RIkin

Rikin implements tools for the fast computation of RNA-RNA interaction kinetics in a detailed RNAup/Intarna-inspired interaction model.

It features accurate modeling of RNA structures and interaction complexes based on their secondary structure and the full Turner nearest neighbor energy model.

Example

Our methods studies the dynamics of the interaction of two RNAs, given as sequences, e.g.

AAAGGGGGGAAAAAAAGGGUGGGAAAAAAAGGGCGGGAAA

and

CCCGCCC

Figure: Kinetics of RNAs can be analyzed and visualized in many ways. The most direct result of the analysis are predicted probabilities over time of single interaction states. These probability progressions can be visualized for the most prominent states.

Methods

The prediction of RNA kinetics is computationally challenging, and the dynamics of RNA-RNA interaction is even substantially more demanding, due to the huge space of possible conformations.

Rikin takes a series of measures to tackle the computational problems:

• RNA interactions are studied at the level of secondary structure with elementary step conformation changes (transitions).

• Secondary structure interaction states are abstracted as RNAup/Intarna like states that each represent an ensemble of interactions. The induced transitions between these states are directly computed by efficient algorithms.

• Sparsification techniques and constraints are applied to further restrict the size of the computational problem in controlled ways.

• The energy landscape of these states is further coarse-grained in two steps: a fast discrete coarse graining is followed by a novel continuous coarse-graining.

• The Master Equation of the corresponding Markov process is solved by the matrix exponentiation method of Padé.

Installation

Installation from Conda package

We recommend to install from the conda package of rikin, using mamba.

Possible create and activate an environment for rikin first:

conda create rikin
conda activate rikin

Then, install by

mamba install -c bioconda rikin

or alternatively conda

conda install -c bioconda rikin

Then, activate the environment

conda activate rikin

and install required R pacakges

Rscript -e "install.packages(c('argparse','shape','RColorBrewer'))"

Compilation/installation from the source repository

The tools can be compiled and installed after cloning the source repository. This requires a build toolchain with C++ compiler and autotools. We describe the installation in a conda environment (and get further specific dependencies from bioconda).

mamba create -n rikin -c bioconda -c conda-forge viennarna locarna r-base

conda activate rikin

autoreconf -i
./configure --prefix=$CONDA_PREFIX PKG_CONFIG_PATH=$CONDA_PREFIX/lib/pkgconfig
make
make install

Finally, see above (installation from conda package) how to install the required R packages.

Usage

RIkin computes kinetics of RNA-RNA interaction in several stages. The complete pipeline consists of the stages

• state enumeration (rikin_enumerate)
• sorting of states (sort)
• first coarse graining (rikin_barriers)
• second coarse graining (rikin_prune)
• solving of the master equation (rikin_xrates.m)
• plotting (rikin_kinetics.R)

We provide the script rikin_pipeline.sh to perform all these pipeline stages in coordination for a pair of given RNAs.

For example

rikin_pipeline.sh -j example \
CGGAGCGACGCUACGUACGGAGCUAGCUGAAACGUAGCAGAGCUAGCUGAAACGUAGCAGAGCUAGCUGAAACGUAGCAGACAGAUACUAGUUUCAAAACUUCCUUGACAGAACCAUUUCAUGUUCAAUAAUGAAAAUUACUUUCACAUGUUUUAGUGGAAAACGUACGUACGUAUCGUAGCGGUUGGACUUACGUAUAC \
GCUGCGAUGCAUGCCUCGAU \
| tee example.out

predicts the kinetics of the interaction between the two given sequences using a set of default parameters. The results are written to subdirectory example and text output is redirected to example.out. In particular, we produce a plot of the dynamics of state probabities in example/example.pdf. This plot was already shown as example above.

All command line tools can be further configured and provide detailed help with options --help or -h.

Related publications

A manuscript describing RIkin is in preparation.

Authors and Contacts

• Rolf Backofen, University of Freiburg

• Sebastian Will, École Polytechnique