*Bio-design automation*: software + biology + robots by +Douglas Densmore and +Swapnil Bhatia Synthetic biology promises to usher in a new era of scientific innovation and discovery [1]. Applications of this technology are broad and diverse [2]. Although applications frequently dominate the headlines, of equal importance are the engineering design principles. If these principles are developed rigorously, they will lay the foundation for the field and enable long-term growth and accessibility. A crucial engineering principle is design automation. Design automation is the process of applying tools (software, hardware, and wetware) to remove manual processes. Often, design automation transforms a high-level system objective ‘input’ (e.g., optimize this metabolic pathway) into a physically realized artifact ‘output’ (e.g., DNA, microbial strain, or protein). In order for design automation to be broadly applied it requires solutions to be based on sound definitions, tractable algorithms, and standardized data formats. Design automation promises to lower costs, increase design reuse, improve design reproducibility, reduce design error, and enable complex designs. Design automation has a rich, 50+ year history in the development of semiconductors, where it is termed electronic design automation (EDA). EDA historically evolved bottom up. Tools were coupled first to the physical fabrication process (1970s), then to design optimization and synthesis (1980s), and finally to design specification and verification (1990s+). This evolution allowed designer focus to shift from semiconductor physics to system optimization and verification. The ability of EDA to allow for this shift was crucial to support the additional computing resources provided by Moores Law. If carefully developed to respect the unique differences of biology from electronics, aspects of this mature discipline can be applied to biological design [3]. This article outlines how specification (how a system should behave), design (what components should constitute a system), assembly (how to put the system together in the physical world), and data management (how to track system information electronically) serve as four distinct research challenges in the emerging bio-design automation (BDA) discipline. A crucial EDA concept is the separation of concerns [4]. Specifications should separate behavioral requirements (e.g., what functions a system has to perform) from performance requirements (e.g., how the system is measured while performing those functions). In addition, the function behavior (e.g., what the system does) should be separated from the structure (e.g., how the system does it). Developing synthetic biological specification languages with these explicit separation abilities will allow modular, reusable descriptions applicable across host organisms, environmental context, and industrial processes [5,6,7]. These languages should provide: (i) formalized behavior and performance specifications allowing for nontrivial relations among design components and multidimensional objective functions; (ii) formalized constraint mechanisms regarding function, structure, and performance requirements of a design; and (iii) executable semantics allowing for the simulation of system behavior and the capability to produce derivative designs adhering to the system constraints. A BDA framework must have a mechanism for automatically transforming a design specification into a representation that can be physically manufactured. This process needs to explore simultaneously the space of valid designs [8] while optimizing those designs [9]. In particular: (i) libraries of genetic elements (e.g., parts) must be assigned to functional requirements respecting the performance characterization of the elements and their compositional behavior; (ii) algorithms should account for multiple genetic elements that map to the same functionality but differ in performance or cost, and should encode sound rules for making expert design decisions; and (iii) a framework for genetic element interactions must present a model that can capture, distinguish, and predict intended and unintended consequences of genetic composition. The manual manufacturing of synthetic biological designs can lead to human error, wasted reagents and consumables, low design throughput, development of nontransferable laboratory-specific techniques, and selective recording of data. An automated genetic manufacturing process [10,11] will provide: (i) assembly strategies formulated to share design intermediates, reduce overall design stages, allow for manufacturing restarts after failure, account for known biological phenomenon (DNA homology for example), and recognize library elements available to replace de novo synthesis; (ii) translation of assembly strategies into detailed, scheduled protocol execution plans, taking into account physical resources available; and (iii) explicit separation between the operations in a protocol and the commands to manufacturing hardware (e.g., liquid handling robotics or microfluidics). Automation relies on making decisions based on quantitative and empirical data. These data must be accurate and capture biological relations. Its storage must be persistent, allow for various query/retrieval methods, and expand with the communitys understanding of biological phenomena. Frameworks that support a flexible data model while providing a programmatic interface to the data can provide a powerful solution [12]. Once curated, such data can allow: (i) machine learning algorithms to create models describing biological relations, requirements, and constraints; (ii) automated redesign of biological systems based on empirical data; and (iii) standardization of design exchange and characterization. Each of these challenge areas can be related in a cyclic design flow, as shown in Figure 1. This ecosystem will allow for a variety of design start points depending on the application and expertise of the designer. Moreover this ecosystem paves the way for a commercialization environment where vertical integration allows for end-to-end solutions and horizontal integration allows for customized solutions within any given design challenge area. The modular nature of the ecosystem with well-defined interfaces will allow for solutions to be developed in isolation and improved upon over time without disrupting the overall design process.... bit.ly/17XNEPX
Posted on: Wed, 27 Nov 2013 19:22:15 +0000
Trending Topics
Recently Viewed Topics
© 2015