 |
 |
 |
 |
| |
|||||
|
|
|
|
|
||||
Designing Defenses Against Biological and Chemical Attacks Olivia Peters 
iological warfare has a long and dishonorablehistory—from Medieval armies catapulting plague victims over city walls, through germ warfare experiments by the Axis Powers in World War II, to the deadly anthrax attacks in the United States in 2001. Waging war through biology is less expensive and time consuming than building the armaments of traditional warfare. And the potential today for turning the positive benefits of biotechnology discoveries into biological threats is enormous. According to Interpol Secretary-General Ronald Noble, al Qaeda intends to use biological weapons at some point and has posted instructions for making them on the Web. To protect its citizens, the United States is continuously preparing for possible attacks. MITRE is applying its diverse expertise, in fields from information technology to the life sciences, to help our sponsors plan and leverage state-of-the-art developments and research in biotechnology. We're currently funding two research projects of our own to this end: Computational Biology and Biotechnology, and Rapid Diagnosis of Biological and Chemical Warfare Agents.
Computational Biotechnology The focus of the Computational Biology and Biotechnology project is to speed response to disease-causing agents—such as anthrax—by identifying points of intervention. We are looking at how such an agent affects the intracellular process and what can be done to halt or delay its effect. To do this, we are re-creating the cellular process in a computer model that we can use to test various intervention strategies. Our first step was to identify how a biological warfare agent disrupts normal cellular processes. If we can understand and model the process, perhaps we can understand how to slow down the effect of the agent long enough for the person to receive treatment. What we are modeling is a cell-signaling pathway, in this case the apoptosis pathway—a process also known as "cell suicide." Suicide happens when a cell receives the message to disintegrate, perhaps because it is infected by an agent that interferes with its protein networks. The apoptosis pathway has been implicated in many traditional diseases, including cancer, and also appears to be a convergence point for several biological agents; the biological agent apparently disrupts normal cell function by engaging it or inhibiting its activation. For example, if you inhale Staphylococcal Enterotoxin B (SEB), the pathogen goes into the cells of your lungs; the cells collapse and your lungs fill up with blood. We're studying the effects of SEB on a cell so we can document what happens to the cell pathway once it has been infected by SEB. How long does it take for SEB to kill a person? How much does it take? How can we stop the process? At what point will intervention save the person infected? Our current cellular model is a network of 93 proteins; this network is the same across most cells and organisms as opposed to many protein networks, which differ by species and sometimes by cell (the pathway to achieve a mechanism in a liver cell may differ from the pathway used in a stomach cell). We are working with collaborators at the Walter Reed Army Institute of Research (WRAIR) Department of Pathology to gather experimental data, and our team has use of its lab facilities for our experiments. Our original model has been translated into a quantitative model using a system of ordinary differential equations that capture the protein network interactions. At the same time, we are developing qualitative models to identify critical pathways and the effects of various feedback loops on patterns of correlation among proteins. We're currently refining the qualitative and quantitative models to make sure they accurately reflect what happens in nature. The qualitative model is used to inform both the quantitative model and the order of experiments that need to be performed. After we complete the model, we will computationally develop and test intervention strategies. We have based our analysis on optimization and qualitative techniques, developing optimization methods to fit computational model outputs to experimentally derived data by varying immeasurable parameters in the model. We used least squares regression optimization to estimate rate constants that cannot be specifically measured in experimental work, and the constrained simplex method provided a way to optimize how the computational and experimental curves match. After we have documented SEB intervention strategies, we'll look at other bioagents and generate the same information for each. There will be a different response for each agent, requiring different intervention techniques. This extended model could also be used to identify potential targets for vaccine development. Although a lot of work has been done in academia on computational modeling of cellular pathways, the focus has not been on biological warfare agents. MITRE's expertise in modeling and simulation and in biotechnology—as well as our access to sensitive government data and labs—has placed us in a unique position to perform this work. Rapid Diagnosis In our Rapid Diagnosis of Biological and Chemical Warfare Agents project,
we are developing a software classifier to quickly identify if a person
has been exposed to a biological or chemical agent, which agent, and when
he or she was exposed. Our first step was to build an exemplar classifier
for cyclosarin (a chemical nerve agent), indicating if a person has been
exposed and to what level. Microarray experiments generate huge, complex datasets that must be correlated and interpreted. We performed feature extraction and dimensionality reduction on the data to identify a combination of genes that indicates exposure to a pathogen. We started with baselines of healthy activity for the genes and then compared these to the same cells exposed to different agents, such as anthrax. The classifiers were trained to differentiate between cells under normal and pathogenic conditions. We received much of our data from several collaborators, including WRAIR and the Edgewood Biological and Chemical Command, which provided experimental data for multiple chemical and biological warfare agents. We're supplementing this data with work being done by MITRE staff at WRAIR's facilities. The classifier selected genes indicative of exposure using methods such as maximum clustering, ideal profile correlation, median based, and maximum likelihood. Often genes are selected that are affected by natural processes (such as time of day), and we are developing a statistically sound method to remove inappropriate genes from our classification. Interestingly, each gene selection method selects different genes, thus we have put some time into understanding the relationship between the sets, as well as the relationship of selected genes to the pathogen under study. Our classification methods have included principal and independent component analysis, naïve Bayesian, empirical Bayesian, and neural networks. These techniques were tested on a microarray community standard dataset, and when the classifiers demonstrated excellent performance we applied them to cyclosarin data. We have been successful in building classifiers for cyclosarin that show with 73 to 79 percent accuracy that a person has been exposed and at what level. We continue to refine the classifiers to improve their overall ability through refined gene sets and classifier combination. So far our classification has been binary (answering the question: has this sample been exposed to this agent, yes/no?). Our next step is to create more complex classifiers, which will be able to assess a sample and determine which pathogen out of many the sample has been exposed to. By developing signatures for each of the different pathogenic agents, we can quickly identify exposure and the specific agent of exposure, as well as provide some clues to the time and dose. These same methods could also be used to allow novel pathogens to be compared to known pathogens, suggesting potential treatments. In this work, we are applying signal processing and information management
approaches to systems biology. Both of these research projects have given
MITRE staff the opportunity to explore areas of significant interest to
our sponsors, while expanding expertise within the company. |
|
||||
| For more information, please contact Olivia Peters using the employee directory. Page last updated: May 24, 2005 | Top of page |
|||||
Solutions That Make a Difference.® |
|
|
Download this issue [3 MB]