AIM Cancer: Automated Integration of Mechanisms in Cancer
Abstract
The assembly of knowledge into executable models that can be simulated increases our understanding of complex systems. We propose to build a framework that will allow for creating and studying causal, explanatory models of complicated systems in which interactions have important causal effects. The modules included in the framework will provide functionality necessary for automation of assembly of information from published literature into network models, and explanation of systems represented by these models. Specifically, the domain of this work are pathways and mechanisms in cancer, but we anticipate that techniques and modules included in the framework will be beneficial for similar future endeavors in other areas of biology or sciences. We are also anticipating application of our framework on infectious diseases, as well as in synthetic biology. Biochemical pathways are interlinked and interdependent within the cell. However, they are often divided into signaling pathways, metabolic networks, gene regulatory networks, etc., and studied only within those pathway categories. Furthermore, cell biology is always contextual and involves the cellular matrices and other stromal/inflammatory cells as the ÔsoilÕ suitable for cancer persistence that need explanation for a full understanding of cancer biology. There exist a number of public databases that catalog biochemical pathways and allow access to their computational models, necessary for reasoning about pathways. Still, complex and paradoxical relationships that are more nuanced and context dependent can be lost in these databases. At the same time, model building is usually done according to specific project goals, for specific cell types, or specific cell processes. In addition, modeling largely depends on the experimental data available and thus, pathway models have different levels of granularity for modeled elements and different degrees of precision for element update functions. To this end, different types of models exist, such as kinetic models, logical models, rule-based models, multi-agent models, statistical models. Moreover, model development of the same system is often repeated by different researchers, due to the fact that the results are not shared in a consistent fashion. To allow for efficient handling of the complexity of the disease, and of the ever-increasing amount of information existing in published work, enabled by large-scale acquisition of data during experiments, it is critical to employ the computational power we have at hand, and even more, existing techniques in natural language processing, data mining, machine learning and design automation. We are, therefore, facing challenges in all three components of automated model design from literature: information extraction through reading, model assembly, and explanation of obtained results. We are addressing these challenges by designing, implementing and integrating a collection of modules that rely on our team s strengths which span several areas critical for the success of this project, namely: computer-aided automation of causal inference, reasoning, model design and analysis, on one side, and knowledge and expertise in cancer biology, systems biology and cell-signaling, on the other. Our proposed framework will use the standardized representation of knowledge, readable by computers, and as such, facilitate further reasoning about missing mechanisms and incorporation of complete pathways of unprecedented scale and accuracy. In turn, this will provide biologists and clinicians the means to efficiently uncover cause and effect relationships, and determine strategies by which these relationships can be altered or manipulated to prevent, cure, or control the disease. Finally, we anticipate that the outcomes of our work will lead to significant speedup not only in biological research and medical practices in cancer, but also in other domains, such as brain science, climate, or social science.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 25, 2018
- Source ID
- W911NF1710135
Entities
People
- Nataša Miškov-Živanov
Organizations
- Army Contracting Command
- Defense Advanced Research Projects Agency
- University of Pittsburgh