Advancing theory and tools for molecular programming

点击翻译按钮获取中文摘要

Molecular programming is the process of designing information-carrying molecules that react, or execute instructions, in purposeful ways. Nucleic acids - DNA and RNA molecules - are interesting to program because of their digital sequences and dynamic structural properties, and because they extend the reach of computational devices by naturally interacting with "wet" biological systems. Researchers in the field of molecular programming are tackling exciting engineering challenges such as building logic circuits, nanoscale structures and even molecular walking robots, from DNA and RNA sequences. In the coming decades, these technologies can offer significant payoffs, in transforming the ways in which we monitor and mediate molecular dynamics within the cell, or in developing "smart drugs" that could assess concentrations of disease indicators and respond accordingly. My research aims to uncover new and better ways to program molecules, as well as limits to molecular programming and energy-efficient computation, through the study of theoretical and empirical models. A primary goal of my research will be to address a significant barrier currently to design and analysis of molecular programs, namely the lack of good biophysical models. Particularly lacking are kinetics models that can accurately predict the rates at which nucleic strands interact through base pair formation and breakage. There is an opportunity now to make progress, by learning model parameters using valuable data from wet lab experiments of molecular programmers. This work will in part involve the creation of a database of experimental nucleic kinetics data, the first of its kind. In addition, I aim to further improve current thermodynamics models by accounting for important energy features (such as coaxial stacking and asymmetric multi-loop penalties) that are not easily incorporated into traditional dynamic programming secondary structure prediction algorithms. These contributions will support molecular programmers in scaling up current designs from hundreds to thousands of interacting strands, as well as biologists who wish to understand RNA conformational dynamics. I also want to understand the fundamental capabilities and limitations of molecular programming by studying models at higher levels of abstraction, namely chemical reaction networks (CRNs) and variants. CRNs provide a natural means for specifying and reasoning about highly distributed, asynchronous molecular programs, and a beautiful theory of what can and cannot be programmed with molecules is emerging from study of these models. I plan to contribute to this theory by developing provably fast algorithms for key problems such as leader election, and by developing models analogous to CRNs that can be useful when designing and reasoning about programs that are contained in single strands.