KinDA: Kinetic DNA strand displacement analyzer

This package provides a framework for analyzing the kinetic behavior of domain-level strand displacement (DSD) reaction networks with nucleotide sequences assigned to each domain. KinDA optionally performs domain-level reaction enumeration using the Peppercorn enumerator (https://github.com/DNA-and-Natural-Algorithms-Group/peppercornenumerator). Reaction enumeration may be skipped if detailed and condensed reactions are given directly to KinDA. Thermodynamic and kinetic statistics about the behavior of resting macrostates and condensed reactions are collected using Multistrand (https://github.com/DNA-and-Natural-Algorithms-Group/multistrand), and may be computed at a desired level of precision using KinDA.

The principles underlying KinDA are introduced in the paper:

"Automated sequence-level analysis of kinetics and thermodynamics for domain-level DNA strand-displacement systems",
Joseph Berleant, Christopher Berlind, Stefan Badelt, Frits Dannenberg, Joseph Schaeffer and Erik Winfree.
Journal of The Royal Society Interface, 2018
(https://royalsocietypublishing.org/doi/full/10.1098/rsif.2018.0107).

General questions and comments should be addressed to Erik Winfree winfree@caltech.edu. For software issues regarding the most recent version, please contact the current maintainer Boyan Beronov beronov@cs.ubc.ca.

Trying out KinDA

Public AWS Image

The easiest way to test out KinDA is through the publicly available Amazon Web Services (AWS) Amazon Machine Image (AMI). This image is available to all AWS users, and can be found in the "Community AMIs" section when creating a new EC2 instance, using the search query "KinDA v0.2". The scripts should run on a "t2.micro" instance, but we often use "c5.9xlarge" instances for serious simulations. matplotlib is installed with the Agg backend default, so it can output files (PDF, etc) but not produce graphics interactively.

A note about AWS regions: The KinDA AMI is currently only available in AWS's four U.S. subdivisions. If you are having trouble finding this AMI, your AWS region may be set outside the U.S. Please contact the project team if you cannot switch your account's region setting and would like us to copy the image to a new region.

Local installation

The following instructions are intended for users wishing to setup KinDA on their own machine. This should not be necessary if using the public AWS AMI, which has KinDA 0.2 and all of its dependencies pre-installed.

KinDA 0.3+ runs on Python 3.10+, and requires a manual installation of the following packages:

• NUPACK 4.0.1+ (http://www.nupack.org)
• Multistrand 2.2+ (https://github.com/DNA-and-Natural-Algorithms-Group/multistrand)

Given the above, you can simply execute

$ pip install .

or

$ pip install -e .

in the root directoy. This will also install all remaining dependencies from the Python Package Index, including:

• Peppercorn enumerator 1.1+ (https://github.com/DNA-and-Natural-Algorithms-Group/peppercornenumerator)

Apptainer container

Alternatively, a fully reproducible, isolated and relocatable installation can be achieved using containerization.

1. Follow the instructions for building a Multistrand container (https://github.com/DNA-and-Natural-Algorithms-Group/multistrand?tab=readme-ov-file#apptainer-container).
2. Build a KinDA container layer on top, see scripts/kinda.def.

Getting Started

Gentle introduction

Input format

In order to use KinDA for analyzing reaction statistics, one has to first create a kinda.System object, which has several functions:

1. it describes the sequences, strands and complexes in a DNA strand-displacement system,
2. it provides convenient access to the reactions and resting sets of such a system, and
3. it holds the corresponding Stats objects.

The user needs to provide (1), while (2) and (3) are computed by KinDA. In particular, there are several ways of creating a reaction system description:

• Directly build up the Python objects defined in the kinda.objects subpackage, as demonstrated in example/analyze.py::create_system().
• Import from an old-style PIL file, as shown in example/analyze.py::import_system().
• Import from related Python packages such as Peppercorn or Multistrand, using the corresponding modules kinda.objects.io_*.

Analysis

The script examples/analyze.py shows how to query basic data for a system described by a PIL file. For a comprehensive example, run the following from within the examples directory, which will take a few hours on an AWS t2.micro instance, and proportionally less time on a faster multiprocessor instance.

$ python -i analyze.py Zhang_etal_Science2007.pil

If you just want to see a script run, but don't have much time, try the (still not so fast) simple toehold-mediated strand displacement example below, which should take less than 100 sec of wall-clock time on a machine with 16 cores.

$ python -i analyze.py simple.pil

You can make all this quicker (or slower) by changing the accuracy target (relative error) and sampling budget (number of Nupack samples / Multistrand trajectories) at the following lines in analyze.py. The comments are hopefully self-explanatory.

rel_error = 0.25
max_sims = 1000

Either way, running these scripts with the python -i flag will dump you into the Python shell at the end, where you can examine your data further, or look at the KinDA documentation, e.g.,

help(rxn_stats)

A deeper dive: The case studies

The directory case_studies contains scripts used to run the simulations described in the paper. You should take a look at the scripts themselves, as instructions for how to configure and run them are often included near the top.

Note that the simulations for Figure 9 were performed using a commandline interface to KinDA (placed in the src/kinda/scripts directory) that is explained in the Figure 9 directory's README.md.

Also note that some scripts produce plots using matplotlib, which on AWS should be able to produce PDF output files, but you won't be able to look at graphics interactively.

In the Amazon AMI, but not the GitHub repository, each case study directory has a subdirectory publication_data with KinDA data files that were generated for the paper. If you are not using the AMI, this data can be obtained from (http://www.dna.caltech.edu/SupplementaryMaterial/KinDA_paper_data/).

Configuration

The file src/kinda/options.py contains optional arguments that may be modified to change the default behavior of Multistrand, Peppercorn, NUPACK, and KinDA. This file must be modified prior to installation to set the default behavior, unless the installation is in editable mode (via pip install -e).

src/kinda/options.py contains four dict objects:

• kinda_params: General parameters for KinDA and its interactions with Multistrand, NUPACK, and Peppercorn.
• multistrand_params: Parameters for Multistrand, used when constructing a multistrand.Options() object.
• nupack_params: Parameters for NUPACK, given directly to NUPACK when calling its sample executable.
• peppercorn_params: Parameters for Peppercorn enumeration, given to the enumerator just prior to enumeration.

The initialization function for the kinda.System() object accepts an optional keyword argument for each of these dicts at runtime. Each keyword argument should be supplied as a dict object, whose key-value pairs will override the defaults.

Version History

0.3 (August 2023)

• Migrated to Python 3.10+ and updated the Python package definition.
• Updated dependencies: Multistrand 2.2, NUPACK 4.0.1, Peppercorn enumerator 1.1, DSDobjects 0.8.
• Created an Apptainer container definition for a fully reproducible installation, see scripts/kinda.def.
• Reworked the order semantics of the internal object hierarchy, leading to more consistent analysis outputs.
• Added an optional rate_model key to the multistrand_params configuration, which invokes a kinetic parameter preset in Multistrand instead of specifying parameters individually.
• Improved the reliability of simulation.*job by switching from the standard library module multiprocessing to the multiprocess package.
• Improved argument handling and output formatting in analysis scripts.

0.2 (April 2019)

0.1 (February 2018)

Contributors

Original authors

Joseph Berleant, Chris Berlind, Stefan Badelt, Frits Dannenberg, Joseph Schaeffer, and Erik Winfree

Current maintainer

Boyan Beronov