FightAIDS@Home - Phase 2
Phase 2 of the project takes results from the molecular docking approach used in Phase 1 and produces a re-evaluation of the top hits found from the virtual screening, to further narrow down potential drug candidates.
While Phase 1 conducted virtual screening of chemical libraries using AutoDock and AutoDock Vina software tools, Phase 2 introduces a completely new computational method into the FightAIDS@Home effort: the BEDAM technique, implemented with Academic IMPACT software, a molecular simulation tool. Academic IMPACT models the thermodynamics of binding of a ligand to a protein including the reorganizational free energy of the docked complex. Much more computer time is needed to evaluate each docked complex, so only the top candidates from Phase 1's virtual screen are examined in this way.
The goal of the FightAIDS@Home project is to find new leads for HIV therapeutics. It is the first step in a long pipeline that may eventually lead to a new drug or therapeutic strategy. Subsequent to virtual screening in Phase 1 and further analysis in Phase 2 for hit identification, the compounds identified must be either purchased or synthesized to be tested experimentally. This is an expensive process, so if our list of hit compounds contains a large number of false positives, much time and money will be wasted. With the BEDAM post-processing of the virtual screening results, we can eliminate much of the false positive results from Phase 1 and save time and money in moving the drug candidates into the drug development pipeline.
Also, the methods that we use on HIV drug discovery are generic and can be applied to any other disease target. This means that other drug search projects on World Community Grid should benefit from our implementation of this computational pipeline.
Phase 2 further refines the results from Phase 1 and helps us identify "false positives," thus saving time and money in future drug development processes.
Yes, we plan to continue to do virtual screens of large chemical libraries against a number of HIV targets. As the biology of the HIV life-cycle continues to be researched by the HIVE Center, and other efforts around the world, new targets and variations of existing targets are discovered and can be used in our Phase 1 activities. We are still currently screening many variations and sites from HIV Protease, Integrase and Reverse Transcriptase.
BEDAM stands for “Binding Energy Distribution Analysis Method”. It is a method developed by the Levy group which uses advanced sampling and analysis techniques to calculate absolute binding free energies based on a foundation in statistical mechanics and data generated from molecular dynamics simulations.
In a collaboration between the Olson group at The Scripps Research Institute and the Levy group at Temple University, BEDAM techniques were recently developed and used in a computational challenge (SAMPL4) demonstrating that docking coupled with subsequent BEDAM processing gives more reliable hits. This challenge used blind data from a pharmaceutical company working on HIV Integrase inhibitors. The AutoDock Vina/BEDAM modeling performed the best among all automated computational predictions submitted in the challenge.
The BEDAM approach uses Academic IMPACT, a computer program developed in the Levy Lab at Temple University to model the thermodynamics of protein-ligand binding. It is not a docking tool, but rather a method to evaluate the energetic and entropic components of the complexes from the molecular interactions predicted by docking. It would take prohibitively long to use the BEDAM approach for the millions of compounds without screening them first using the much faster docking software first, as done in Phase 1. Since the energetic estimates from the docking calculations are approximate -- to allow the high throughput needed to screen millions of compounds -- there are many so-called "false positives" in the list of top ligand protein complexes. The BEDAM analysis has been shown to potentially eliminate many of these false positive hits.
The simulations researchers need to carry out are typically very long running and complex, each of which would take several months to run on a single device. Instead, researchers split up each simulation for a given drug candidate into shorter running research tasks: much smaller and more manageable pieces. These pieces of work can be run independently and simultaneously on a volunteer devices.
However for FightAIDS@Home – Phase 2, the research tasks within a single drug candidate simulation are dependent on each other where the output of one task is used as the input to the next. This means longer research tasks within each drug candidate simulation which can’t be run simultaneously.
To handle this complexity, we are using two different, but related mechanisms called trickle messaging and intermediate uploads to allow us to track your progress through a research task and manage the handover of that task from one volunteer to the next to get it completed in the shortest time possible. This way, we can track the progress of the long simulations to ensure that computations are not delayed or lost, while the researchers get the valuable results back as quickly as possible. In addition, volunteers acquire their credits sooner too.
Trickle messaging is a capability that allows your device to communicate with us while working on a research task to report the progress being made by your device. This allows us to determine whether work should continue on that research task or whether insufficient progress is being made and therefore that task should be handed over to another volunteer for processing.
This capability is particularly useful to a project like FightAIDS@Home – Phase 2 because of the nature of its research tasks, which may require more processing time to complete and can therefore be started by one volunteer and completed by another, without losing the progress made by the first volunteer.
A trickle up message is when your device sends a message back to World Community Grid at certain processing milestones to inform us that you’re still making progress on the current research task. Along with intermediate results sent to us by your device, we use this information to:
- Validate your work up to that point and grant credit accordingly;
- Determine whether sufficient progress is being made by your device or whether the task should be handed over to another volunteer.
We are using a mechanism called intermediate uploads whereby at certain processing milestones, your device would send us back partial results for the research task your device is currently working on. This allows us to validate the work you have completed up to that point and helps the researchers examine and interpret the results being returned by the volunteers.
Using the information your device intermittently sends to us during the processing of a research task, we determine the likelihood of your device finishing the task before the completion deadline. If we determine that you are very likely to miss that deadline or you have already missed the deadline, we would send your device a trickle down message to instruct it to stop working on that research task and we then pass it along on to another volunteer.
There are two types of trickle down messages:
- Soft stop: Instruction for your device to continue until the next milestone before stopping the computation of the current research task. This happens when your device is not making sufficient progress on the current calculation. We would then hand over your partial result to another volunteer to continue working on.
- Hard stop: Instruction for your device to stop working on the current research task immediately. This happens when you have already passed the processing deadline without sending in the final full result or that there may be a communication error in sending us your progress. In this case, we would hand over the research task from the point of your last intermediate milestone to another volunteer to resume working on.
In either case, you will be awarded credit for the work you completed up to the point of the last checkpoint.
This mechanism allows work to be completed quicker and for the researchers to receive valuable results sooner.
Unlike many of our research projects, FightAIDS@Home – Phase 2 requires that research tasks within each simulation, and small processing steps within each task, to be carried out in sequence. This means that it would take much longer to get the results to the researchers.
Trickle messaging and intermediate upload capabilities allow us instead to move the same research task from one volunteer to the next, without losing progress along the way. Overall, this assures progress, shortens and stabilizes the processing time required to complete research tasks and speeds up the rate at which valuable results can be returned to the researchers.
While a typical FightAIDS@Home – Phase 2 simulation might take up to a year to complete, using these capabilities means that it can be completed in as little as two months.
Unlike many of our research projects, FightAIDS@Home – Phase 2 does not require redundancy, where the same research task is sent to two devices and the results are compared for consistency. Instead, this project will be using various processing metrics during the computation of a research task to validate that the task is progressing without errors.
Here is a video of the FightAIDS@Home - Phase 2 project graphics:
The right portion of the screen saver shows both the target and candidate compound molecules, depicted as a collection of small spheres that represent the atoms of each molecule. These are the specific molecules that your device is currently working on.
The progress bar towards the bottom of the screen saver represents approximately how much of the current task your device has processed. When it reaches 100%, the computation is complete and the results will then be sent back to World Community Grid, where they will be packaged and delivered to the FightAIDS@Home researchers.
The small spheres represent the atoms in both the target molecule and candidate molecule currently being processed by your device.
The Scripps Research Institute in La Jolla, California, is the largest private biomedical research institute in the US and the home of the research team behind the FightAIDS@Home Project.
