Designing RNA Sequences With Azoarcus-like Secondary Structures

This is an implementation of a pipeline for sequence design given a target structure with simple pseudoknots, without explicitly modelling pseudoknots in the design process.

Prerequisites

Besides packages installable via pip, the following is required:

ViennaRNA (Python bindings and RNAPKplex)
pKiss
RNAblueprint
the proof-of-concept implementation libbpdist

Caveats

Although the commandline options of the scripts are documented and accessible via -h, usage is not always ergonomic.

Usage & Structure

Design Pipeline

The design pipeline is implemented in azoarcus_design.py (and pipeline.py). Per default, its output is in FASTA format printed to stdout. To compute properties of designed sequences, use -i, specify reference data via -S, -T.

Constrained Approach

This is the default design approach. Example usage:

python3 azoarcus_design.py -n 10 -j 4 -s 0.05 -T <target structure> -C <sequence constraints> > sequencedesigns.fasta
python3 azoarcus_design.py -j 4 -s -i sequencedesigns.fasta -T <target structure> -S <reference sequence> > sequencedesigns.csv

-s 0.05 adjusts the stopping threshold for the objective function. Structural constraints are extracted from pseudoknots in the target structure; those should be reflected in sequence constraints specified via -C.

Example input data:

GUGCCUUGCGCCGGGAAACCACGCAAGGGAUGGUGUCAAAUUCGGCGAAACCUAAGCGCCCGCCCGGGCGUAUGGCAACGCCGAGCCAAGCUUCGGCGCCUGCGCCGAUGAAGGUGUAGAGACUAGACGGCACCCACCUAAGGCAAACGCUAUGGUGAAGGCAUAGUCCAGGGAGUGGCGAAAGUCACACAAACCGG # reference sequence
...(((((((..((....)).)))))))...((((((....((((((...((...((((((....))))))..))...))))))(((...(.((((((....)))))).)..)))...[.[[[[[...))))))((((...(((....)))..))))......]]]]]]..((.(((((....))))).....)).. # target structure
GUGNCNNNNNNNNNGAAANNNNNNNNGNNANNNNNNCNAAUNCGNCNNNNNCUAAGNNNNNNNNNNNNNNUAUGNNNNNGNCGNNCCANNNNNNNNNNNNNNNNNNNNNNNNGGNGUAGAGACUANNNGNNNNNNNNCUAAGNNNNNNNNUAUGNNNNNNNCAUAGUCCNNNNNNNNNNGAAANNNNNNNNNNNNNG # complete sequence constraints
GUGNNNNNNNNNNNGAAANNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNGAGACUANNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNUAGUCCNNNNNNNNNNNNNNNNNNNNNNNNNNNG # minimal sequence constraints

Alternative Approach

The alternative approach does not require structural constraints but two target structures obtained by splitting a pseudoknotted structure. This variant of the pipeline is toggled using -P. Example usage:

python3 azoarcus_design.py -P -n 10 -j 4 -s 0.15 -T <first split structure> -T <second split structure> -C <sequence constraints> > sequencedesigns.fasta
python3 azoarcus_design.py -P -j 4 -s -i sequencedesigns.fasta -T <original target structure> -T <first split structure> -T <second split structure> -S <reference sequence> > sequencedesigns.csv

-s 0.15 is a higher threshold than before (the reason is illustrated in maxned_threshold.py). Structural constraints are extracted from pseudoknots in the target structure; those should be reflected in sequence constraints specified via -C.

Example input data:

GUGCCUUGCGCCGGGAAACCACGCAAGGGAUGGUGUCAAAUUCGGCGAAACCUAAGCGCCCGCCCGGGCGUAUGGCAACGCCGAGCCAAGCUUCGGCGCCUGCGCCGAUGAAGGUGUAGAGACUAGACGGCACCCACCUAAGGCAAACGCUAUGGUGAAGGCAUAGUCCAGGGAGUGGCGAAAGUCACACAAACCGG # reference sequence
...(((((((..((....)).)))))))...((((((....((((((...((...((((((....))))))..))...))))))(((...(.((((((....)))))).)..)))...[.[[[[[...))))))((((...(((....)))..))))......]]]]]]..((.(((((....))))).....)).. # target structure
...(((((((..((....)).)))))))...((((((....((((((...((...((((((....))))))..))...))))))(((...(.((((((....)))))).)..))).............))))))((((...(((....)))..))))..............((.(((((....))))).....)).. # first split structure
...(((((((..((....)).))))))).............((((((...((...((((((....))))))..))...))))))(((...(.((((((....)))))).)..)))...(.(((((.........((((...(((....)))..))))......))))))..((.(((((....))))).....)).. # second split structure
GUGNCNNNNNNNNNGAAANNNNNNNNGNNANNNNNNCNAAUNCGNCNNNNNCUAAGNNNNNNNNNNNNNNUAUGNNNNNGNCGNNCCANNNNNNNNNNNNNNNNNNNNNNNNGGNGUANNNNNNNNNNGNNNNNNNNCUAAGNNNNNNNNUAUGNNNNNNNCANNNNNNNNNNNNNNNNGAAANNNNNNNNNNNNNG # complete-alt sequence constraints
GUGNNNNNNNNNNNGAAANNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNG # minimal-alt sequence constraints

Neutral Paths

neutralpaths.py computes neutral path lengths using pKiss structure prediction.

Although not required for neutral paths, the option-T computes the expected Hamming distance between two random sequences compatible to a shared structure, by both uniform sampling and an estimation using the number of paired and unpaired positions.

Energy Parameter Sets

eparams.py may be used to assess different energy parameter sets (in format compatible to ViennaRNA) using RNAfold, pKissand RNAPKplex. Note that the parameter files are not provided here but are available in ViennaRNA and here (requires conversion using RNAparconv). For this, pKiss was used at version 2.2.12 and RNAPKplex was patched (as of ViennaRNA 2.4.18, RNAPKplex seems to work better).

Dot Plots

dotplot.py was used to produce dotplots of base pair probability matrices and single secondary structures.

fncnt/azodesign