Contents

1 Introduction

The miRanalyzer standalone version requires the following third party software that needs to be installed first:

• Vienna RNA package for RNA Secondary Structure Prediction and ComparisonVienna package
• Bowtie - An ultrafast memory-efficient short read aligner (Bowtie)
• Weka: Data Mining Software in Java (Weka). The tested version is weka 3.5.3. NOTE that the “weka.jar” file (named exactly like this, i.e. without version) must be in the same folder as the miRanalyzer.jar (see below).

2 Downloads

In this section the available programs and perl scripts can be downloaded. Please read next section on how to install

• groupReads: this script can be used to convert fastq to RC (read-count) format.
• makeSeqObj: this java jar can be used to to prepare the genome sequences (see below). Please note that the current version has a very inefficient memory usage and you might assign several Gigas of memory (for example -Xmx5000m).
• miRanalyzer jar file: the newest version of the miRanalyzer algorithm (miRanalyzer_0.2, 02/14/2011)
• models for version 0.2: the newest model files (they may just work with the newest version!)
• start-up DB: the start-up database (see below)

3 Installation

miRanalyzer relies on a huge number of libraries like mature microRNAs, chromosome sequences, bowtie indexes, etc. These data needs to be stored in a local file-base database. Before using miRanalyzer, this database needs to be generated.

3.1 General structure of the database

The easiest way to generate the basic structure of the database is to download the miRanalyzer start-up DB following this steps:

• download the start-up DB to your miRanalyzer directory (like /home/user)
• extract the tar: tar xvzf miRanalyzerDB.tgz

Now you will see a directory named miRanalyzerDB which is the base directory . Inside the base directory you will find 4 directories, the miRanalyzer jar file, the makeSeqObj jar file, the fastq convertion perl script and a example input file from Drosophila melanogaster. Please note that all folders must have exactly the names as used in this manual, otherwise miRanalyzer can not access the data. The folders are the following:

• bowtie: the folder where the bowtie indexes must be (see below)
• model: this folder holds the model files for the prediction of new microRNAs. Download all model files here: miRanalyzer model files
• out: the default output folder, a folder with the name of the input file will be generated in this directory.
• seqOBJ: the genome sequences in miRanalyzer format (to download see the table below).

The bowtie folder needs several subfolders which need to be named exactly as follows:

• genome: the indexes of the whole genome sequences. The names must be the same as used for the genome sequences (in seqOBJ folder)
• mature: the indexes for the mature microRNAs. Download here all indexes and put them into the ’mature’ folder.
• maturestar: the indexes for the maturestar microRNAs. Download here all indexes and put them into the ’maturestar’ folder.
• maturestarunobs: the indexes for the maturestar microRNAs which are not in miRBase, i.e. that have not been observed before. Download here all indexes and put them into the ’maturestarunobs’ folder.
• hairpin: the indexes for the hairpin - precursor- sequences of the microRNAs. Download here all indexes and put them into the ’hairpin’ folder.
• translibs: the indexes of the other libraries which should be used. The files can be generated with bowtie-built. The basename of colorspace libraries must finish in ’_c’. For example rfam_c*. Note that, however the input name would just be ’rfam’, as miRanalyzer add the ’_c’ in the case that color space sequences have been indicated in the input.

3.2 Populate the database

3.2.1 Bowtie indexes

The bowtie indexes for the miRBase can be downloaded above. Please, note that the mature microRNA sequences are to short to work with normal read lengths. Therefore 25 Gs are added in order to be able to use Bowtie. Please, make sure using the latest version of miRBase indexes provided here, or take into account the 25 Gs added when preparing your own microRNA references. Please, use the bowtie-built algorithm or download the bowtie indexes from the bowtie pagefor the genome assemblies and transcribed libraries.

3.2.2 Genome sequence data

For the prediction of new microRNAs, miRanalyzer needs to extract sequences from the genome assemblies. In order to facilitate a rapid access to this information, miRanalyzer uses preprocessed fasta files. Table 1 shows the assemblies which are currently available and provides the download links. The user can generate these files for other species or unpuplished assemblies using the makeSeqObj Java program. The program takes a (multi)fasta as input and generates the corresponding zip file.

3.2.3 Models

The latest models for animal and plant prediction of new microRNAs can be found here (if not downloaded along with the start-up DB). They must be placed in the “models” folder within the miRanalyzer directory.

4 Using miRanalyzer

Once the database is generated, miRanalyzer can be used. In general, the command line parameters must be given in the following format: parameterName=value. The program uses the following input parameters:

4.1 Mandatory parameters

• input=file: the name of the input file. The input can be given in read-count format (’read sequence’ ’read frequency’) or in multi fasta format (see the manual for more information)
• dbPath=path: the absolutecomplete path to the miRanalyzer database
• species=’assembly name’: Assembly/basename of bowtie index, i.e. hg18,mm8,rn4,etc.
• speciesShort=’short name of species’: This must be the abbreviation of the species used in miRBase for example hsa for Homo sapiens or mmu for Mus musculus.
• kingdom=plant or kingdom=animal: In order to set the models and features for the prediction of new microRNAs the program needs to know whether the species is a plant or animal.
• bowtiePath=path: the path with the Bowtie binaries

4.2 Optional parameters

• translibs=’parameter string’: the additional libraries which should be used. This parameter actually takes into account four different values which are ’encoded’ into a parameter string. The parameter string has the following format: ’name of bowtie basename’:’name of library’:’maximum number of considered matches’:’remove value’. Only the basename of the bowtie index is mandatory, the other three are optional. If not set, the default values are: a) ’name of library’ is set to the name of the bowtie basename (miRanalyzer uses this name in the output), b) ’maximum number of considered matches’ is set to 20 (this value is used as bowtie parameter -k –> the maximum number of reported alignments) and c) ’remove value’ is set to 5, i.e. if a read matches to more that 5 reference sequences it is removed and therefore not used in the prediction of new microRNAs.
• solid=true: if this parameter is set, then the input is treated as color space format.
• output=’output folder’: by default the output is written in the ’out’ folder in the miRanalyzer directory. The default folder is named after the input. For example, if the input is called “hsa_G0.txt”, then within the ’out’ folder a subfolder will be generated named ’hsa_G0’ where the output is written.
• minReadLength=value: the minimum read length. Note that this length is also used as seed length in the bowtie alignment against known microRNAs and the genome (default: 17).
• minReadLengthTrans=value: the seed length in bowtie alignments against the transcribed libraries (default: 20).
• maxReadLength=value: the maximum read length. miRanalyzer performs an internal re-grouping with this length (default: 26).
• noMM=value: allowed number of mismatches to known microRNAs (default: 1)
• noMMTrans=value: allowed number of mismatches to transcribed libraries (default: 1)
• noMMGenome=value: allowed number of mismatches to the genome (default: 1)
• score=value: the minimal score so that the candidate is considered a new microRNA. The value must be between 0 and 1, although very low values will lead to drastically overpredicting. (default: 0.9)
• minNoPositives=value: miRanalyzer predicts using 5 models (5 different negative sets). This parameter determines the minimum number of models which predicts a candidate to be a new microRNA (default: 3).
• maxNoGenome=value: the maximum number of considered matches to genome (-k option in bowtie) (default: 8)
• maxNoKnown=value: the maximum number of considered matches to known microRNAs (-k option in bowtie) (default: 10)

4.3 An example

This would be an example to launch the example data file provided with the start-up db. Note that you need to adapt the dbPath and bowtiePath to your local values. For very big input files, it might be necessary to increase the memory (-Xmx option of Java VM).

java -Xmx2000m -jar miRanalyzer.jar input=SRR069503.rc dbPath=/home/user/miRanalyzer species=dm3 speciesShort=dme kingdom=animal bowtiePath=/usr/local/bin/bowtie64 translibs=rfam:RFam:15:5

5 miRanalyzer output

miRanalyzer produces several different of output files. In the following * can be mature, maturestar, matureunobs, hairpin or any of the transcribed libraries.

5.1 *_unique.txt:

A summary of the reference sequences which have been uniquely matched. That means all reads just map to this and not any other reference sequence. The follwing columns exist:

• name: the name of the reference sequence
• #uniqueReads: the number of unique reads mapped
• readCount: the read count sum of all unique reads
• norm_expressed_all: the read count divided by the total read count in the experiment, times 100 (percentage)
• norm_expressed_mapped: the read count divided by the read count of all unique reads mapped to this library, times 100 (percentage)

5.2 *_ambig.txt

Sometimes a read maps with the same quality (number of mismatches and length) to more than one reference sequence. In such cases, miRanalyzer groups together those reference sequences and reports them as ambigous matches. We do so, as sometimes the members of a family have very similar sequences and cannot be readily distinguished. However, by means of this ambigous matches at least the microRNA family can be detected. The format of the file is the same as the “*_unique” file.

5.3 *_reads.txt

A summary on the reads level.

• ReadSequence: the sequence of the read
• ReadCount: the read count (number of copies)
• NoMatches: the number of reference sequences the read has mapped to
• alignLength: the length of the longest alignment
• name: the names of the reference sequences the read mappes to
• readID: a unique ID assigned by miRanalyzer

5.4 *_parsed.txt

The parsed bowtie output for the library. It holds just the longest alignments with the lowest number of mismatches (see Bowtie manual for more details).

5.5 *_grouped.txt

All reference sequences mapped by at least one read (no distinction between unique and ambigous matches!).

• name: the name of the reference sequence
• readCountSum: the sum of all read counts mapped to this reference sequence
• uniqueReadSum: the number of unique reads mapped to this reference
• readsString: is an encoded string with information of all reads mapped to this sequence. The information on the different reads is separated by ’#’. For each read the read sequence, the read count, and the position in the alignment (0-based) is given (separated by ’:’).

5.6 newMicroRNA.txt & newMicroRNA_pure.txt

The “newMicroRNA.txt” and “newMicroRNA_pure.txt” files hold a summary on the newly predicted microRNAs. Depending on the configuration, some reads which mapped to the transcriptome or RFam are also used for the prediction of new microRNAs. Those reads migth be more prone to cause false positive predictions and therefore, in the file “newMicroRNA_pure.txt” only those microRNAs with previously unmapped reads are reported. The columns of the files are:

• name: the name of the candidate
• chrom: the chromosome
• chromStart: the start coordinate
• chromEnd: the end coordinate
• strand: the strand
• #unique reads: the number of unique reads
• read count: the read count sum of all unique reads
• #posPredictions: the number of times this new microRNA has been predicted (in either of the 5 models)
• norm. read count: the normalized read count (divided by the total read count of all reads in the analysis)
• cluster sequence: the sequence formed by all reads

5.7 Candidates.txt and Candidates_pure.txt

“Candidates.txt” and “Candidates_pure.txt” files contain additional information on the newly predicted microRNAs. The first column is the name of the candidate. Then follows a string which holdsseparated by “;”: the precursor sequence, the secondary structure, the mean free energy of the structure, the chromosome, chromosomic coordinate start, chromosomic coordinate end, the number of times it was predicted and the reads which compose this new microRNA. The different reads are separated by ’:’. The information for the different reads is the following (separated by ’@’): read sequence, position in precursor (0-based), read count and aligment length.