Hands-on examples for the application of IntaRNA
In the following, we list some examples how to use IntaRNA in specific application scenarios. The examples are presented in detail in our publication
- How to do RNA-RNA interaction prediction? A use-case driven
handbook using IntaRNA
- Martin Raden and Milad Miladi
- In: Lorenz, R. (eds) RNA Folding. Methods in Molecular Biology, vol 2726. Humana, New York, NY. https://doi.org/10.1007/978-1-0716-3519-3_9
Example 2.1: Local Installation via conda (Linux System with bash Shell)
First, we need to install conda (if not already available).
To that end, we assume you use a Linux-like system with a bash command line shell.
On Microsoft Windows, you can emply the Windows Subsystem for Linux (WSL) for that purpose.
# download Miniconda installer
wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda.sh
# note: "-b" runs the installer in non-interactive silent mode (agreeing to licence etc.)
# starting conda installation process
bash ~/miniconda.sh -b -p $HOME/miniconda
Everytime when starting a new shell (and thus before running IntaRNA), we need to ensure conda is among the tools we can run, i.e. its tools are among the program search path.
You can simplify this, by adding the export line to your .bashrc script in your home directory.
# add conda to your program search path
export PATH="$HOME/miniconda/bin:$PATH"
# installing IntaRNA via conda
conda install -c conda-forge -c bioconda "intarna>3" python=3.10 # python needed for viennarna package :/
# checking installed IntaRNA version
IntaRNA --version
It might be that your conda environment comes with a preinstalled python version that exceeds the dependency specifications of the viennarna package, a central dependency of IntaRNA.
In that case, you need to specify a python version that meets the requirements within the conda install call.
For example at the time of writing (Dec 2023), I used for conda 23.10.0 to install IntaRNA 3.3.2 with viennarna 2.5.1 the following call.
conda install -c conda-forge -c bioconda "intarna>3" python=3.10 # needed for viennarna 2.5.1
Example 3.1: Standard Input and Output
The bacterial sRNA OxyS of Echerichia coli is known to interact with the mRNA fhlA as shown by Argaman and Altuvia (2000).
From that publication, we extracted the sequences shown in Figure 7 provided here in the files fhlA.fasta and OxyS.fasta.
We use --tIdxPos0 to index the mRNA in relation to the position of its start codon.
# minimal call with with explicit fhlA index information
IntaRNA -t fhlA.fasta -q OxyS.fasta --tIdxPos0=-53
produces
fhlA|NC000913|-53..+60|doi:10.1006/jmbi.2000.3942:Fig7
-15 +7
| |
5'-AGU...GCUU A-- AA UG AUAC...GAC-3'
UCCUGGA C CAAAA UCAU
+++++++ | ||||| |||:
AGGACCU G GUUUU AGUG
3'-GCCU CUA GC CA CAAC...AAG-5'
| |
104 81
OxyS|NC_000913|56..164|doi:10.1006/jmbi.2000.3942:Fig7
interaction energy = -5.59 kcal/mol
Therein, base pairs denoted with + are part of putative seed regions (typically most reliable) while | and : represent AU,CG and GU base pairs, respectively.
Example 3.2: Detailed Output of Exact Prediction
To investigate the details of interactions, we suggest the IntaRNAexact personality.
IntaRNAexact --intLenMax=60 -t fhlA.fasta --tIdxPos0=-53 -q OxyS.fasta --outMode=D
--outMode=D provides more details for the interaction:
fhlA|NC000913|-53..+60|doi:10.1006/jmbi.2000.3942:Fig7
-15 -9
| |
5'-AGU...GCUU ACAA...GAC-3'
UCCUGGA
+++++++
AGGACCU
3'-GCCU CUAG...AAG-5'
| |
104 98
OxyS|NC_000913|56..164|doi:10.1006/jmbi.2000.3942:Fig7
interaction seq1 = -15..-9
interaction seq2 = 98..104
interaction energy = -5.57 kcal/mol
= E(init) = 4.1
+ E(loops) = -15.6
+ E(dangleLeft) = -0.4
+ E(dangleRight) = -0.96
+ E(endLeft) = 0.5
+ E(endRight) = 0.5
: E(hybrid) = -11.86
+ ED(seq1) = 0.62
: Pu(seq1) = 0.36569
+ ED(seq2) = 5.67
: Pu(seq2) = 0.000101064
seed seq1 = -15..-9
seed seq2 = 98..104
seed energy = -5.57
seed ED1 = 0.62
seed ED2 = 5.67
seed Pu1 = 0.36569
seed Pu2 = 0.000101064
It is also useful to study suboptimal interactions using -n:
IntaRNAexact --intLenMax=60 -t fhlA.fasta --tIdxPos0=-53 -q OxyS.fasta \
-n 10 --outMode=C --outCsvSort=E --outCsvCols="E,Pu2,subseqDB,hybridDB"
Note: \ at the line end will continue the command in the next line.
You can remove the \ along with the line break if needed.
E;Pu2;subseqDB;hybridDB
-5.57;0.000101064;-15UCCUGGA&98UCCAGGA;-15|||||||&98|||||||
-5.54;5.61948e-07;-15UCCUGGAACAACAAAAUGUCAU&81GUGAACUUUUGCGGAUCUCCAGGA;-15|||||||.|..|||||..||||&81||||..|||||..|...|||||||
-3.98;5.61948e-07;-15UCCUGGAACAACAAAAUGUCA&82UGAACUUUUGCGGAUCUCCAGGA;-15|||||||.|..|||||..|||&82|||..|||||..|...|||||||
-3.87;5.52904e-07;-15UCCUGGAACAACAAAAUGUCAUAU&79ACGUGAACUUUUGCGGAUCUCCAGGA;-15|||||||.|..|||||..||||.|&79|.||||..|||||..|...|||||||
-3.62;0.0167597;34CAAGGGU&24ACCCUUG;34|||||||&24|||||||
-3.42;6.6094e-07;-15UCCUGGAACAACAAAA&87UUUUGCGGAUCUCCAGGA;-15|||||||.|..|||||&87|||||..|...|||||||
-3.22;5.99629e-07;-15UCCUGGAACAACAAAAUGUC&83GAACUUUUGCGGAUCUCCAGGA;-15|||||||.|..|||||..||&83||..|||||..|...|||||||
-2.74;5.44005e-07;-15UCCUGGAACAACAAAAUGUCAUAUACAC&75GCCAACGUGAACUUUUGCGGAUCUCCAGGA;-15|||||||.|..|||||..||||.....|&75|.....||||..|||||..|...|||||||
-2.72;6.71751e-07;-15UCCUGGAACAACAAA&88UUUGCGGAUCUCCAGGA;-15|||||||.|..||||&88||||..|...|||||||
-2.7;6.6094e-07;-15UCCUGGAACAACAAAAUGU&85ACUUUUGCGGAUCUCCAGGA;-15|||||||.|..|||||.||&85|||||||..|...|||||||
The output reveals the second interaction site on rank 5 that was reported by Argaman and Altuvia (2000).
Example 3.3: Corrected Prediction of 2nd RRI in Two-site Concurrent Model
Given the first interaction site predicted in Example 3.2, we can predict an interaction model of two concurrently formed interaction sites using the following command.
IntaRNAexact --intLenMax=60 -t fhlA.fasta --tIdxPos0=-53 -q OxyS.fasta \
--tAccConstr="b:-15--9" --qAccConstr="b:98-104" --energyAdd="-5.57"
--energyAdd provides the interaction energy of the first site while the --*AccConstr accessibility constraints ensure that the regions involved in the first interaction are considered blocked for the prediction of the second interaction site.
Thus, IntaRNA predicts a concurrent second site (in accordance with literature) and reports the energy of the join interaction formation of both sites.
fhlA|NC000913|-53..+60|doi:10.1006/jmbi.2000.3942:Fig7
+34 +40
| |
5'-AGU...ACAA UGUU...GAC-3'
CAAGGGU
+++++++
GUUCCCA
3'-GCC...UGAA AUUU...AAG-5'
| |
30 24
OxyS|NC_000913|56..164|doi:10.1006/jmbi.2000.3942:Fig7
interaction energy = -9.19 kcal/mol
Example 3.4: Anchored Prediction
When specific base pairs of an interaction are known, e.g. from mutation experiments, this information can be provided to constrain the IntaRNA prediction. Here we use the interaction of the bacterial sRNA GcvB with its mRNA target phoB. Sequences GcvB.fasta and phoB.fasta were taken from the NCBI reference genome NC_000913.
IntaRNAexact -q GcvB.fasta -t phoB.fasta --tIdxPos0=-200 --intLenMax=60 -n 2
When using an informed standard prediction mode, the following interactions are predicted.
phoB|NC_000913|b1130|-200..+100|genom-subsequence
-56 -34
| |
5'-GAG...CCCC C AUC CUAU...GCC-3'
CAUAAC ACAUA GCGUUACA
||:||| |||:| ++++++++
GUGUUG UGUGU UGUAGUGU
3'-UUU...GUUU U --- UGGC...UCA-5'
| |
85 66
GcvB|NC_000913
interaction energy = -9.44 kcal/mol
phoB|NC_000913|b1130|-200..+100|genom-subsequence
-18 -11
| |
5'-GAG...AUUA AGAA...GCC-3'
AGACAGGG
++++++++
UCUGUCCC
3'-UUU...CCUG AUUU...UCA-5'
| |
159 152
GcvB|NC_000913
interaction energy = -9.34 kcal/mol
Here the known interaction from (Coornaer et al., 2013) is only second in rank. Incorporating the experimental information from the publication using
IntaRNAexact -q GcvB.fasta -t phoB.fasta --tIdxPos0=-200 --intLenMax=60 -n 2 \
--seedTQ="-17|||||&154|||||"
provides variants of the validated interaction site.
phoB|NC_000913|b1130|-200..+100|genom-subsequence
-18 -11
| |
5'-GAG...AUUA AGAA...GCC-3'
AGACAGGG
|+++++||
UCUGUCCC
3'-UUU...CCUG AUUU...UCA-5'
| |
159 152
GcvB|NC_000913
interaction energy = -9.34 kcal/mol
phoB|NC_000913|b1130|-200..+100|genom-subsequence
-17 -11
| |
5'-GAG...UUAA AGAA...GCC-3'
GACAGGG
+++++||
CUGUCCC
3'-UUU...CUGU AUUU...UCA-5'
| |
158 152
GcvB|NC_000913
interaction energy = -8.53 kcal/mol
Example 3.5: Seed-region-constraint Prediction
In case seed regions of RNAs, i.e. regions known to be involved in interactions, are known, their specification can provide a powerful tool to enhance the prediction quality. For instance, GcvB is known to feature three regions R1, R2 and R3 that can be involved in regulatory interactions.
Regions R1 and R2 correspond to the positions ranges 75-93 and 129-145 in the GcvB sequence of S. Typhimurium provided in GcvB.ST.fasta. When predicting interactions with its the regulatory region of its mRNA target ilvE provided in ilvE.fasta using
IntaRNAexact -q GcvB.ST.fasta -t ilvE.fasta --tIdxPos0=-200 --intLenMax=60
the top-ranked interaction is heavily disrupting local intra-molecular structure elements.
ilvE|NC_003197|STM3903|-200..+100|genom-subsequence
-56 -32
| |
5'-GGU...GAUC --- G C C AAAU...GGU-3'
UGC CA AGCG UGC ACAUCACAAC
||| || |:|: ::| ++++++++++
ACG GU UUGU GUG UGUAGUGUUG
3'-UUC...GUUA UUU G U U GCAU...UCA-5'
| |
91 64
GcvB|NC_003197
interaction energy = -8.32 kcal/mol
When restricting seed prediction to the known regions R1 and R2, using
IntaRNAexact -q GcvB.ST.fasta -t ilvE.fasta --tIdxPos0=-200 --intLenMax=60 \
--seedQRange="75-93,129-145"
the correct and experimentally verified interaction is predicted.
ilvE|NC_003197|STM3903|-200..+100|genom-subsequence
-37 -30
| |
5'-GGU...ACAU AUCC...GGU-3'
CACAACAA
++++++++
GUGUUGUU
3'-UUC...GUUU GUGU...UCA-5'
| |
85 78
GcvB|NC_003197
interaction energy = -6.38 kcal/mol
Example 3.6: Target Screen on Genomic Regions
Using the regulatory regions around all start codons of mRNAs in Echerichia coli provided in NC_000913.fa, we can do target prediction for the sRNA ChiX given in ChiX_NC_000913.fasta.
To prepare multiple target screens (not done here, but typically the case), we first generate and store the accessibility information of all putative targets.
# Precomputation of target accessibility via dummy call (pseudo query + no output)
IntaRNA -t NC_000913.fa --out=tAcc:NC_000913.ed -q AAA -n 0 --out=/dev/null --noSeed
Afterwards, we can do a fast target scan using
# Multi-threaded target screen using precomputed target accessibility data
IntaRNA -t NC_000913.fa -q ChiX_NC_000913.fasta \
--intLenMax=60 \
--outMode=C --outCsvCols="E,id1,start1,end1,id2,start2,end2" --outCsvSort=E \
--tAcc=E --tAccFile=NC_000913.ed \
--threads=0 \
> NC_000913.ChiX.csv
and the resulting top-5 predictions are
E;id1;start1;end1;id2;start2;end2
-66.23;b0481;20;78;ChiX_NC_000913;2;60
-16.41;b0681;182;193;ChiX_NC_000913;45;56
-15.45;b0619;164;175;ChiX_NC_000913;46;57
-12.9;b0352;157;192;ChiX_NC_000913;45;84
-12.51;b0847;224;253;ChiX_NC_000913;26;55
Note, b0481 is the gene encoding ChiX and thus highly complementary.
This shows that intra-molecular structure prediction and interaction prediction are quite related tasks.