Towards an Optimal DNA-Templated Molecular Assembler

Jakob L. Andersen, Christoph Flamm, Martin M. Hanczyc, Daniel MerkleALIFE 2014 · DOI

TL;DR

This paper introduces a computational framework for designing and verifying DNA-templated synthesis, aiming to automate the creation of complex molecules. By combining graph grammars with strategy-based exploration, they demonstrate how to translate high-level synthesis plans into executable DNA programs, tackling the challenge of finding optimal assembly strategies and minimizing unwanted side products.

What Problem Does This Paper Address?

The "ideal synthesis" aims for simple, efficient, and environmentally friendly one-step production of complex molecules. While multi-step one-pot reactions approximate this ideal, designing them, especially with DNA-templated approaches, is complex. DNA-templated synthesis uses DNA strands to bring reactants together, enhancing reaction rates and specificity. However, for a given target molecule, numerous synthesis plans and corresponding DNA programs can exist. The core problem this paper addresses is how to automatically infer optimal low-level DNA programs from high-level synthesis plans or target molecules, ensuring correctness and minimizing side products. This involves navigating a vast chemical space and translating chemical logic into a computational framework.

Methods

The authors propose a computational framework that integrates two key approaches:

  • Graph Grammars: This formalism is used to represent chemical transformations. Molecules and reactions are modeled as labeled graphs, and reactions are defined as graph rewrite rules (specifically using the Double Pushout approach). This allows for a formal and systematic way to describe chemical processes.
  • Strategy Framework: This framework, adapted from previous work on exploring chemical spaces, defines how to systematically apply graph grammar rules. Strategies are functions that transform a set of molecular graphs (the "state" of the chemical space) by applying reaction rules. This allows for the exploration and generation of new chemical compounds and complexes.

The paper combines these by first defining a chemical space using graph grammars and then using the strategy framework to "compile" a high-level synthesis plan into a DNA program. This program is essentially a sequence of operations (adding monomers, instruction strands, release strands) that are executed computationally. The framework allows for the verification of these DNA programs by checking if the final state contains only the desired product, thus identifying side effects.

Datasets

Not discussed in the paper. The paper focuses on artificial chemical spaces and a real-world example from a prior publication (McKee et al., 2010).

Key Insights and Ideas

  • Bridging High-Level Plans and Low-Level Execution: The paper demonstrates a concrete computational approach to translate abstract chemical synthesis goals into concrete, executable DNA programs. This is a crucial step towards an "organic synthesis compiler."
  • The Non-Triviality of Optimality: Finding an "optimal" DNA assembler program is not straightforward. The number of DNA tags used significantly impacts the program's length and complexity. Reducing the number of tags can lead to more complex programs and potential side effects, highlighting a trade-off.
  • Graph Grammars as a Unified Framework: Using graph grammars for both chemical space representation and DNA program execution provides a consistent and powerful formalism for handling complex chemical transformations and interactions.
  • Verification as a Core Component: The framework inherently supports the verification of DNA programs, allowing researchers to identify and troubleshoot unwanted side products, which is critical for reliable synthesis.

Key Results

The paper presents three main examples to illustrate their framework:

  1. Artificial Synthesis (Four Reactants): A straightforward, side-effect-free DNA program was generated for a simple synthesis plan using four distinct DNA tags. This demonstrates the basic functionality of the framework.
  2. Real-World Synthesis (McKee et al., 2010): The framework was used to verify the correctness of a previously published multi-step DNA-templated synthesis of an oligomer. The program successfully reproduced the side-effect-free synthesis of the target molecule.
  3. Optimality Exploration (Four Reactants, Two Adapters): This example highlights the challenge of optimizing the number of DNA tags. A "naïve" approach with fewer tags resulted in unwanted side products, while a "correct" approach required a more complex, 9-step program to ensure purity. This demonstrates the trade-off between tag count and program complexity.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the current framework focuses on verification and generation of specific programs, rather than fully automatic inference of optimal synthesis plans from scratch.
  • The combinatorial explosion of possible DNA-tagged compounds and complexes is a significant challenge, even for toy examples.
  • Future Work:
    • Automatic Inference of Optimal Programs: A major future direction is developing methods to automatically infer optimal DNA assembly programs for a given synthesis plan or target molecule.
  • Multi-Objective Optimization: The authors suggest that optimality in the synthesis plan and optimality in the DNA assembler program might be competing qualities, requiring a multi-objective optimization approach.
  • Expanding Instruction Set: Exploring more complex DNA strand interactions, such as hairpin structures, could potentially reduce the number of required DNA tags and simplify programs.

Summary updated

BibTeX
@inproceedings{andersen2014towards,
  author = {Jakob L. Andersen and Christoph Flamm and Martin M. Hanczyc and Daniel Merkle},
  title = {{Towards an Optimal DNA-Templated Molecular Assembler}},
  booktitle = {ALIFE},
  year = {2014},
  doi = {10.7551/978-0-262-32621-6-ch090},
}
Figure