Our task is to investigate how synthetic biochemical systems can be designed to carry out algorithms and compute; what models of computation arise from biochemical processes and how they can be programmed; and how to "compile" abstract descriptions of biomolecular algorithms down to specific synthetic DNA sequences that implement the desired computation in the laboratory.
Like the carefully orchestrated molecular processes that occur within living cells, biomolecular computation can in principle occur autonomously, without the need for any external intervention during the computation. Being able to design and understand such systems is our ultimate goal. We are exploring several interconnected paradigms of biomolecular computation, based loosely on processes that are ubiquitous throughout living organisms:
See also non-technical discussion of our work: A description of our collaboration with Bernie Yurke, which started in 2003. A summary of a talk I gave at the University of Washington in 2007. An interview in Discover Magazine for another perspective on our research, circa 2009. An Economist article from 2012. A brief history of DNA computing circa 2015.
Algorithmic self-assembly of DNA tiles (inspired by crystals, microtubules, and virus capsids on the biological side, and by Wang tiles on the mathematical) encodes information in the geometric arrangement of tiles, and performs logical steps by the selective addition of tiles as geometrically compatible sites. Algorithmic self-assembly may be ideally suited for bottom-up self-fabrication of complex nanostructures. A major question concerns how to reduce error rates during assembly; we are investigating "proofreading" logic for error-resilient algorithmic growth, as well as methods to programmably control the nucleation of self-assembled structures. Both theoretical and experimental projects are ongoing.In vitro RNA transcriptional circuits are a stripped-down, bare-bones version of genetic regulatory networks in the cell; signals are carried by the concentration of specific RNA transcripts; RNA polymerase and RNase regulate the production and degradation of RNA. In vitro RNA transcriptional circuits should allow dynamic control of biomolecular processes -- at the time scale of minutes. On the theory side, we have shown how these networks can function as biochemical neural networks; on the experimental side, we have demonstrated and characterized a two-node bistable circuit. Future research aims at spatial patterning in reaction-diffusion conditions, and at measuring stochastic behavior due to small copy numbers in small volumes.
Biochemical circuits, such as cellular signal transduction cascades, are logically related to boolean circuits. For example, a given enzyme molecules may be either phosphorylated ("on") or not ("off"). Phosphorylation cascades are ideal for the study of reliable computation in the presence of thermal "noise". More generally, one may ask how to design formal chemical reaction networks to perform computation, and how stochastic noise is shaped by network activity. Experimentally, we are constructing DNA-based logic gates that can be "wired" into arbitrary circuits.
Chemical self-replication and evolution must have gotten started somehow, way back when. We are using algorithmic self-assembly to investigate a radical hypothesis of Graham Cairns-Smith, that life got started as clay crystals that reproduced patterns as they grew. On paper, at least, it appears that simple crystal growth mechanisms are sufficient for very complex Darwinian evolution.
RNA and DNA hybridization and folding are essential processes for all DNA computing, and can perform complex logical operations in their own right. Realistic yet tractable models of nucleic acid interactions form the foundation for higher-level descriptions of DNA nanodevices, and allow for automated design of DNA sequences for DNA structures and devices. We are developing fast simulation algorithms for simulating folding at the secondary structure level.
See our publications page for a detailed log of our progress in these directions.
Erik Winfree, 10/16/01, 3/17/06, 11/17/07, 2/5/15, 3/20/16