| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: FightAIDS@Home
Thread: Possible BUG in Autodock |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 19
|
|
| Author |
|
|
mgl_ALPerryman
FightAIDS@Home, GO Fight Against Malaria and OpenZika Scientist USA Joined: Aug 25, 2007 Post Count: 283 Status: Offline Project Badges: 
|
Hi Falconet,
Thank you for paying close attention and for listing the details of the issue you noticed. I've asked the other experts in the lab, and this does not seem like an actual bug to us. To be certain we will need the IBM staff to voice their thoughts, since this issue involves the workflow management for jobs that were halted and then resumed (which is something that the FightAIDS@Home scientists have no control over and little knowledge about). But when I read everything on this thread in detail, I was reassured. Since the output you provided kept listing "Successful Completion" for the autogrid4 and autodock4 components of each run, everything should be fine. Observing different "Best Energy" values for different docking runs (i.e., for different "Docking numbers") of a particular ligand is to be expected. Since AutoDock uses a genetic algorithm with random number seeds in its search process, different runs generally produce different values for the best energy binding mode observed. This is one of the reasons why we perform at least 100 independent docking runs per job (i.e., per compound-target pair). A more negative "best energy" value will replace inferior values in the output you can see, but the best energy obtained during each and every run is reported to us in the dlg files (i.e., the docking log files, which contain a huge amount of very detailed data that we then re-cluster and sort). Most importantly, when we use our local, well-tested version of the AutoGrid and AutoDock4.2 code to re-test the performance of the best-ranking ligands in a particular experiment, the best energy values produced in our local re-tests are generally identical to the best energy values reported in the results from World Community Grid (i.e., when they are not exactly the same, they are within 0.01 to 0.1 kcal/mol difference in the best energy value). If there was a significant bug in the version of AutoDock that was ported to World Community Grid, then we would see very different "best energy" values during our local re-tests. Consequently, the members of the FightAIDS@Home team agree that this is not an actual bug. But thank you for bringing it to our attention, anyway. Thank you very much for your continued support, Alex L. Perryman, Ph.D. PS--next time you notice a potential issue, please also send us the details of the particular ligand, target, and work unit number that were involved. |
|||
|
||||
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges: 
|
Hi Falconet, Thank you for paying close attention and for listing the details of the issue you noticed. I've asked the other experts in the lab, and this does not seem like an actual bug to us. To be certain we will need the IBM staff to voice their thoughts, since this issue involves the workflow management for jobs that were halted and then resumed (which is something that the FightAIDS@Home scientists have no control over and little knowledge about). But when I read everything on this thread in detail, I was reassured. Since the output you provided kept listing "Successful Completion" for the autogrid4 and autodock4 components of each run, everything should be fine. Observing different "Best Energy" values for different docking runs (i.e., for different "Docking numbers") of a particular ligand is to be expected. Since AutoDock uses a genetic algorithm with random number seeds in its search process, different runs generally produce different values for the best energy binding mode observed. This is one of the reasons why we perform at least 100 independent docking runs per job (i.e., per compound-target pair). A more negative "best energy" value will replace inferior values in the output you can see, but the best energy obtained during each and every run is reported to us in the dlg files (i.e., the docking log files, which contain a huge amount of very detailed data that we then re-cluster and sort). Most importantly, when we use our local, well-tested version of the AutoGrid and AutoDock4.2 code to re-test the performance of the best-ranking ligands in a particular experiment, the best energy values produced in our local re-tests are generally identical to the best energy values reported in the results from World Community Grid (i.e., when they are not exactly the same, they are within 0.01 to 0.1 kcal/mol difference in the best energy value). If there was a significant bug in the version of AutoDock that was ported to World Community Grid, then we would see very different "best energy" values during our local re-tests. Consequently, the members of the FightAIDS@Home team agree that this is not an actual bug. But thank you for bringing it to our attention, anyway. Thank you very much for your continued support, Alex L. Perryman, Ph.D. PS--next time you notice a potential issue, please also send us the details of the particular ligand, target, and work unit number that were involved. Thank you for your time to adress this issue. I'm glad it's not a bug! Thanks again!!  - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges:
|
Dr. Perryman,
----------------------------------------something off-topic here. A couple months ago the Discovering Dengue Drugs Together Project team posted an update in which they say that all their AutoDock hits from the project phase 1 failed and they actually were expecting it. The fact that they were expecting it makes me wonder if the FAAH should run a second docking software to double check their results, i.e, run AutoDock on the grid and then run another docking software on the grid to test the same compounds. You can read about it here http://www.worldcommunitygrid.org/forums/wcg/...ead,31014_offset,0#317082 and it was Rickjb who pointed out this issue. - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
armstrdj
Former World Community Grid Tech Joined: Oct 21, 2004 Post Count: 695 Status: Offline Project Badges: 
|
Falconet,
Sorry for the delay in responding, there is nothing to worry about with the energy values being printed. It is not a bug in Autodock but perhaps a poor labeling of what is being output. If you are interested the details follow. The energy values you see being output to stderr are coming from the World Community Grid code added to AutoDock. We keep track of the best energy value seen so far for the graphics code. After each docking it checks the new energy value and if it is less than the current best value it updates shared memory with the new value. The line you see sometimes after a docking "Updating Best Energy for WU: -7.96" is outputing the current best energy, in this case -7.96, and not the new value, -9.053500, which replaces -7.96. On a restart the best energy value is restored for the graphics code and this is why you see the -9.053500. Sorry for the confusion this has caused but rest assurred everything is good with the results. Thanks, armstrdj |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges:
|
Falconet, Sorry for the delay in responding, there is nothing to worry about with the energy values being printed. It is not a bug in Autodock but perhaps a poor labeling of what is being output. If you are interested the details follow. The energy values you see being output to stderr are coming from the World Community Grid code added to AutoDock. We keep track of the best energy value seen so far for the graphics code. After each docking it checks the new energy value and if it is less than the current best value it updates shared memory with the new value. The line you see sometimes after a docking "Updating Best Energy for WU: -7.96" is outputing the current best energy, in this case -7.96, and not the new value, -9.053500, which replaces -7.96. On a restart the best energy value is restored for the graphics code and this is why you see the -9.053500. Sorry for the confusion this has caused but rest assurred everything is good with the results. Thanks, armstrdj I see. Thanks armstrdj! - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
mgl_ALPerryman
FightAIDS@Home, GO Fight Against Malaria and OpenZika Scientist USA Joined: Aug 25, 2007 Post Count: 283 Status: Offline Project Badges:
|
Hi Falconet,
I just read the thread that you linked. They said that their virtual screens with AutoDock on World Community Grid did not find better hits than the hits that they had previously identified. They might have discovered some new hits, but they just weren't as potent as the other hits they found earlier. When it comes to the success rates of a virtual screen, many factors are involved, such as the particular docking program used for that system, the choice of the target structure, how the target's input file was prepared, which library of compounds was used, how the input files for those compounds were prepared, how the docking calculations were performed (i.e., what set of run parameters were used), how the results were sorted, and how the top compounds were selected for purchasing and testing in "wet lab" assays. Luck helps, too. ;) The idea of using a different docking program to perform a second round of testing is a great idea. In fact, I recently used a protocol like that to identify two active fragments from the results of FAAH Experiment 28. First, I used the AutoDock results from FightAIDS@Home to identify a couple hundred promising compounds. I then re-docked those compounds on our cluster at TSRI to confirm the results from World Community Grid. I then docked all of the top compounds again using the new docking program "AutoDock VINA," which uses a different search algorithm and a different scoring function. Out of the compounds that scored very well with both AutoDock and VINA, 10 fragments were purchased and assayed by our collaborators. In the preliminary results, 2 of these 10 compounds were shown to inhibit HIV protease. These two active fragments are still being tested in additional assays, but I should be able to start writing a paper on these results soon. This idea of using a "consensus docking approach" with two different docking programs is a central theme in the proposal I am writing to start a malaria project on World Community Grid (i.e., AutoDocking Molecules to Fight Malaria = AMFM). Thank you very much for your support, Dr. Alex Perryman |
|||
|
||||
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges:
|
Hi Falconet, I just read the thread that you linked. They said that their virtual screens with AutoDock on World Community Grid did not find better hits than the hits that they had previously identified. They might have discovered some new hits, but they just weren't as potent as the other hits they found earlier. When it comes to the success rates of a virtual screen, many factors are involved, such as the particular docking program used for that system, the choice of the target structure, how the target's input file was prepared, which library of compounds was used, how the input files for those compounds were prepared, how the docking calculations were performed (i.e., what set of run parameters were used), how the results were sorted, and how the top compounds were selected for purchasing and testing in "wet lab" assays. Luck helps, too. ;) The idea of using a different docking program to perform a second round of testing is a great idea. In fact, I recently used a protocol like that to identify two active fragments from the results of FAAH Experiment 28. First, I used the AutoDock results from FightAIDS@Home to identify a couple hundred promising compounds. I then re-docked those compounds on our cluster at TSRI to confirm the results from World Community Grid. I then docked all of the top compounds again using the new docking program "AutoDock VINA," which uses a different search algorithm and a different scoring function. Out of the compounds that scored very well with both AutoDock and VINA, 10 fragments were purchased and assayed by our collaborators. In the preliminary results, 2 of these 10 compounds were shown to inhibit HIV protease. These two active fragments are still being tested in additional assays, but I should be able to start writing a paper on these results soon. This idea of using a "consensus docking approach" with two different docking programs is a central theme in the proposal I am writing to start a malaria project on World Community Grid (i.e., AutoDocking Molecules to Fight Malaria = AMFM). Thank you very much for your support, Dr. Alex Perryman Thanks for your reply! And those are great news! Best of luck! - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 445 Status: Offline Project Badges:
|
This idea of using a "consensus docking approach" with two different docking programs is a central theme in the proposal I am writing to start a malaria project on World Community Grid (i.e., AutoDocking Molecules to Fight Malaria = AMFM). Thank you very much for your support, Dr. Alex Perryman Since you're interested in a malaria project, you may want to read a research paper I found recently to determine if Rosetta@Home can supply software suitable for that project. http://www.nature.com/nature/journal/v473/n7346/full/nature09937.html |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
WCG uses a version of the Rosetta software [HPF/HPF2], except the version was locked in so that the databases build with that version are consistent. Malaria was a [brief] target. See google hits: http://www.google.com/search?q=malaria+human+proteome+folding
----------------------------------------edit: one of the papers of HPF/HPF2 specifically mentions malaria http://biology.as.nyu.edu/docs/IO/6579/hpf-results.pdf snip: Future Directions: We will continue this project with a second phase. The second phase will be called Human Proteome Folding 2 (HPF2, we could of thought up a more creative name, i know...) would take important proteins with interesting novel predictions from HPF1 and refine those predicted structures (with something we call all-atom mode) to a higher level of resolution/accuracy (HPF1 == fold resolution, broad fold-function survey; HPF2 == higher res for more detailed conclusions). We will focus on cancer biomarkers, proteins expressed at key times in the infection cycle of malaria and human secreted proteins. We will use a different mode of the Rosetta program to generate higher resolution structures... [Edit 2 times, last edit by Former Member at May 30, 2011 6:51:02 PM] |
||
|
|
|