Constructing explanatory process models from biological data and knowledge

Multiple Levels of Heuristic Reasoning Processes in Scientific Model Construction

Science historians have recognized the importance of heuristic reasoning strategies for constructing theories, but their extent and degree of organization are still poorly understood. This paper first consolidates a set of important heuristic strategies for constructing scientific models from three books, including studies in the history of genetics and electromagnetism, and an expert think-aloud study in the field of mechanics. The books focus on qualitative reasoning strategies (processes) involved in creative model construction, scientific breakthroughs, and conceptual change. Twenty four processes are examined, most of which are field-general, but all are heuristic in not being guaranteed to work. An organizing framework is then proposed as a four-level hierarchy of nested reasoning processes and subprocesses at different size and time scales, including: Level (L4) Several longer-time-scale Major Modeling Modes, such as Model Evolution and Model Competition; the former mode utilizes: (L3) Modeling Cycle Phases of Model Generation, Evaluation, and Modification under Constraints; which can utilize: (L2) Thirteen Tactical Heuristic Processes, e.g., Analogy, Infer new model feature (e.g., by running the model), etc.; many of which selectively utilize: (L1) Grounded Imagistic Processes, namely Mental Simulations and Structural Transformations. Incomplete serial ordering in the framework gives it an intermediate degree of organization that is neither anarchistic nor fully algorithmic. Its organizational structure is hypothesized to promote a difficult balance between divergent and convergent processes as it alternates between them in modeling cycles with increasingly constrained modifications. Videotaped think-aloud protocols that include depictive gestures and other imagery indicators indicate that the processes in L1 above can be imagistic. From neurological evidence that imagery uses many of the same brain regions as actual perception and action, it is argued that these expert reasoning processes are grounded in the sense of utilizing the perceptual and motor systems, and interconnections to and possible benefits for reasoning processes at higher levels are examined. The discussion examines whether this grounding and the various forms of organization in the framework may begin to explain how processes that are only sometimes useful and not guaranteed to work can combine successfully to achieve innovative scientific model construction.

Nobel Turing Challenge: creating the engine for scientific discovery

Scientific discovery has long been one of the central driving forces in our civilization. It uncovered the principles of the world we live in, and enabled us to invent new technologies reshaping our society, cure diseases, explore unknown new frontiers, and hopefully lead us to build a sustainable society. Accelerating the speed of scientific discovery is therefore one of the most important endeavors. This requires an in-depth understanding of not only the subject areas but also the nature of scientific discoveries themselves. In other words, the "science of science" needs to be established, and has to be implemented using artificial intelligence (AI) systems to be practically executable. At the same time, what may be implemented by "AI Scientists" may not resemble the scientific process conducted by human scientist. It may be an alternative form of science that will break the limitation of current scientific practice largely hampered by human cognitive limitation and sociological constraints. It could give rise to a human-AI hybrid form of science that shall bring systems biology and other sciences into the next stage. The Nobel Turing Challenge aims to develop a highly autonomous AI system that can perform top-level science, indistinguishable from the quality of that performed by the best human scientists, where some of the discoveries may be worthy of Nobel Prize level recognition and beyond.

Synthetic biology as it relates to CAM photosynthesis: challenges and opportunities

To meet future food and energy security needs, which are amplified by increasing population growth and reduced natural resource availability, metabolic engineering efforts have moved from manipulating single genes/proteins to introducing multiple genes and novel pathways to improve photosynthetic efficiency in a more comprehensive manner. Biochemical carbon-concentrating mechanisms such as crassulacean acid metabolism (CAM), which improves photosynthetic, water-use, and possibly nutrient-use efficiency, represent a strategic target for synthetic biology to engineer more productive C3 crops for a warmer and drier world. One key challenge for introducing multigene traits like CAM onto a background of C3 photosynthesis is to gain a better understanding of the dynamic spatial and temporal regulatory events that underpin photosynthetic metabolism. With the aid of systems and computational biology, vast amounts of experimental data encompassing transcriptomics, proteomics, and metabolomics can be related in a network to create dynamic models. Such models can undergo simulations to discover key regulatory elements in metabolism and suggest strategic substitution or augmentation by synthetic components to improve photosynthetic performance and water-use efficiency in C3 crops. Another key challenge in the application of synthetic biology to photosynthesis research is to develop efficient systems for multigene assembly and stacking. Here, we review recent progress in computational modelling as applied to plant photosynthesis, with attention to the requirements for CAM, and recent advances in synthetic biology tool development. Lastly, we discuss possible options for multigene pathway construction in plants with an emphasis on CAM-into-C3 engineering.

A strategy for selecting data mining techniques in metabolomics

There is a general agreement that the development of metabolomics depends not only on advances in chemical analysis techniques but also on advances in computing and data analysis methods. Metabolomics data usually requires intensive pre-processing, analysis, and mining procedures. Selecting and applying such procedures requires attention to issues including justification, traceability, and reproducibility. We describe a strategy for selecting data mining techniques which takes into consideration the goals of data mining techniques on the one hand, and the goals of metabolomics investigations and the nature of the data on the other. The strategy aims to ensure the validity and soundness of results and promote the achievement of the investigation goals.

Two kinds of knowledge in scientific discovery

Research on computational models of scientific discovery investigates both the induction of descriptive laws and the construction of explanatory models. Although the work in law discovery centers on knowledge-lean approaches to searching a problem space, research on deeper modeling tasks emphasizes the pivotal role of domain knowledge. As an example, our own research on inductive process modeling uses information about candidate processes to explain why variables change over time. However, our experience with IPM, an artificial intelligence system that implements this approach, suggests that process knowledge is insufficient to avoid consideration of implausible models. To this end, the discovery system needs additional knowledge that constrains the model structures. We report on an extended system, SC-IPM, that uses such information to reduce its search through the space of candidates and to produce models that human scientists find more plausible. We also argue that although people carry out less extensive search than SC-IPM, they rely on the same forms of knowledge--processes and constraints--when constructing explanatory models.

Equation discovery for systems biology: finding the structure and dynamics of biological networks from time course data

Reconstructing biological networks, such as metabolic and signaling networks, is at the heart of systems biology. Although many approaches exist for reconstructing network structure, few approaches recover the full dynamic behavior of a network. We survey such approaches that originate from computational scientific discovery, a subfield of machine learning. These take as input measured time course data, as well as existing domain knowledge, such as partial knowledge of the network structure. We demonstrate the use of these approaches on illustrative tasks of finding the complete dynamics of biological networks, which include examples of rediscovering known networks and their dynamics, as well as examples of proposing models for unknown networks.

A multi-layered approach to protein data integration for diabetes research

OBJECTIVE

Recent advances in high-throughput experimental techniques have enabled many protein-protein interactions to be identified and stored in large databases. Understanding protein interactions is fundamental to the advancement of science and medical knowledge, unfortunately the scale of the requires an automated approach to analysis. We describe our graph-mining techniques to identify important structures within protein-protein interaction networks to aid in human comprehension and computerised analysis.

METHODS AND MATERIALS

We describe our techniques for characterizing graph type and associated properties which is constructed from data collated from the Human Protein Reference Database. Using random graph rewiring comparative techniques and cross-validation with other identification methods a further analysis of the identified essential proteins is presented to illustrate the accuracy of these measures. We argue for using techniques based upon graph structure for separating and encapsulating proteins based upon functionality.

RESULTS

We demonstrate how rational Erdos numbers may be used as a method to identify collaborating proteins based solely upon network structure. Further, by using dynamic cut-off limit it demonstrates how collaboration subgraphs can be generated for each protein within the network, and how graph containment can be used as a means of identifying which of many possible graphs are likely to be actual protein complexes. The demonstration protein interaction network built for diabetes is found to be a scale-free, small-world graph with a power-law degree distribution of interactions on nodes. These findings are consistent with many other protein interaction networks.

Biomedical Communications: Collaborative Research in Scientific Visualization, Online Learning, and Knowledge Translation

Collaborative research in biomedical communications investigates the role of visual media in scientific discovery and in patient and health professional education. The spectrum of work is broad and includes the visualization of scientific knowledge and simulation of hypothetical models of health and disease, as well as the design of audience-centered interactive visual media. The work cited supports the notion that research-based visual media can contribute to the core missions of science: discovery, communication, collaboration, and education.

Qualitative analysis of the relation between DNA microarray data and behavioral models of regulation networks

We introduce a mathematical framework that allows to test the compatibility between differential data and knowledge on genetic and metabolic interactions. Within this framework, a behavioral model is represented by a labeled oriented interaction graph; its predictions can be compared to experimental data. The comparison is qualitative and relies on a system of linear qualitative equations derived from the interaction graph. We show how to partially solve the qualitative system, how to identify incompatibilities between the model and the data, and how to detect competitions in the biological processes that are modeled. This approach can be used for the analysis of transcriptomic, metabolic or proteomic data.