WORKFLOW

WORKFLOW 
---

The user can upload either FASTQ or FASTA files bearing reads using the 
web interface or the MIRPIPE FTP server. These should ideally be 
compressed (.zip, .gz) to reduce upload time. The pipeline can fully 
process raw reads originating from Illumina, 454, IonTorrent or Sanger 
sequencing instruments including adapter trimming.
--- Parameters:
- Reads: File bearing reads in FASTQ or FASTA format (ideally zip 
compressed for Galaxy). This file can either be uploaded using the Galaxy 
Upload Tool (Helpful Tools / Get Data / Upload Files) or using an account 
on our FTP server. The latter is only possible after the registration of a 
user in Galaxy, which automatically creates an account with the same 
username and password on the FTP server 
(ftp://bioinformatics.mpi-bn.mpg.de/). The data will be deleted from the 
server after two weeks.


A reference FASTA database bearing mature target miRNAs can either be 
selected from the preprocessed current miRBase release 20 data harbouring 
30424 entries of 206 species or can be uploaded by the user in FASTA 
format. The user can optionally choose a subset of the miRBase reference 
miRNAs bearing only miRNAs of the desired organism to limit the comparison 
the e.g. the closest relative. If the chosen reference FASTA file does not 
obey to the naming convention of miRBase (<species>-miR-<#>-<suffix>), the 
"family name clustering" parameter should be turned off.
--- Parameters:
- Reference database: Preprocessed DBs (full miRBase or miRNAs of only one 
species) or any user uploaded FASTA file bearing mature miRNAs. The 
correct miRBase file can be downloaded for offline usage: 
ftp://mirbase.org/pub/mirbase/CURRENT/mature.fa.gz.


The raw read data is then processed to optionally remove an adapter 
sequence and trim for a minimum quality (default Q20). Only reads of the 
desired size range are selected to limit the pool to likely mature miRNAs 
(default: 18-28 nt).
--- Parameters:
- Adapter sequence: Nucleotide sequence of the adapter to be removed from 
the 3' end using Cutadapt. By default the larger of the following values 
is used as the maximum mismatch number:  1, 10% of the adapter length. 
These values can be changed inside the mirpipe.pl script.
- Minimum read length: Minimum length of a read after trimming to be 
considered in the analysis.
- Minimum base quality: Minimum phred quality for FASTQ data. Nucleotides 
with lower quality will be trimmed. This parameter is not used if FASTA 
formatted read data is supplied.
- Maximum read length: Maximum length of a read to be considered in 
analysis.


Duplicate reads are collapsed to decrease the number of necessary homology 
searches (the number of duplicates per read is noted). Only those 
sequences present a minimum number of times (default = 5) are kept for 
further analyses. This measure is intended to remove unique reads which 
frequently denote sequencing errors or lowly expressed miRNAs that can not 
be reliably quantified. Setting this parameter to "1" will increase 
sensitivity at the cost of an increased false positive rate.
--- Parameters:
- Minimum read copy number: A read sequence must be present at least this 
number of times to be included.


Read counts from isomiRs of the same miRNA are combined. These isomiR read 
sequences may only differ by the 3' end and are thus putatively encoded by 
the same gene and bear the same target specificity. This function allows 
the summary of putatively functionally equivalent isomiRs resulting from 
imperfect digestion by the RNases Drosha and Dicer or RNA-Editing by 
specialized enzymes resulting in 3' modification. Only the final 3' 
nucleotide may differ between two sequences to be counted as isoforms of 
the same miRNA and only the longest isoform sequence is used in the next 
step to reduce the amount of homology searches per miRNA.


The resulting read sequences are compared versus the chosen reference 
database of miRNAs. Sensitivity and specificity of this BLASTN homology 
search can be controlled using various parameters. Parameters are 
optimized for small query sequences (-num_alignments 15 -word_size 7 
-evalue 10 -dust no -strand plus). The resulting hits are filtered to 
exclude those with too many mismatches ((read length - alignment length) + 
mismatches + gaps = final mismatches).
--- Parameters:
- Maximum mismatches: Maximum number of mismatches allowed between 
reference miRNA and read sequence ((read length - alignment length) + 
mismatches + gaps = final mismatches). This parameter controls the size of 
the miRNA clusters: more mismatches allowed = larger clusters.


Mature miRNAs and their precursors are optionally collated by name on the 
family level to remove redundancy (ex. 
bta-miR-200a,oan-miR-200a-3p,tgu-miR-200a-3p -> miR-200a). Otherwise the 
complete miRNA names given in the reference database are carried over 
resulting in more detailed but also more redundant output. Turning off the 
family name clustering can be advisable in case the reference database of 
miRNA sequences does not obey to the naming convention of miRBase 
(<species>-miR-<#>-<suffix>).
--- Parameters:
Family name clustering: Collapse the names of all variants of a miRNA to 
the miRNA family (ex. bta-miR-200a,oan-miR-200a-3p,tgu-miR-200a-3p -> 
miR-200a).


Detected reference miRNA families per read are scored based on the minimum 
number of mismatches. If a read matched equally well versus multiple miRNA 
families, the respective families are joined by single linkage clustering. 
By default only those read sequences that are at least 5% as abundant as 
the most abundant sequence per miRNA family cluster are denoted (ex. most 
abundant sequence = 100 reads, cutoff = 5 reads). This is intended to 
further suppress reads resulting from sequencing errors or biological 
miRNA variations that are expressed near the detection limit.
--- Parameters:
Minimum cluster abundance: Remove read sequences from a cluster that are 
less than x% as abundant as the most abundant sequence. This is intended 
to suppress reads resulting from sequencing errors or biological miRNA 
variations that are expressed near the detection limit. This parameter 
controls the size of the miRNA clusters: lower minimum cluster abundance = 
larger clusters.


In order to achieve congruent results yielding one count value per miRNA, 
miRNA family clusters are finally split. Since some of the reads match 
multiple miRNAs equally well, these reads are counted fully for all of the 
respective miRNAs. This can lead to a situation where the summarized read 
counts of all miRNAs can be higher than the amount of reads totally 
matching. Each miRNA is associated with an ambiguity value, denoting the 
share of reads that could not be placed clearly (e.g. 11/89 reads 
ambiguous = 0.12). If this value is high, the respective miRNA count may 
be misleading. Finally, the most abundant sequence matching a miRNA is 
given (primary sequence) as well as the number of reads matching it.



OUTPUT FILES
---


- 1. mirpipe_cluster.tsv: MIRPIPE miRNA clusters = output of one read 
sequence per line

This file is centred on the different read sequences found per miRNA 
cluster that result from biological and technical variation. Only those 
read sequences that are >=5% as abundant as the most abundant sequence per 
cluster are denoted by default. If a read matched equally well versus 
multiple miRNAs, the respective miRNAs or miRNA clusters are joined by 
single linkage clustering.

Columns:
Cluster     Cluster number
Sequence    Read sequence
Count       Summarized read count for all duplicates of this read
miRNA       Name of miRNA or miRNA families

Example (sorted for cluster number, expression):
Cluster  Sequence                     Count  miRNA
     90  CAGTACTGTGATAACTGAAGAA          33  miR-101a
     90  CTACTGTGATAACTGACT              17  miR-101c,miR-101a


- 2. mirpipe_cluster.fasta: MIRPIPE cluster sequences

All sequences reported in the MIRPIPE miRNA cluster's file in fasta format.

Example:
>miR-101a count=33
CAGTACTGTGATAACTGAAGAA
>miR-101a,miR-101c count=17
GTACTGTGATAACTGACT


- 3. mirpipe_mirna.tsv: MIRPIPE miRNAs on 5% level = output of one miRNA 
per line

This file includes one count value per miRNA and can directly serve as 
input for subsequent differential expression analyses. It is based on 
clusters of highly similar miRNAs, where a clear assignment of reads is 
not always possible, since the same read can match equally well to 
multiple reference miRNAs. Only those miRNA sequences are reported that 
are >5% as abundant as the most abundant sequence in its cluster.

Columns:
miRNA                   Name of miRNA or miRNA family
Count		        Summarized read count including isomiRs, 
biological + technical sequence variations
Ambiguous reads         Ratio of reads that mapped equally well to other 
miRNAs inside the miRNA family cluster
Cluster                 miRNA family cluster number
Primary sequence        Most abundant sequence for this miRNA inside the 
cluster
Primary sequence count  Count of the most abundant sequence for this miRNA 
inside the cluster
Cluster members         A comma-separated list of all members of the miRNA 
family cluster

Example (sorted for cluster number, expression):
miRNA     Expression  Ambiguity  Cluster  Primary Sequence     PS Count  
Cluster members
miR-101a         143       0.12       90  CAGTACTGTGATAACTGAAGAA     33  
miR-101a,miR-101c
miR-101c          17          1       90  GTACTGTGATAACTGACT         17  
miR-101a,miR-101c



EXAMPLE
---

The following example shows a MIRPIPE result using default parameters. Two 
miRNAs (miR-2478,miR-3968) were joined into a miRNA cluster based on 
BLASTN results.

mirpipe_cluster.tsv
Cluster Sequence                Count   miRNA
192	ATCCCACTTCTGACACCA	69	miR-2478
192	ATCCCACTCTCAACACCA	11	miR-3968
192	ATCCCACTCCTGACACCA	11	miR-2478,miR-3968
192	ATCCCATTCTTGACACCA	9	miR-2478
192	TCGAATCCCACTCCTGACACCA	6	miR-3968
192	AATCCCACTCTCAACACCA	5	miR-3968
192	TCAAATCCCACTCTCAACACCA	5	miR-3968

mirpipe_cluster.fasta:
>miR-2478 count=69
ATCCCACTTCTGACACCA
>miR-3968 count=5
AATCCCACTCTCAACACCA
>miR-3968 count=11
ATCCCACTCTCAACACCA
>miR-2478,miR-3968 count=11
ATCCCACTCCTGACACCA
>miR-2478 count=9
ATCCCATTCTTGACACCA
>miR-3968 count=5
TCAAATCCCACTCTCAACACCA
>miR-3968 count=6
TCGAATCCCACTCCTGACACCA

mirpipe_mirna.tsv:
miRNA        Count Ambiguity  Cluster   Primary Sequence        Primary 
Sequence Reads  Cluster members
miR-2478	89	0.12	192	ATCCCACTTCTGACACCA	69	
miR-2478,miR-3968
miR-3968	38	0.29	192	ATCCCACTCTCAACACCA	11	
miR-2478,miR-3968

The mirpipe_cluster.tsv file depicts the best BLASTN hit per read sequence 
based on the least number of mismatches. Sequences are sorted for 
expression from top to bottom with the least expressed sequence still at 
least 5% as abundant as the most expressed sequence (69 <> 5). The two 
miRNAs were joined to a cluster because one of the read sequences showed a 
BLASTN hit which fit equally well to both reference sequences (192	
ATCCCACTCCTGACACCA	11	miR-2478,miR-3968). If another query had 
found that e.g. miR-2478 and miR-1000 had resulted in equally similar 
homologies, the two clusters would have been joined to 
miR-2478,miR-3968,miR-1000.

The mirpipe_cluster.fasta file shows all read sequences found in 
mirpipe_cluster.tsv converted to FASTA format.

The mirpipe_mirna.tsv file attempts to include one count value per miRNA 
in order to facilitate later quantification. The count values for each 
sequence detected per miRNA are summarized (e.g.: miR-2478 = 69 + 11 + 9 = 
89, miR-3968 = 11 + 11 + 6 + 5 + 5 = 38). Since some of the reads matched 
two different miRNAs equally well (miR-2478,miR-3968 = 11), these reads 
are counted fully for both miRNAs. This leads to a situation where the 
summarized read counts of all miRNAs can be higher than the amount of 
reads totally matching. Each miRNA is associated with an ambiguity value, 
denoting the share of reads that could not be placed clearly (e.g. 
miR-2478: 11/89 ambiguous = 0.12). If this value is high, the respective 
miRNA count may be misleading. Finally, the most abundant sequence 
matching a miRNA is given (primary sequence) as well as the number of 
reads matching it.