Authors: Sebastian Mackowiak & Marc Friedländer
This is miRDeep2 developed by Sebastian Mackowiak & Marc Friedländer. miRDeep2 discovers active known or novel miRNAs from deep sequencing data (Solexa/Illumina, 454, ...).
(minor edits to README, TUTORIAL, CHANGELOG, and FAQ, convertion to
Markdown, trailing whitespace removal & CI setup by Marcel Schilling)
Linux system, 2GB Ram, enough disk space dependent on your deep sequencing data
MacOSX with Xcode and gcc compiler installed. (This can be obtained from the appstore, if there are any issues with installing it please look for help online).
To compile the Vienna package it may be necessary to have GNU grep installed since the MacOSX grep is BSD based and sometimes not accepted by the installer. To get a GNU grep you could for example install homebrew by typing
ruby -e "$(curl -fsSL \
https://raw.githubusercontent.com/Homebrew/install/master/install)"(the link could be out of date, in that case look up online what to do)
After that typing
brew tap homebrew/dupes; brew install grepwill install GNU grep as ggrep in /usr/local/bin/
Type
perl install.plFollow the instructions given below
First download all necessary packages listed here
- bowtie short read aligner
- Vienna package with RNAfold
- SQUID library goto Squid and download it
- randfold
- Perl package PDF::API2
When packages are downloaded
- attach the miRDeep2 executable path to your PATH
echo 'export PATH=$PATH:your_path_to_mirdeep2/src' >> ~/.bashrc-
unzip bowtie-0.11.3-bin-linux-x86_64.zip -
put the bowtie directory into your
PATHvariable, e.g.
echo 'export PATH=$PATH:your_path_tobowtie' >> ~/.bashrc-
tar xvvzf ViennaRNA-1.8.4.tar.gz -
cdto the Vienna dir -
type
./configure --prefix=your_path_to_Vienna/install_dir
make
make install- add Vienna binaries to your
PATHvariable, e.g.
echo 'export PATH=$PATH:your_path_to_Vienna/install_dir/bin' >> ~/.bashrc-
tar xxvzf squid-1.9g.tar.gz -
tar xvvzf randfold-2.0.tar.gz -
cd randfold2.0 -
edit Makefile, e.g.
emacs Makefile:
change line with INCLUDE=-I. to
INCLUDE=-I. -I<your_path_to_squid-1.9g> -L<your_path_to_squid-1.9g>,
e.g. INCLUDE=-I. -I/home/Pattern/squid-1.9g/ -L/home/Pattern/squid-1.9g/
-
make -
add randfold to your
PATHvariable, e.g.
echo 'export PATH=$PATH:your_path_to_randfold' >> ~/.bashrc-
tar xvvzf PDF-API2-0.73.tar.gz -
cdto your PDF_API2 directory -
then type in
perl Makefile.PL INSTALL_BASE=your_path_to_miRDeep2 LIB=your_path_to_miRDeep2/lib
make
make test
make install- add your library to the
PERL5LIB, e.g.
echo \
'export PERL5LIB=PERL5LIB:your_path_to_miRDeep2/lib/perl5' \
>> ~/.bashrc-
cdto your mirdeep2 directory (the one containinginstall.pl) -
touch install_successful -
start a new shell session to apply the changes to environment variables
To test if everything is installed properly type in
bowtieRNAfold -hrandfoldmake_html.pl
You should not get any error messages. Otherwise something is not correctly installed.
Everything that is download by the installer will be in a directory called
<your_path_to_mirdeep2>/essentials
miRDeep2 analyses can be performed using the three scripts miRDeep2.pl,
mapper.pl and quantifier.pl.
Wrapper function for the miRDeep2.pl program package. The script runs all
necessary scripts of the miRDeep2 package to perform a microRNA detection deep
sequencing data anlysis.
- A FASTA file with deep sequencing reads,
- a FASTA file of the corresponding genome,
- a file of mapped reads to the genome in miRDeep2 ARF format,
- an optional FASTA file with known miRNAs of the analysed species, and
- an optional FASTA file of known miRNAs of related species.
- A spreadsheet and
- an HTML file
with an overview of all detected miRNAs in the deep sequencing input data.
| option | description |
|---|---|
‑a <int> |
minimum read stack height that triggers analysis. Using this option disables automatic estimation of the optimal value. |
‑b <int> |
minimum score cut-off for predicted novel miRNAs to be displayed in the overview table. This score cut-off is by default 0. |
‑c |
disable randfold analysis |
‑t <species> |
species being analyzed - this is used to link to the appropriate UCSC browser |
‑u |
output list of UCSC browser species that are supported and exit |
‑v |
remove directory with temporary files |
‑q <file> |
miRBase.mrd file from quantifier module to show miRBase miRNAs in data that were not scored by miRDeep2 |
The miRDeep2 module identifies known and novel miRNAs in deep sequencing data. The output of the mapper module can be directly plugged into the miRDeep2 module.
The user wishes to identify miRNAs in mouse deep sequencing data, using default
options.
The miRBase_mmu_v14.fa file contains all miRBase mature mouse miRNAs, while
the miRBase_rno_v14.fa file contains all the miRBase mature rat miRNAs.
The 2> will pipe all progress output to the report.log file.
miRDeep2.pl reads_collapsed.fa genome.fa reads_collapsed_vs_genome.arf \
miRBase_mmu_v14.fa miRBase_rno_v14.fa precursors_ref_this_species.fa \
-t Mouse 2>report.logThis command will generate
- a directory with PDFs showing the structures, read signatures and score breakdowns of novel and known miRNAs in the data,
- an HTML webpage that links to all results generated (
result.html), - a copy of the novel and known miRNAs contained in the webpage but in text
format which allows easy parsing (
result.csv), - a copy of the performance survey contained in the webpage but in text format
(
survey.csv), and - a copy of the miRNA read signatures contained in the PDFs but in text format
(
output.mrd).
The user wishes to identify miRNAs in deep sequencing data from an animal with no related species in miRBase:
miRDeep2.pl reads_collapsed.fa genome.fa reads_collapsed_vs_genome.arf \
none none none 2>report.logThis command will generate the same type of files as example use 1 above. Note that there it will in practice always improve miRDeep2 performance if miRNAs from some related species is input, even if it is not closely related.
Processes reads and/or maps them to the reference genome.
Default input is
- a file in FASTA,
seq.txtorqseq.txtformat.
More input can be given depending on the options used.
The output depends on the options used (see below).
Either
- a FASTA file with processed reads, or
- an ARF file with with mapped reads, or
- both
are output.
| option | description |
|---|---|
‑a |
input file is seq.txt format |
‑b |
input file is qseq.txt format |
‑c |
input file is FASTA format |
| option | description |
|---|---|
‑h |
parse to FASTA format |
‑i |
convert RNA to DNA alphabet (to map against genome) |
‑j |
remove all entries that have a sequence that contains letters other than a, c, g, t, u, n, A, C, G, T, U, or N. |
‑k <seq> |
clip 3' adapter sequence |
‑l <int> |
discard reads shorter than <int> nts |
‑m |
collapse reads |
‑p <genome> |
map to genome (must be indexed by bowtie-build). The genome string must be the prefix of the bowtie index. For instance, if the first indexed file is called h_sapiens_37_asm.1.ebwt then the prefix is h_sapiens_37_asm. |
‑q |
map with one mismatch in the seed (mapping takes longer) |
| option | description |
|---|---|
‑s file |
print processed reads to this file |
‑t file |
print read mappings to this file |
| option | description |
|---|---|
‑u |
do not remove directory with temporary files |
‑v |
outputs progress report |
The mapper module is designed as a tool to process deep sequencing reads and/or map them to the reference genome. The module works in sequence space, and can process or map data that is in sequence FASTA format. A number of the functions of the mapper module are implemented specifically with Solexa/Illumina data in mind. For example on how to post-process mappings in color space, see example use 5:
The user wishes to parse a file in qseq.txt format to FASTA format, convert
from RNA to DNA alphabet, remove entries with non-canonical letters (letters
other than a, c, g, t, u, n, A, C, G, T, U, or N), clip
adapters, discard reads shorter than 18 nts and collapse the reads:
mapper.pl reads_qseq.txt -b -h -i -j -k TCGTATGCCGTCTTCTGCTTGT -l 18 -m \
-s reads_collapsed.faThe user wishes to map a FASTA file against the reference genome.
The genome has already been indexed by bowtie-build.
The first of the indexed files is named genome.1.ebwt:
mapper.pl reads_collapsed.fa -c -p genome -t reads_collapsed_vs_genome.arfThe user wishes to process the reads as in example use 1 and map the reads as in example use 2 in a single step, while observing the progress:
mapper.pl reads_qseq.txt -b -h -i -j -k TCGTATGCCGTCTTCTGCTTGT -l 18 -m \
-p genome -s reads_collapsed.fa -t reads_collapsed_vs_genome.arf -vThe user wishes to parse a GEO file to FASTA format and process it as in example use 1. The GEO file is in tabular format, with the first column showing the sequence and the second column showing the read counts:
geo2fasta.pl GSM.txt > reads.fa
mapper.pl reads.fa -c -h -i -j -k TCGTATGCCGTCTTCTGCTTGT -l 18 -m \
-s reads_collapsed.faThe user has already removed 3' adapters in color space and has mapped the
reads against the genome using the BWA tool. The BWA output file is named
reads_vs_genome.sam. Notice that the BWA output contains extra fields that
are not required for SAM format. Our converter requires these fields and thus
may not work with all types of SAM files. The user wishes to generate
reads_collapsed.fa and reads_vs_genome.arf to input to miRDeep2:
bwa_sam_converter.pl reads_vs_genome.sam reads.fa reads_vs_genome.arf
mapper.pl reads.fa -c -i -j -l 18 -m -s reads_collapsed.faThe module maps the deep sequencing reads to predefined miRNA precursors and determines by that the expression of the corresponding miRNAs. First, the predefined mature miRNA sequences are mapped to the predefined precursors. Optionally, predefined star sequences can be mapped to the precursors too. By that the mature and star sequence in the precursors are determined. Second, the deep sequencing reads are mapped to the precursors. The number of reads falling into an interval 2 nt upstream and 5 nt downstream of the mature/star sequence is determined.
- A FASTA file with precursor sequences,
- a FASTA file with mature miRNA sequences,
- a FASTA file with deep sequencing reads, and
- optionally a FASTA file with star sequences and the 3 letter code of the species of interest.
- A 2 column table file called
miRNA_expressed.csvwith miRNA identifiers and its read count, - a file called
miRNA_not_expressed.csvwith all miRNAs having 0 read counts, - a signature file called
miRBase.mrd, - a file called
expression.htmlthat gives an overview of all miRNAs the input data, and - a directory called
pdfsthat contains for each miRNA a PDF file showing its signature and structure.
| option | description |
|---|---|
‑t |
list all values allowed for the species parameter that have an entry at UCSC |
quantifier.pl precursors.fa mature.fa reads.fa star.fa/none species/none \
timestamp/none pdfIt creates a file called result.html that gives an overview of miRDeep2
detected miRNAs (known and novel ones). The HTML file lists up each detected
miRNA and provides among others information on its miRDeep2 score, reads mapped
to its mature, loop and star sequence, the mature, star and consensus precursor
sequences themselves and provides links to BLAST, BLAT, mirBase for miRBase
miRNAs and to a PDF file that shows the signature and structure.
- A miRDeep2 output.mrd file and
- a miRDeep2 survey.csv file
- A
result.htmlfile with an entry for each provisional miRNA that contains information about its assigned Id, miRDeep2 score, estimated probability that the miRNA candidate is a true positive, rfam alert, total read count, mature read count, loop read count, star read count, significant randfold p-value, miRBase miRNA, example miRBase miRNA with the same seed, BLAT, BLAST, consensus mature sequence, consensus star sequence and consensus precursor sequence. Furthermore, the miRBase miRNAs existent in the input data but not scored by miRDeep2 are listed. - A directory called
pdfsthat contains for each provisional miRNA ID a PDF with its signature and structure. - A file called
result.csv(when option-cis used) that contains the same entries as the HTML file.
| option | description |
|---|---|
‑v <int> |
only output hairpins with score above <int> |
‑c |
also create overview in excel format |
‑k <file> |
supply file with known miRNAs |
‑s <file> |
supply survey file if score cutoff is used to get information about how big is the confidence of resulting reads |
‑f <file> |
miRDeep2 output MRD file |
‑e |
report complete survey file |
‑g |
report survey for current score cutoff |
‑w <project_folder> |
automatically used when running webinterface, otherwise don't use it |
‑r <file> |
Rfam file to check for already reported small RNA sequences |
‑q <file> |
miRBase.mrd file produced by quantifier module |
‑x <file> |
signature.arf file with mapped reads to precursors |
‑t <org> |
specify the organism from which your sequencing data was obtained |
‑u |
print all available UCSC input organisms |
‑d |
do not generate PDFs |
‑y |
timestamp |
‑z |
switch is automatically used when script is called by quantifier.pl |
‑o |
print reads in PDF signature sorted by their 3 letter code in front of their identifier |
make_html.pl -f miRDeep_outfile -s survey.csv -c -e -y 123456789Removes 3' end adaptors from deep sequenced small RNAs. The script searches for occurrences of the six first nucleotides of the adapter in the read sequence, starting after position 18 in the read sequence (so the shortest clipped read will be 18 nts). If no matches to the first six nts of the adapter are identified in a read, the 3' end of the read is searched for shorter matches to the 5 to 1 first nts of the adapter.
- A FASTA file with the deep sequencing reads and the adapter sequence (both in RNA or DNA alphabet).
- A FASTA file with the clipped reads.
FASTA IDs are retained. If no matches to the adapter prefixes are identified in a given read, the unclipped read is output.
clip_adapters.pl reads.fa TCGTATGCCGTCTTCTGCTTGT > reads_clipped.faIt is possible to clip adapters using more sophisticated methods. Users are encouraged to test other methods with the miRDeep2 modules.
Collapses reads in the FASTA file to ensure that each sequence only occurs
once.
To indicate how many times reads the sequence represents, a suffix is added to
each FASTA identifier. E.g. a sequence that represents ten reads in the data
will have the _x10 suffix added to the identifier.
- A FASTA file, either in standard format or in the collapsed suffix format.
- A FASTA file in the collapsed suffix format.
| option | description |
|---|---|
‑a |
outputs progress |
collapse_reads.pl reads.fa > reads_collapsedSince the script reads all FASTA entries into a hash using the sequence as key, it can potentially use more than 3 GB memory when collapsing very big datasets, >50 million reads. In this case, the user can partition the reads (for instance based on the 5' nucleotide), collapse separately and concatenate.
This script is a wrapper for excise_precursors.pl, which it calls one or more
times, incrementing the height of the read stack required for initiating
excision until the number of excised precursors falls below a given threshold.
- The reference genome in FASTA format,
- the mapped reads in
.arfformat, - a filename that the excised precursors will be written to, and
- the maximal number of precursors that should be reported.
| option | description |
|---|---|
‑a |
Output progress to screen. |
excise_precursors_iterative.pl genome.fa reads_vs_genome.arf \
potential_precursors.fa 50000 -aExcises precursors from the genome using the mapped reads as guidelines.
- The reference genome in FASTA format and
- the mapped reads in
.arfformat.
- The excised precursors in FASTA format.
| option | description |
|---|---|
‑a <integer> |
Only excise if the highest local read stack is <integer> reads high (default 2). |
‑b |
Output progress to screen. |
excise_precursors.pl genome.arf reads_vs_genome.arf -bPerforms simple filtering of entries in a FASTA file.
- A FASTA file.
- A filtered FASTA file.
| option | description |
|---|---|
‑a <int> |
only output entries where the sequence is minimum int nts long |
‑b |
remove all entries that have a sequence that contains letters other than a, c, g, t, u, n, A, C, G, T, U, or N. |
‑s |
output progress |
fastaparse.pl reads.fa -a 18 -s > reads_no_short.faThis script only prints out the FASTA entries that match an ID in the ID file.
- A FASTA file and a file with IDs, one ID per line.
- A FASTA file containing the FASTA entries that match an ID.
| option | description |
|---|---|
‑a |
only prints out entries that has an id that is not present in the ID file. |
fastaselect.pl reads.fa reads_select.ids > reads_select.faScans a file searching for the suffixes that are generated by
collapse_reads.pl (e.g. _x10).
It sums up the integer values in the suffixes and outputs the sum. If a given
id occurs multiple times in the file, it will multi-count the integer value of
the ID. It will also only count the first integer occurrence in a given line.
- Any file containing the suffixes that are generated by
collapse_reads.pl.
This will typically be a FASTA file or a list of IDs.
- The sum of integer values (the total read count).
find_read_count.pl reads_collapsed.faParses GSM format files into FASTA format.
- GSM files in tabular format.
The first column should be sequences and the second column the number of times the sequence occurs in the data.
- A FASTA file, one sequence per line (the sequences are expanded).
geo2fasta.pl GSM.txt > reads.faParses seq.txt or qseq.txt output from the Solexa/Illumina platform to
FASTA format.
- A
seq.txtor qseq.txtfile.
By default seq.txt.
- A FASTA file, one entry for each line of
seq.txt.
The entries are named seq plus a running number that is incremented by one
for each entry. Any . characters in the seq.txt file is substituted with an
N.
| option | description |
|---|---|
‑a |
format is qseq.txt |
illumina_to_fasta.pl s_1.qseq.txt -a > reads.faFor each potential miRNA precursor input, the miRDeep2 core algorithm either discards it or assigns it a log-odds score that reflects the probability that the precursor is a genuine miRNA.
Default input is
- an ARF file with the read signatures and
- an RNAfold output file with the structures of the potential miRNA precursors.
- A .mrd file with all potential miRNA precursors that are scored.
| option | description |
|---|---|
‑h |
print this usage |
‑s |
FASTA file with reference mature miRNAs from one or more related species |
‑t |
print filtered |
‑u |
limited output (only ids) |
‑v |
cut-off (default 1) |
‑x |
sensitive option for Sanger sequences |
‑y |
file with randfold p-values |
‑z |
consider Drosha processing |
miRDeep2_core_algorithm.pl signature.arf potential_precursors.str \
-s miRBase_related_species.fa -y potential_precursors.rand > output.mrdThe -z option has not been thoroughly tested.
Performs simple filtering of entries in an .arf file.
Default input is
- an
.arffile.
- A filtered
.arffile.
| option | description |
|---|---|
‑a <int> |
Discard mappings of edit distance higher than this |
‑b <int> |
Discard mappings of read queries shorter than this |
‑c <int> |
Discard mappings of read queries longer than this |
‑d <file> |
Discard read queries not in this file |
‑e <file> |
Discard read queries in this file |
‑f <file> |
Discard reference dbs not in this file |
‑g <file> |
Discard reference dbs in this file |
‑h |
Discard remaining suboptimal mappings |
‑i <int> |
Discard remaining suboptimal mappings and discard any reads that have more remaining mappings than this |
‑j |
Remove any unmatched nts in the very 3' end |
‑k |
Output progress to standard output |
parse_mappings.pl reads_vs_genome.arf -a 0 -b 18 -c 25 -i 5 \
> reads_vs_genome_parsed.arfPerforms a designated number of rounds of permuted controls (for details, see Friedländer et al., Nature Biotechnology, 2008).
The permutation controls estimate the number of false positives produced by a
miRDeep2_core_algorithm.pl run.
The input to perform_controls.pl should be
- a file containing the exact command line used to initiate the
miRDeep2_core_algorithm.plrun, - the structure file input to
miRDeep2_core_algorithm.pl, and - the desired rounds of controls.
- A file in
.mrdformat.
The output of each control run is separated by a line permutation integer.
The mean number of entries output by the control runs gives an estimate of the
false positives produced. The further contents (besides the number of entries)
of the .mrd output by perform_controls.pl is not biologically meaningful.
| option | description |
|---|---|
‑a |
Output progress to screen |
perform_controls.pl line potential_precursors.str 100 \
> output_controls.mrdIn a file output by RNAfold, each entry can be partitioned into an 'id' part
and an 'other' part, consisting of the dot-bracket structure, sequence, mfe
etc. This scripts reads all 'id' parts into a hash and pairs them with 'other'
parts from random entries. This is used by the perform_controls.pl script.
- An RNAfold output file.
- An RNAfold output file with IDs moved to random entries.
permute_structure.pl potential_precursors.str \
> potential_precursors_permuted.strPrepares the signature file to be input to the miRDeep2_core_algorithm.pl
script.
- A FASTA file with deep sequencing reads and
- a FASTA file with precursors.
- A signature file in
.arfformat.
| option | description |
|---|---|
‑a <file> |
FASTA file with the sequences of known mature miRNAs for the species. These sequences will not influence the miRDeep scoring, but will subsequently make it easy to estimate sensitivity of the run. |
‑b |
Output progress to screen |
prepare_signature.pl reads_collapsed.fa potential_precursors.fa \
-a miRBase_this_species.fa > signature.arfSubstitutes us and Us to Ts.
This is useful since bowtie does not match Us to Ts.
- A FASTA file.
- A substituted FASTA file.
rna2dna.pl reads_RNA_alphabet.fa > reads_DNA_alphabet.faThis script identifies potential precursors whose structure is basically consistent with Dicer recognition. Since running randfold is time-consuming, it is practical only to estimate p-values for those potential precursors that actually fold into hairpin structures.
- An ARF file with the read signatures and
- an RNAfold output file with the structures of the potential miRNA precursors.
- A list of ids, separated by newlines.
select_for_randfold.pl signature.arf potential_precursors.str \
> potential_precursors_for_randfold.idsSurveys miRDeep2 performance at score cut-offs from -10 to 10.
Default input is
- a
.mrdfile output by themiRDeep2_core_algorithm.plscript.
- A .csv file with performace statistics.
| option | description |
|---|---|
‑a <file> |
file outputted by controls |
‑b <file> |
mature miRNA FASTA reference file for the species |
‑c <file> |
signature file |
‑d <int> |
read stack height necessary for triggering excision |
survey.pl output.mrd -a output_controls.mrd -b miRBase_this_species.fa \
-c signature.arf -d 2 > survey.csvIt converts a bowtie bwt mapping file to a mirdeep arf file.
- A file in
bwtformat.
- A file in
mirdeeparfformat.
It converts a bwa sam mapping file to a mirdeep arf file.
- A
bwacreated file insamformat.
- A file in
mirdeeparfformat.
It gives information about the bwa FLAG in a bwa created mapping file in
sam format.
- A FLAG number created by
bwa.
- Information about the alignment created by
bwa.
Removes 3' end adaptors from deep sequenced small RNAs. The script searches for occurrences of the six first nucleotides of the adapter in the read sequence, starting after position 18 in the read sequence (so the shortest clipped read will be 18 nts). If no matches to the first six nts of the adapter are identified in a read, the 3' end of the read is searched for shorter matches to the 5 to 1 first nts of the adapter.
- A FASTA file with the deep sequencing reads and
- the adapter sequence (both in RNA or DNA alphabet).
- A FASTA file with the clipped reads.
FASTA IDs are retained. If no matches to the adapter prefixes are identified in a given read, the unclipped read is output.
clip_adapters.pl reads.fa TCGTATGCCGTCTTCTGCTTGT > reads_clipped.faIt is possible to clip adapters using more sophisticated methods. Users are encouraged to test other methods with the miRDeep2 modules.
It checks the supplied genome FASTA file for its correctness.
Identifier lines are not allowed to contain whitespaces and must be unique.
Sequence lines are not allowed to contain characters others than
A, C, G, T, N, a, c, g, t, or n.
- A genome file in FASTA format
It checks the supplied mapping file for its correctness. Each line in the file must be in the ARF format.
- A mapping file in ARF format.
It checks the supplied mature_miRNA FASTA file for its correctness.
Identifier lines are not allowed to contain whitespaces and must be unique.
Sequence lines are not allowed to contain characters others than A, C, G,
T, N, a, c, g, t, or n.
- A mature miRNA file in FASTA format.
It checks the supplied reads file for its correctness.
Each identifier line must have the format of '>name_uniqueNumber_xnumbere.g.>xyz_1_x20. See also file format_descriptions.txt` for more detailed
informations.
- A mapping file in ARF format.
Extracts mature and precursor sequences from miRBase fasta files for species of interest.
- A fasta file from miRBAase
- One or more species three letter code abbreviations
- A fasta file in a proper format usable by quantifier.pl and miRDeep2.pl.
- Multiline sequences from input files are put on a single line and MacOS and Windows linebreaks/carriage returns are removed
extract_miRNAs.pl mature_miRBase.fa hsa > mature_hsa.fa
extract_miRNAs.pl hairpin_miRBase.fa hsa > hairpin_hsa.fa
extract_miRNAs.pl mature_miRBase.fa mmu,chi > mature_other.fa
