Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Completed Research Forum: FightAIDS@Home Thread: Better tools for AIDS drug research / Scripps team publishes a paper |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 12
|
Author |
|
ErikaT
Former World Community Grid Admin USA Joined: Apr 27, 2009 Post Count: 912 Status: Offline Project Badges: |
Hello all,
The Scripps team has published a paper proving the effectiveness of a method to more accurately predict bindings between protein targets and drug candidates, which could benefit FightAIDS@Home and other World Community Grid drug discovery projects. Please read this News article to find out more. Thanks! ErikaT |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I am glad to hear that this method reduces the number of false positives. I hope somebody can point me toward an explanation of this result that I can understand. The news article has been carefully simplified but . . .
Lawrence |
||
|
[CSF] Thomas Dupont
Veteran Cruncher Joined: Aug 25, 2013 Post Count: 685 Status: Offline |
Thanks for the heads-up Erika
---------------------------------------- |
||
|
cjslman
Master Cruncher Mexico Joined: Nov 23, 2004 Post Count: 2082 Status: Offline Project Badges: |
If Lawrence was asking for something more simpler... I didn't even get past the title of the paper !!! But I agree with everybody that this is good news and hope that it can be incorporated into the WCG projects soon.
----------------------------------------CJSL Crunching for a better world... |
||
|
Seoulpowergrid
Veteran Cruncher Joined: Apr 12, 2013 Post Count: 799 Status: Offline Project Badges: |
@lawrencehardin
----------------------------------------It is much cheaper to run these tests on a computer than physically combining the materials in a lab. But computers are not 100% accurate and will sometimes give false positives - something that looks like a cure or a step in the right direction...but it's not. Things that look like potential cures have to be physically tested in a lab which isn't cheap and takes a long(?) amount of time. Also as budgets are always tight, scientists might run the risk of having 50 possible "hits" or possible cures, but only enough money to test 20 in a year. The Scripps people, taking part in the SAMPL4 competition, found a way to make the computer modeling more accurate = less false positives = less time/money wasted as there are less bad computer results that need to be checked in a lab. The more accurate these things are the better future WUs can be at finding drug candidates for diseases. edit: "I didn't even get past the title of the paper !!!" Honestly the title is way too technical for me too! Which brings up the idea to create not just the "Lay Person Abstract" but the "Lay person Title!" My personal attempt would be "Making AutoDock Vina WUs more accurate: The results of a science challenge" I know that is really simplifying things but I couldn't get through the technical version Cheers :) [Edit 1 times, last edit by Seoulpowergrid at Jun 25, 2014 12:48:05 AM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello,
These are some good questions and ideas for future updates on our published articles. A layman's title might be "Post-docking techniques improve on AutoDock Vina's Ranking for the Catalytic Domain of HIV-1 Integrase". Below is a short explanation with a longer one to follow for those who would like some technical explanations. Thanks, Dan ---------- Short explanation: The goal of virtual screening is to rank a collection of molecules, in context of a protein structure, so that if binders in the set exist, they will be high-ranking. Seoulpowergrid mentioned some good points about concepts like "having enough money to test 20 in a year." Say we had a collection of 100 molecules, and there were 10 good "hits" in the set. Docking would be perfect if these 10 molecules were in the top 10 ranks. This is not usually the case (if ever). General outcomes of our (MGL) paper and the Levy Lab paper: the Common Pharmacophore Engine and BEDAM re-score AutoDock Vina's results with more accuracy (BEDAM does so more effectively than CPE). Now imagine in this scaled-down example of 100 molecules that we could only test 2 compounds. Hopefully, it is clear that ranking is quite important. Even in a top-10 ranking, it is quite possible to have 8 hits present, but if they are not ranked 3-8, we have just tested 2 false positives. Long explanation: Terminology BEDAM Binding Energy Distribution Analysis Method (...protein-ligand affinities The original paper can be found here . pharmacophore a 3-dimensional collection of molecular interactions. In our study, a set of known binders were used to create a set of pharmacophores to re-score Vina results, giving better scores to those docked poses matching any pharmacophores. enrichment factor ratio of proportion of hits found in the test set over the proportion of hits in the total set. Enrichment factor is a way to measure virtual screening. A higher number indicates a better enrichment of hits in a tested set as compared to the whole set of molecules. Both methods used were ways to increase enrichment given a set of docked poses. CPE slightly increased enrichment, and BEDAM greatly enhanced enrichment. BEDAM calculates a new score based in thermodynamics simulations. Another technique (rank-difference ratio) in this paper attempts to improve choice of a receptor for virtual screening. This has yet to be tested experimentally although trained with crystallographic data. Fortunately, experiments submitted to the WCG will help us study dockings to multiple receptors, leading to experimental testing through our collaborators. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Indeed we have recently published two papers, with another about to be submitted that couples our Docking codes, AutoDock and AutoDock Vina with the BEDAM software from the Levy Laboratory (now at Temple University). A brief rationale and explanation of the computations follows:
Docking software produces two kinds of results: 1) the predicted pose of a ligand molecule bound to the target receptor (protein or other biomolecule); and 2) a score for that pose, which represents the strength of the binding interaction, and which is used during the docking to pick out the best pose. Because docking typically requires computing millions of poses, and evaluating a score for each pose, the score evaluation must be very fast, and thus we must trade accuracy for speed. Because of this, while the pose of the ligand tends to be pretty accurate, the score (which is interpreted as a binding free energy) is not. When it comes to actually obtaining the best ranked chemical compounds (either through purchase or synthesis) based solely on the docking score, we tend to get a number of "false positives" -- compounds that docking predicts will bind strongly, but when tested in the laboratory, actually do not. False positives are costly and time-consuming in the drug development pipeline. The Levy group had developed software called BEDAM (Binding Energy Analysis Method) which uses molecular dynamics (in implicit solvent) to sample the energetics of reorganization that occurs upon ligand binding. This includes both the reorganization of solvent (water), and of the region of the target protein involved in the binding. It also gives a better accounting of the entropic change of the system. Needless to say these computations are very expensive in terms of computer time, and cannot be run during the docking itself. However, since the poses produced by the docking are more reliable than the docking score, they are a good starting point for the BEDAM computations. Thus we use our docking scores to pick the best 100 or so poses, then submit them to BEDAM for recalculation of the binding free energy of the interaction. The papers that have been published describe the results of this procedure, which show a very significant reduction in the number of false positives. We are now in the process, in collaboration with the Levy Group, of getting the BEDAM code ported to the FightAIDS@Home Project on the World Community Grid. With the computing power that volunteers make available for our research, this should greatly improve the results of our virtual screening experiments. |
||
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3265 Status: Offline Project Badges: |
We're getting another application? Cool!
----------------------------------------AMD Ryzen 5 1600AF 4C/8T 3.2 GHz - 85W AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W Intel Z3740 4C/4T 1.8 GHz - 6W [Edit 1 times, last edit by Falconet at Jun 25, 2014 6:36:58 PM] |
||
|
Seoulpowergrid
Veteran Cruncher Joined: Apr 12, 2013 Post Count: 799 Status: Offline Project Badges: |
Wonderful news and congratulations on the improved accuracy and new publications!
----------------------------------------Indeed we have recently published two papers, with another about to be submitted Really quick I want to say Please please please contact the forces that be in charge of this page: http://boinc.berkeley.edu/wiki/Publications_b...ects#World_Community_Grid and get your information updated. This seems to be the definitive list of BOINC related publications and for me personally was big motivation for me to join WCG and this project. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
First let me thank mgl_DNSantiago for his long explanation, which explained what I was missing in the news. This is a method for ranking drug candidates on our computers which also eliminates some false positives. Then let me thank mgl1 for explaining how the software went about doing this.
I am pleased to hear that we can use our computers to help with this. Lawrence |
||
|
|