save modeling result. Then use the trajectory of these values in your modeling. You will get your answer if you try modeling a known structure target, so that you can affirm model evaluation through GDT scores. Good Luck, Ashish Ashish Runthala, Lecturer, Structural Biology Cell, Biological Sciences Group, BITS, Pilani Rajasthan, INDIA ----- Original Message ----- From: "modeller usage-request" <modeller_usage-request(a)salilab.org> To: "modeller usage" <modeller_usage(a)salilab.org> Sent: Saturday, July 30, 2011 12:47:57 PM GMT +05:30 Chennai, Kolkata, Mumbai, New Delhi Subject: modeller_usage Digest, Vol 10, Issue 108 Send modeller_usage mailing list submissions to modeller_usage(a)salilab.org To subscribe or unsubscribe via the World Wide Web, visit https://salilab.org/mailman/listinfo/modeller_usage or, via email, send a message with subject or body 'help' to modeller_usage-request(a)salilab.org You can reach the person managing the list at modeller_usage-owner(a)salilab.org When replying, please edit your Subject line so it is more specific than "Re: Contents of modeller_usage digest..." Today's Topics: 1. Re: Good Quality Score (Modeller Caretaker) 2. Regarding Proper Alignment (Ashish Runthala) 3. Re: Regarding Proper Alignment (Modeller Caretaker) 4. Selection of Model (MUSTAFA BUGHIO) ---------------------------------------------------------------------- Message: 1 Date: Thu, 28 Jul 2011 17:18:07 -0700 From: Modeller Caretaker <modeller-care(a)salilab.org> Subject: Re: [modeller_usage] Good Quality Score To: Brian Tsui <btsui17(a)yahoo.com> Cc: "modeller_usage(a)salilab.org" <modeller_usage(a)salilab.org> Message-ID: <4E31FC3F.5010601(a)salilab.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed On 7/26/11 8:33 PM, Brian Tsui wrote: > In the Madhusudhan paper, the cutoff was at least 70% for a SO of 3.5 A. Do you mean the SO was 70% for a cutoff of 3.5A? > What number does that correspond to in the salign output? The percentage of equivalent (i.e. within the cutoff) positions is the qscorepct member of the SalignData object returned by salign(). Set the cutoff with the rms_cutoff argument. And make sure you have QUALITY in the output argument. > For the quality score, is there an objective number that I can use for > every alignment to see whether it is good or not? As I said, only for structure alignments. > Lastly, for "iterative salign," is it recommended to run the regular > version first or just jump straight to using the iterative salign? The > Madhusudhan paper seems to forgo the use of iterative salign if the > alignment is 70% at a SO of 3.5 I'm assuming you're looking at fig 1 in the paper here. There are three iterative procedures within that flowchart; one of them is skipped if the first procedure (simpler alignment using a subset of features) found a reasonable starting guess for the last procedure (structural alignment). The second procedure is simply an attempt to improve this guess if the first failed. Ben Webb, Modeller Caretaker -- modeller-care(a)salilab.org http://www.salilab.org/modeller/ Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage ------------------------------ Message: 2 Date: Fri, 29 Jul 2011 12:19:49 +0530 (IST) From: Ashish Runthala <ashishr(a)bits-pilani.ac.in> Subject: [modeller_usage] Regarding Proper Alignment To: modeller <modeller_usage(a)salilab.org> Message-ID: <30722239.171381311922189682.JavaMail.root(a)mailserver.bits-pilani.ac.in> Content-Type: text/plain; charset=utf-8 Dear Modellers, May somebody tell me the logic of using rms_cutoff=3.5 while comparing and aligning multiple structures. Even when you alter, no difference come up in the generated model. Is this upper bound a relaxed in every modeling case. It may not be so, when same family proteins are aligned. So this strict upper bound on C? distance may leave several large chunks unaligned. Variable gap penalty makes the alignment correct through flexible allowed superposition, but why changing rms_cutoff didn't generate better models, not even 0.1 GDT_TS improvement was not observed. Thanks Ashish Ashish Runthala, Lecturer, Structural Biology Cell, Biological Sciences Group, BITS, Pilani Rajasthan, INDIA ------------------------------ Message: 3 Date: Fri, 29 Jul 2011 15:43:30 -0700 From: Modeller Caretaker <modeller-care(a)salilab.org> Subject: Re: [modeller_usage] Regarding Proper Alignment To: Ashish Runthala <ashishr(a)bits-pilani.ac.in> Cc: modeller <modeller_usage(a)salilab.org> Message-ID: <4E333792.1070802(a)salilab.org> Content-Type: text/plain; charset=UTF-8; format=flowed On 7/28/11 11:49 PM, Ashish Runthala wrote: > May somebody tell me the logic of using > rms_cutoff=3.5 while comparing and aligning multiple structures. Even > when you alter, no difference come up in the generated model. Changing rms_cutoff will not change the alignment. The cutoff is simply the distance below which two residues are considered to be structurally equivalent when aligned, so it will only affect measures such as the structural overlap (and quality score). Ben Webb, Modeller Caretaker -- modeller-care(a)salilab.org http://www.salilab.org/modeller/ Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage ------------------------------ Message: 4 Date: Sat, 30 Jul 2011 12:17:54 +0500 From: MUSTAFA BUGHIO <mustafab4(a)gmail.com> Subject: [modeller_usage] Selection of Model To: modeller_usage(a)salilab.org Message-ID: <CALHZR253S3mXhzH-=j4a8nM10__Tv+F+G2DNigfJHgZu=BEPjQ(a)mail.gmail.com> Content-Type: text/plain; charset="iso-8859-1" Dear Ben I wanna repeat the question already asked by some one regarding the selection of best model criterion. I have developed pentamer Structure using model-multichain.py command i want to know on what basis i should select the best model out of 20 models. I have checked both modeller objective function scores and also Dope score but there i could not observe any correlation between the two. Can i get your input in this regard Regards -- MUSTAFA BUGHIO Junior Research Fellow Lab # P-133 Computational Medicinal Chemistry and Drug Design Panjwani Center For Molecular Medicine and Drug Research, International Center for Chemical Sciences, University of Karachi, Karachi 75270 Pakistan, +923313425010

1 0

Selection of Model
by MUSTAFA BUGHIO 30 Jul '11

30 Jul '11

Dear Ben I wanna repeat the question already asked by some one regarding the selection of best model criterion. I have developed pentamer Structure using model-multichain.py command i want to know on what basis i should select the best model out of 20 models. I have checked both modeller objective function scores and also Dope score but there i could not observe any correlation between the two. Can i get your input in this regard Regards -- MUSTAFA BUGHIO Junior Research Fellow Lab # P-133 Computational Medicinal Chemistry and Drug Design Panjwani Center For Molecular Medicine and Drug Research, International Center for Chemical Sciences, University of Karachi, Karachi 75270 Pakistan, +923313425010

1 0

Regarding Proper Alignment
by Ashish Runthala 29 Jul '11

29 Jul '11

Dear Modellers, May somebody tell me the logic of using rms_cutoff=3.5 while comparing and aligning multiple structures. Even when you alter, no difference come up in the generated model. Is this upper bound a relaxed in every modeling case. It may not be so, when same family proteins are aligned. So this strict upper bound on Cα distance may leave several large chunks unaligned. Variable gap penalty makes the alignment correct through flexible allowed superposition, but why changing rms_cutoff didn't generate better models, not even 0.1 GDT_TS improvement was not observed. Thanks Ashish Ashish Runthala, Lecturer, Structural Biology Cell, Biological Sciences Group, BITS, Pilani Rajasthan, INDIA

2 1

Good Quality Score
by Brian Tsui 28 Jul '11

28 Jul '11

Dear all, In the output for salign, there is a raw quality score, a percentage quality score, and many other values. What value should I focus on to make sure that I have a good alignment? Is there a cutoff that I should use? If so, are there references that I could cite? Thanks, Brian

2 3

How to build customized salign result?
by Ashish Runthala 25 Jul '11

25 Jul '11

Dear Modellers, Tell me the solution to a query. Suppose I have 4 templates for a target sequence. And while using salign_multiple.py, I want to start the progressive alignment specifically from template 2 only, to let others follow this template2 alignment. Tell me the way to make salign force the progressive alignment starting on from a specific template only? Thanks in advance, Ashish -------- Ashish Runthala, Lecturer, Structural Biology Cell, Biological Sciences Group, BITS, Pilani Rajasthan, INDIA

2 1

Dope profile of tetramer
by Matthias Schmidt 22 Jul '11

22 Jul '11

Hi, I have a model of a tetramer, and I would like to evaluate/show only the DOPE profile for one chain, because the tetramer is symmetric. Here is the script I used for calculating the DOPE profile: from modeller import * from modeller.scripts import complete_pdb log.verbose() # request verbose output env = environ() env.libs.topology.read(file='$(LIB)/top_heav.lib') # read topology env.libs.parameters.read(file='$(LIB)/par.lib') # read parameters # read model file mdl = complete_pdb(env, 'model.B99990001.pdb') # Assess with DOPE: s = selection(mdl) # all atom selection s.assess_dope(output='ENERGY_PROFILE NO_REPORT', file='model.profile', normalize_profile=True, smoothing_window=15) and for plotting: import pylab import modeller def get_profile(profile_file, seq): """Read `profile_file` into a Python array, and add gaps corresponding to the alignment sequence `seq`.""" # Read all non-comment and non-blank lines from the file: f = file(profile_file) vals = [] for line in f: if not line.startswith('#') and len(line) > 10: spl = line.split() vals.append(float(spl[-1])) # Insert gaps into the profile corresponding to those in seq: for n, res in enumerate(seq.residues): for gap in range(res.get_leading_gaps()): vals.insert(n, None) # Add a gap at position '0', so that we effectively count from 1: vals.insert(0, None) return vals e = modeller.environ() a = modeller.alignment(e, file='3zrs-align-nterm.ali') model = get_profile('3zrs-nterm.profile', a['3zrs-nterm']) template = get_profile('3zrs-nterm.profile', a['3zrs-nterm']) # Plot the template and model profiles in the same plot for comparison: pylab.figure(1, figsize=(10,6)) pylab.xlabel('Alignment position') pylab.ylabel('DOPE per-residue score') pylab.plot(model, color='red', linewidth=2, label='Model') pylab.plot(template, color='green', linewidth=2, label='Template') pylab.legend() pylab.savefig('dope_profile.png', dpi=65) These scripts are from the tutorial on the website. Best, Matthias

2 1

SALIGN ERROR to clarify
by Nava Moran 21 Jul '11

21 Jul '11

dear Modeller, trying to upload a sequence in fasta mode to SALIGN I got this error: Sorry, your request was not processed due to the following reason: * directory (/modbase5/home/salign/todo/1299736963_SA_230/upload/unzip/) Core(10)_Helices is still tainted at /usr/lib/perl5/5.8.8/File/Find.pm line 837. *is this my fault? or something else? nava

2 1

Re: [modeller_usage] modeller_usage Digest, Vol 10, Issue 104
by Ashish Runthala 21 Jul '11

21 Jul '11

Dear Sir Benn, You are absolutely correct about CA-CA distance deviations, but when you choose 2 templates with CA-CA distances 5, 6 angstroms, then it will give you a better model as compared to 5, 10 angstroms set. Secondly, precise sequence alignment is always neccessary, but correctly aligned incorrect templates will still give bad models and probably shifted residues as was the general case till CASP5 and CASP6. Iterative Structural alignment also forces mathematical satisfaction of alignment, with no justification of biological significance of optimally placed gaps. So we still need better alignments. Now I request your expert reply against these views. Ashish Runthala, Lecturer, Structural Biology Cell, Biological Sciences Group, BITS, Pilani Rajasthan, INDIA Message: 2 Date: Thu, 21 Jul 2011 08:36:05 -0700 From: Modeller Caretaker <modeller-care(a)salilab.org> Subject: Re: [modeller_usage] Multiple Sequence Alignment vs Single Alignment To: Daniel Li <leeyuhang(a)yahoo.com> Cc: modeller_usage(a)salilab.org Message-ID: <4E284765.1020106(a)salilab.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed On 7/21/11 12:16 AM, Daniel Li wrote: > From reading it seems as if multiple templates should usually generate > better scores but running 3 trials of 600 with different sets of > templates for different proteins I still get better results from my > single template models. So is there something I have to manipulate with > the scripts and parameters? Multiple templates generally do give better models, but they are more sensitive to alignment errors - the templates have to be quite well aligned with each other, otherwise they will present conflicting information (e.g. one template says that a given CA-CA distance is 5 angstroms, and another 10 angstroms) which is difficult to satisfy. Ben Webb, Modeller Caretaker -- modeller-care(a)salilab.org http://www.salilab.org/modeller/ Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage ------------------------------ _______________________________________________ modeller_usage mailing list modeller_usage(a)salilab.org https://salilab.org/mailman/listinfo/modeller_usage End of modeller_usage Digest, Vol 10, Issue 104 ***********************************************

1 0

Multiple Sequence Alignment vs Single Alignment
by Daniel Li 21 Jul '11

21 Jul '11

Hi, I have been reading about modeller usage for awhile and ran some tests with different models from the dual specificity phosphatase family. However, whenever I run the modelling with salign for multiple templates, my GA341 score is many folds lower (under 0.1) and the dope score/molpdf scores are lower as well compared to just a single template. The templates that I use all give me a decent 50%+ identity based on output from build profile and I generally use 3 templates for my multiple template modeling.From reading it seems as if multiple templates should usually generate better scores but running 3 trials of 600 with different sets of templates for different proteins I still get better results from my single template models. So is there something I have to manipulate with the scripts and parameters?I'm sorry for the vague question because I'm not too sure what kind of additional information to provide but I hope someone can provide me with more insight into this. Thank you! Daniel

2 1

Re: [modeller_usage] modeller_usage Digest, Vol 10, Issue 103
by Ashish Runthala 21 Jul '11

21 Jul '11

Dear Daniel, Multiple templates do improve the prediction accuracy as compared to the single best templates. This has been published in several papers. This has been reviewed correctly here Anna Tramontano: Protein Struture Prediction: Concepts and Applications Wiley-VCH; 2006. But here is a trick too, first of all you didn't mention what you actually used. Please answer these queries to help you practically with the best possible answer. 1. Did you use salign_iterative/salign_multiple? 2. What was the length of your sequence? 3. Did you do increased sampling? All the Scores that you discussed won't allow you to rank templates and select their correct set. Hence, it is beneficial if you try to take the correct set of a few biologically related templates. Good Luck, Ashish Ashish Runthala, Lecturer, Structural Biology Cell, Biological Sciences Group, BITS, Pilani Rajasthan, INDIA ----- Original Message ----- From: "modeller usage-request" <modeller_usage-request(a)salilab.org> To: "modeller usage" <modeller_usage(a)salilab.org> Sent: Thursday, July 21, 2011 12:59:26 PM GMT +05:30 Chennai, Kolkata, Mumbai, New Delhi Subject: modeller_usage Digest, Vol 10, Issue 103 Send modeller_usage mailing list submissions to modeller_usage(a)salilab.org To subscribe or unsubscribe via the World Wide Web, visit https://salilab.org/mailman/listinfo/modeller_usage or, via email, send a message with subject or body 'help' to modeller_usage-request(a)salilab.org You can reach the person managing the list at modeller_usage-owner(a)salilab.org When replying, please edit your Subject line so it is more specific than "Re: Contents of modeller_usage digest..." Today's Topics: 1. Re: ModWeb details -> Modeller? (Modeller Caretaker) 2. Re: Help with modeller 9.9 in windows (Modeller Caretaker) 3. Multiple Sequence Alignment vs Single Alignment (Daniel Li) ---------------------------------------------------------------------- Message: 1 Date: Mon, 18 Jul 2011 12:13:13 -0700 From: Modeller Caretaker <modeller-care(a)salilab.org> Subject: Re: [modeller_usage] ModWeb details -> Modeller? To: rbelew(a)ucsd.edu Cc: modeller_usage(a)salilab.org Message-ID: <4E2485C9.3040006(a)salilab.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed On 7/17/11 10:57 PM, richard k belew wrote: > i've done a couple of runs using the ModWeb server, and i'm wondering > how to recapitulate its results using Modeller directly? ModWeb gives > a .pdb file as a result, with a header that tells the template it > selected, and regions in it that were used for modeling. but: > - what method is used for selecting this template? > - is it possible to get the alignment that was used? If you asked for your results to be uploaded into ModBase, you can download the alignment file for each model from that web interface. The header of the alignment file specifies the method that was used to find the template (e.g. sequence-sequence, sequence-profile). If you instead chose to have the models sent directly to you, the alignment file isn't available. But it might still be on our system somewhere - if you email modweb(a)salilab.org with your ModWeb job ID we can see if we can find it for you. Ben Webb, Modeller Caretaker -- modeller-care(a)salilab.org http://www.salilab.org/modeller/ Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage ------------------------------ Message: 2 Date: Mon, 18 Jul 2011 12:21:16 -0700 From: Modeller Caretaker <modeller-care(a)salilab.org> Subject: Re: [modeller_usage] Help with modeller 9.9 in windows To: "Greene, Brandon Lee" <blgree2(a)emory.edu> Cc: "modeller_usage(a)salilab.org" <modeller_usage(a)salilab.org> Message-ID: <4E2487AC.1000003(a)salilab.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed On 7/18/11 7:22 AM, Greene, Brandon Lee wrote: > There are three inorganic residues in the first chain and one in the > second placed as "." in my ali file. Your alignment file does not match the PDB. Modeller reads the residues sequentially, so what it will see is - A chain amino acids - B chain amino acids - A chain ligands - B chain ligands So your sequence should also be in this order (XXXXXX/XXXXXX/.../.). Alternatively, you can edit the PDB file to place the A chain ligands immediately after the A chain amino acids and leave your alignment unchanged. > Using this structure and the template ali file to do the alignment (w/ > salign) I get poor alignment in the PAP file, but not in the PIR file??? > (P/---XXXX instead of P---/XXX in PAP). Those two alignments are identical. Since neither gaps nor chain breaks are "real" residues, their relative ordering is irrelevant - it only matters which "real" residues (amino acids or ligands) they come between. > "No aligned template residues for BLK residue; number of templates: 1 > Make sure that each BLK residue in your target sequence is aligned with > your templates..." Just like it says, a BLK (".") residue in the target sequence must be aligned with a residue in the template. Usually this means you need a "." residue in the template in the same place as the "." in the target. You can't align a "." with a chain break or gap, because "." means to copy the aligned residue directly from the template to the target, so it needs a "real" residue to copy from. Ben Webb, Modeller Caretaker -- modeller-care(a)salilab.org http://www.salilab.org/modeller/ Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage ------------------------------ Message: 3 Date: Thu, 21 Jul 2011 00:16:06 -0700 (PDT) From: Daniel Li <leeyuhang(a)yahoo.com> Subject: [modeller_usage] Multiple Sequence Alignment vs Single Alignment To: modeller_usage(a)salilab.org Message-ID: <1311232566.71184.YahooMailClassic(a)web161902.mail.bf1.yahoo.com> Content-Type: text/plain; charset="iso-8859-1" Hi, I have been reading about modeller usage for awhile and ran some tests with different models from the dual specificity phosphatase family. However, whenever I run the modelling with salign for multiple templates, my GA341 score is many folds lower (under 0.1) and the dope score/molpdf scores are lower as well compared to just a single template.?The templates that I use all give me a decent 50%+ identity based on output from build profile and I generally use 3 templates for my multiple template modeling.From reading it seems as if multiple templates should usually generate better scores but running 3 trials of 600 with different sets of templates for different proteins I still get better results from my single template models. So is there something I have to manipulate with the scripts and parameters?I'm sorry for the vague question because I'm not too sure what kind of additional information to provide but I hope someone can provide me with more insight into this. Thank you! Daniel

1 0