Physical Design of Biological Systems

摘要

It may seem premature to worry about the design of of livingrnsystems when we do not yet fully understand their operation.rnHowever, the path to understanding is clear, and design willrninevitably follow. It’s not too early to think about how thisrnmight be structured.rnBiological systems, as opposed to the systems consideredrnat ISPD so far, are grown rather than constructed. This processrnis not yet understood in detail but work is in progress,rnwith full understanding to follow. Living creatures are builtrnby a combination of discrete events (cell division) driven byrnanalog variables (concentrations and gradients). Every cellrnin the adult organism has a lineage, a path back to the originalrnegg cell. This is currently under active investigation inrnthe case of the fruit fly, and this talk will describe this effort,rnthe tools that are used, and some of the results so far.rnAfter construction, the operation of the organism is drivenrnby continued chemical signalling, and in the case of animals,rnfaster operations mediated by the nervous system. Here toornunderstanding is coming slowly, driven by studies of modelrnorganisms such as the fruit fly and rodents. All of these arernunder active investigation, and this talk will give a quickrnsummary of progress to date.rnThe partner of analysis is design - creating a system withrncertain desirable features, or lack of particular drawbacks.rnHistorically, design of living systems has been done by crossbreeding.rnThough developed empirically before recordedrnhistory, this algorithm resembles simulated annealing or (notrnsurprisingly) genetic algorithms. Although there have beenrnsome notable successes, this method is limited to local configurationrnchanges, and by the long time scales required.rnHow can design be made faster and more specific? To dornthis, we need both physical tools that can implement desiredrnchanges, and understanding to know what changes to implement.rnThe physical tools needed to create new capabilitiesrnare being developed at both at the level of DNA and genesrn(corresponding very roughly to the level of transistors andrngates or functional units.) This work is well along and is notrnlikely to be the limiting step.rnHowever, using these technologies to achieve desired goalsrnwill require new physical design tools - first to predict thernresults of changes (think of SPICE and logic simulators), tornsuggest ways to achieve desired goals (think logic synthesis)rnand optimize the end result (think design for testability, designrnfor manufacturabilty, and and design centering). Thisrnis a huge opportunity for physical design tools at all levels.rnSome of this is already underway. Industrial scale productionrnof medicine and chemicals currently use relatively minorrnmodifications to cell pathways that already exist. Tools tornunderstand and optimize these complex metabolic networksrnare under active construction and use. At the current level ofrnunderstanding this usually works by enhancing or blockingrnpaths within the existing system.rnAt the next level, synthetic biologists seek to constructrnnovel systems, but using the same basic architecture of livingrnthings. Although only working for systems of a few variablesrnat present, this holds the potential of creating behaviors thatrnare completely independent of those that already exist.rnAt the third level, it must be possible to specify and constructrnneural systems with specified behaviors. Nature doesrnthis, where the same construction methods create the behaviorrnof honeybees and elephants. Here, design tools arerna long ways off since how nervous systems are constructedrnis one of the biggest mysteries of biology. In the broadestrnoutline, the instructions that tell how to build such a systemrnmust be composed of at least two parts - a part that tellsrnroughly which units must be connected, and then rules thatrntell how the detailed connections are made once the generalrnoutline is done. Here the current physical design challenge isrncreating tools to help us understand how this works. Toolsrnto simulate, optimize, or create these nervous systems willrnbe one of greatest opportunities in future physical design.rnFinally, one wonderful attribute of biological systems isrnthat all but the simplest learn and adapt, and are enormouslyrnrobust to minor perturbations and malfunctions ofrnconstituent components. Once again, design tools are prematurernbut the opportunities are enormous.rnIn short, those who believe we are near the end of the roadrnin physical design are thinking only of existing techniques forrnbuilding electronics. In the larger view, the opportunitiesrnahead vastly exceed what has been done already.