Seqcluster¶

With bcbio installed

If you already have `bcbio`_, seqcluster comes with it. If you want the last development version:

/bcbio_anaconda_bin_path/seqcluster_install.py --upgrade

Binstar binary

install conda if you want an isolate env:

wget http://repo.continuum.io/miniconda/Miniconda-latest-Linux-x86_64.sh
bash Miniconda-latest-Linux-x86_64.sh -b -p ~/install/seqcluster/anaconda

You can install directly from binstar (only for linux):

~/install/seqcluster/anaconda/conda install seqcluster bcbio-nextgen -c bioconda

With that you will have everything you need for the python package. The last step is to add seqcluster to your PATH (see below).

Go to Tools dependecies below to continue with the installation.

Step by step

If you want to install step by step from a new conda environment:

~/install/seqcluster/anaconda/bin/conda install -c bioconda bcbio-nextgen

Link binary to any folder is already in your path:

ln -s ~/install/seqcluster/anaconda/bin/seqcluster* ~/install/seqcluster/linuxbrew/bin/.

Note: After installation is highly recommended to get the last updated version doing:

seqcluster_install.py --upgrade

Tools dependecies¶

For seqcluster command:

• bedtools
• samtools
• rnafold (for HTML report)

For some steps of a typical small RNA-seq pipeline (recommended to use directly `bcbio`_ ):

• STAR, bowtie
• fastqc
• cutadapt (install with bioconda using the same python env than seqcluster.

You will need to link the cutadapt binary to your PATH)

easy installation

Strongly recommended to use bcbio installation if you work with sequencing data. But if you want a minimal installation:

pip install fabric
seqcluster_install --upgrade
mkdir -p $PATH_TO_TOOLS/bin
seqcluster_install --tools $PATH_TO_TOOLS

After that you will need to add to your path: export PATH=$PATH_TO_TOOLS/bin:$PATH

Data¶

Easy way to install your small RNA seq data with cloudbiolinux. Seqcluster has snipped code to do that for you. Recommended to use `bcbio`_ for the pipeline since will install everything you need in a single step bcbio_nextgen.py upgrade -u development --tools --genomes hg19 --aligners bowtie.

But If you want to run seqcluster step by step an example of hg19 human version it will be (another well annotated supported genome is mm10):

Download genome data:

seqcluster_install --data $PATH_TO_DATA --genomes hg19 --aligners bowtie2 --datatarget smallrna

If you want to install STAR indexes since gets kind of better results than bowtie2 (warning, 40GB memory RAM needed):

seqcluster_install --data $PATH_TO_DATA --genomes hg19 --aligners star

R package¶

Install isomiRs package for R using devtools:

devtools::install_github('lpantano/isomiRs')

To install all packages used by the Rmd report:

Rscript -e 'source(https://raw.githubusercontent.com/lpantano/seqcluster/master/scripts/install_libraries.R)'

clustering of small RNA sequences¶

seqcluster generates a list of clusters of small RNA sequences, their genome location, their annotation and the abundance in all the sample of the project

REMOVE ADAPTER

I am currently using cutadapt:

cutadapt --adapter=$ADAPTER --minimum-length=8 --untrimmed-output=sample1_notfound.fastq -o sample1_clean.fastq -m 17 --overlap=8 sample1.fastq

COLLAPSE READS

To reduce computational time, I recommend to collapse sequences, also it would help to apply filters based on abundances. Like removing sequences that appear only once.

seqcluster collapse -f sample1_clean.fastq -o collapse

Here I am only using sequences that had the adapter, meaning that for sure are small fragments.

PREPARE SAMPLES

seqcluster prepare -c file_w_samples -o res --minl 17 --minc 2 --maxl 45

the file_w_samples should have the following format:

lane1_sequence.txt_1_1_phred.fastq      cc1
lane1_sequence.txt_2_1_phred.fastq      cc2
lane2_sequence.txt_1_1_phred.fastq      cc3
lane2_sequence.txt_2_1_phred.fastq      cc4

two columns file, where the first column is the name of the file with the small RNA sequences for each sample, and the second column in the name of the sample.

The fastq files should be like this:

@seq_1_x11
CCCCGTTCCCCCCTCCTCC
+
QUALITY_LINE
@seq_2_x20
TGCGCAGTGGCAGTATCGTAGCCAATG
+
QUALITY_LINE
</pre>

Where _x[09] indicate the abundance of that sequence, and the middle number is the index of the sequence.

This script will generate: seqs.fastq and seqs.ma. * seqs.fastq: have unique sequences and unique ids * seqs.ma: is the abundance matrix of all unique sequences in all samples

ALIGNMENT

You should use an aligner to map seqs.fa to your genome. A possibility is bowtie or STAR. From here, we need a file in BAM format for the next step. VERY IMPORTANT: the BAM file should be sorted

bowtie -a --best --strata -m 5000 INDEX seqs.fastq -S | samtools view -Sbh /dev/stdin | samtools sort -o /dev/stdout temp > seqs.sort.bam

or

STAR --genomeDir $star_index_folder --readFilesIn res/seqs.fastq --alignIntronMax 1  --outFilterMultimapNmax 1000 --outSAMattributes NH HI NM --outSAMtype BAM SortedByCoordinate

CLUSTERING

seqcluster cluster -a res/Aligned.sortedByCoord.out.bam  -m res/seqs.ma -g $GTF_FILE  -o res/cluster -ref PATH_TO_GENOME_FASTA --db example

• -a is the SAM file generated after mapped with your tool, which input has been seqs.fa
• -m the previous seqs.fa
• -b annotation files in bed format (see below examples) [deprecated]
• -g annotation files in gtf format (see below examples) [recommended]
• -i genome fasta file used in the mapping step (only needed if -s active)
• -o output folder
• -ref genome fasta file. Needs fai file as well there. (i.e hg19.fa, hg19.fa.fai)
• -d create debug logging
• -s construction of putative precursor (NOT YET IMPLEMENTED)
• –db (optional) will create sqlite3 database with results that will be used to browse data with html web page (under development)

Example of a bed file for annotation (the fourth column should be the name of the feature):

chr1    157783  157886  snRNA   0       -

Strongly recommend gtf format. Bed annotation is deprecated. Go here to know how to download data from hg19 and mm10.

Example of a gtf file for annotation (the third column should be the name of the feature and the value after gene name attribute is the specific annotation):

chr1    source  miRNA      1       11503   .       +       .       gene name 'mir-102' ;

hint: scripts to generate human and mouse annotation are inside seqcluster/scripts folder.

OUTPUTS

• counts.tsv: count matrix that can be input of downstream analyses
• size_counts.tsv: size distribution of the small RNA by annotation group
• seqcluster.json: json file containing all information
• log/run.log: all messages at debug level
• log/trace.log: to keep trace of algorithm decisions

Interactive HTML Report¶

This will create html report using the following command assuming the output of seqcluster cluster is at res:

seqcluster report -j res/seqcluster.json -o report -r $GENONE_FASTA_PATH

where $GENOME_FASTA_PATH is the path to the genome fasta file used in the alignment.

Note: you can try our new visualization tool!

• report/html/index.html: table with all clusters and the annotation with sorting option
• report/html/[0-9]/maps.html: summary of the cluster with expression profile, annotation, and all sequences inside
• report/html/[0-9]/maps.fa: putative precursor

An example of the output is below:

Easy start with bcbio-nextgen.py¶

Note:If you already are using bcbio, visit bcbio to run the pipeline there.

To install the small RNA data:

bcbio_nextgen.py upgrade -u development --tools --datatarget smallrna

Options to run in a cluster

It uses ipython-cluster-helper to send jobs to nodes in the cluster

• –parallel should set to ipython
• –scheduler should be set to sge,lsf,slurm
• –num-jobs indicates how much jobs to launch. It will run samples independently. If you have 4 samples, and set this to 4, 4 jobs will be launch to the cluster
• –queue the queue to use
• –resources allows to set any special parameter for the cluster, such as, email in sge system: M=my@email.com

Read complete usability here: https://github.com/roryk/ipython-cluster-helper An examples in slurm system is:

--parallel ipython --scheduler slurm --num-jobs 4 --queue general

Output

• one folder for each analysys, and inside one per sample

• adapter: *clean.fastq is the file after adapter removal, *clean_trimmed.fastq is the collapse clean.fastq, *fragments.fastq is file without adapter, *short.fastq is file with reads < 16 nt.
• align: BAM file results from align trimmed.fastq
• mirbase: file with miRNA anotation and novel miRNA discovery with mirdeep2
• tRNA: analysis done with tdrmapper [citation needed]
• qc: *_fastqc.html is the fastqc results from the uncollapse fastq file

• seqcluster: is the result of running seqcluster. See its documentation for further information.
• report/srna-report.Rmd: template to create a quick html report with exploration and differential expression analysis. See example here

seqcluster¶

• counts.tsv: count matrix that can be input of downstream analyses. nloci will be 0 always that the meta-cluster has been resolved successfully. For instance, it can happen that you got sequences you have a bunch of sequences mapping to hundreds of different places on the genome, then seqcluster doesn’t resolve that, and put everything under the larger region covered by those sequences. So, mainly, 0 all are good rows. The ann column is just where the meta-clusters overlap with. It can happen that one name appears many times if different locations of the meta-cluster map to different copies of that feature. OR if the annotation file used had multiple lines for that.
• read_stats.tsv: number of reads for each sample after each step in the analysis. Meant to give a hint if we lose a lot of information or not.
• size_counts.tsv: size distribution of the small RNA by annotation group. (position, reads, cluster)
• seqcluster.json: json file containing all information. This file is used as the input of the report suit.
• log/run.log: all messages at debug level
• log/trace.log: to keep trace of algorithm decisions

Report¶

Beside the static HTML report that you can get using report subcommand, you can download this HTML. (watch the repository to get notifications of new releases.)

• Go inside seqclusterViz folder
• Open reader.html
• Upload the seqcluster.db file generated by report subcommand.
• Start browsing your data!

Meaning of different sections:

• Top-left table shows list of meta-clusters, user can filter by number ID or keywords.
• Top-right table shows positions where this meta-cluster has been detected.
• Expression profile along precursor: Lines are number of reads in that position of the precursor. It is sum of the log2 RPM of the expression for each sample.
• Table: raw counts for each sample and sequence. Only top 100 are shown.
• secondary structure: The region with more sequences inside meta-cluster is used to plot the secondary structure. Colors refers to abundance in each position. Darker means more abundance.

An example of the HTML code:

miRQC data¶

About

mirRQC project

samples overview:

>> Universal Human miRNA reference RNA (Agilent Technologies, #750700), human brain total RNA (Life Technologies, #AM6050), human liver total RNA (Life Technologies, #AM7960) and MS2-phage RNA (Roche, #10165948001) were diluted to a platform-specific concentration. RNA integrity and purity were evaluated using the Experion automated gel electrophoresis system (Bio-Rad) and Nanodrop spectrophotometer. All RNA samples were of high quality (miRQC A: RNA quality index (RQI, scale from 0 to 10) = 9.0; miRQC B: RQI = 8.7; human liver RNA: RQI = 9.2) and high purity (data not shown). RNA was isolated from serum prepared from three healthy donors using the miRNeasy mini kit (Qiagen) according to the manufacturer’s instructions, and RNA samples were pooled. Informed consent was obtained from all donors (Ghent University Ethical Committee). Different kits for isolation of serum RNA are available; addressing their impact was outside the scope of this work. Synthetic miRNA templates for let-7a-5p, let-7b-5p, let-7c, let-7d-5p, miR-302a-3p, miR-302b-3p, miR-302c-3p, miR-302d-3p, miR-133a and miR-10a-5p were synthesized by Integrated DNA Technologies and 5′ phosphorylated. Synthetic let-7 and miR-302 miRNAs were spiked into MS2-phage RNA and total human liver RNA, respectively, at 5 × 106 copies/μg RNA. These samples do not contain endogenous miR-302 or let-7 miRNAs, which allowed unbiased analysis of cross-reactivity between the individual miR-302 and let-7 miRNAs measured by the platform and the different miR-302 and let-7 synthetic templates in a complex RNA background. Synthetic miRNA templates for miR-10a-5p, let-7a-5p, miR-302a-3p and miR-133a were spiked in human serum RNA at 6 × 103 copies per microliter of serum RNA or at 5-times higher, 2-times higher, 2-times lower and 5-times lower concentrations, respectively. All vendors received 10 μl of each serum RNA sample.

Commands

Data was download from GEO web with this script. The following 2 configs were used for the two sets: mirqc samples and non mirqc samples. Samples were analyzed with bcbio with the following commands

report

Report showing part of the output report of bcbio pipelines together with some validations are here.

Processing of reads¶

REMOVE ADAPTER

I am currently using cutadapt.

cutadapt --adapter=$ADAPTER --minimum-length=8 --untrimmed-output=sample1_notfound.fastq -o sample1_clean.fastq -m 17 --overlap=8 sample1.fastq

COLLAPSE READS

To reduce computational time, I recommend to collapse sequences, also it would help to apply filters based on abundances. Like removing sequences that appear only once.

seqcluster collapse -f sample1_clean.fastq -o collapse

Here I am only using sequences that had the adapter, meaning that for sure are small fragments. The output will be named as sample1_clean_trimmed.fastq

Prepare databases¶

For human or mouse, follows this instruction to download easily miRBase files. For other species you only need hairpin.fa and miRNA.str from miRBase site. Highly recommended to filer hairpin.hsa to contain only the desired species.

miRNA/isomiR annotation with JAVA¶

MIRALIGNER

Download the tool from miraligner repository.

Download the mirbase files (hairpin and miRNA) from the ftp and save it to DB folder.

You can map the miRNAs with.

java -jar miraligner.jar -sub 1 -trim 3 -add 3 -s hsa -i sample1_clean_trimmed.fastq -db DB  -o output_prefix

Cite

SeqBuster is a bioinformatic tool for the processing and analysis of small RNAs datasets, reveals ubiquitous miRNA modifications in human embryonic cells. Pantano L, Estivill X, Martí E. Nucleic Acids Res. 2010 Mar;38(5):e34. Epub 2009 Dec 11.

miRNA/isomiRs annotation with python¶

A new function to annotate miRNA/isomiR sequences using BAM files aligned to miRBase precursors or fastq files to align from scratch has been added to seqcluster:

seqcluster seqbuster --out results --hairpin hairpin.fa --mirna miRNA.str --species hsa input_file.fastq ...

If the input file is a BAM file, seqcluster will parse it to produce miRNA annotation, including isomiRs. If the input is FASTQ/FASTA file, seqcluster will map with the new C implementation of miraligner and annotate miRNAs and isomiRs as before.

Multiple files can be given to analyze all of them serially. Files inside the output folder are:

• raw mirna annotation to all posible mirnas (*.premirna)
• count file for miRNAs (counts_mirna.tsv)
• count file for isomiRs (counts.tsv)

NOTE: Check comparison of multiple tools for miRNA annotation.

Post-analysis with R¶

Use the outputs to do differential expression, clustering and descriptive analysis with this package: isomiRs

Manual of miraligner(JAVA)¶

options

Add -freq if you have your fasta/fastq file with this format and you want a third column with the frequency (normally value after x character):

>seq_1_x4
CACCGCTGTCGGGGAACCGCGCCAATTT

Add -pre if you want also sequences that map to the precursor but outside the mature miRNA

• Parameter -sub: mismatches allowed (0/1)
• Parameter -trim: nucleotides allowed for trimming (max 3)
• Parameter -add: nucleotides allowed for addition (max 3)
• Parameter -s: species (3 letter, human=>hsa)
• Parameter -i: fasta file
• Parameter -db: folder where miRBase files are(one copy at miraligner-1.0/DB folder)
• Parameter -o: prefix for the output files
• Parameter -freq: add frequency of the sequence to the output (just where input is fasta file with name matching this patter: >seq_3_x67)
• Parameter -pre: add sequences mapping to precursors as well

input

A fasta/fastq file reads:

>seq
CACCGCTGTCGGGGAACCGCGCCAATTT

or tabular file with counts information:

CACCGCTGTCGGGGAACCGCGCCAATTT 45

output

Track file *.mirna.opt: information about the process

Non mapped sequences will be on *.nomap

Header of the *.mirna.out file:

• seq: sequence

• freq/name: depending on the input this column contains counts (tabular input file) or name (fasta file)

• mir: miRNA name

• start: start of the sequence at the precursor

• end: end of the sequence at the precursor

• mism: nucleotide substitution position | nucleotide at sequence | nucleotide at precursor

• addition: nucleotides at 3 end added:

  precursor         => cctgtggttagctggttgcatatcc
  annotated miRNA   =>   TGTGGTTAGCTGGTTGCATAT
  sequence add:  TT =>   TGTGGTTAGCTGGTTGCATATTT
  

• tr5: nucleotides at 5 end different from the annonated sequence in miRBase:

  precursor             => cctgtggttagctggttgcatatcc
  annotated miRNA   =>   TGTGGTTAGCTGGTTGCATAT
  sequence tr5:  CC => CCTGTGGTTAGCTGGTTGCATAT
  sequence tr5:  tg =>     TGGTTAGCTGGTTGCATAT
  

• tr3: nucleotides at 3 end different from the annotated sequence in miRBase:

  precursor         => cctgtggttagctggttgcatatcc
  annotated miRNA   =>   TGTGGTTAGCTGGTTGCATAT
  sequence tr3: cc  =>   TGTGGTTAGCTGGTTGCATATCC
  sequence tr3: AT  =>   TGTGGTTAGCTGGTTGCAT
  

• s5: offset nucleotides at the begining of the annotated miRNAs:

  precursor         => agcctgtggttagctggttgcatatcc
  annotated miRNA   =>     TGTGGTTAGCTGGTTGCATAT
  s5                => AGCCTGTG
  

• s3:offset nucleotides at the ending of the annotated miRNAs:

  precursor         =>  cctgtggttagctggttgcatatccgc
  annotated miRNA   =>    TGTGGTTAGCTGGTTGCATAT
  s3                =>                     ATATCCGC
  

• type: mapped on precursor or miRNA sequences

• ambiguity: number of different detected precursors

Example:

seq                 miRNA           start   end     mism    tr5     tr3     add     s5      s3      DB amb
TGGCTCAGTTCAGCAGGACC    hsa-mir-24-2    50      67      0       qCC     0       0       0       0       precursor 1
ACTGCCCTAAGTGCTCCTTCTG  hsa-miR-18a*    47      68      0       0       0       tG      ATCTACTG        CTGGCA  miRNA 1

Web-servers¶

TFmiR: disease-specific miRNA/transcription factor co-regulatory networks v1.2. It uses results from UP/DOWN regulated miRNA/Genes and allows to focus in only one disease to create different type of relationships between miRNA/TF/Gene. Easy to use. Probably need to filter the output sometime due to the big networks that can result from an analysis.

Diana-TarBase v7.0: Database for validated miRNA targets. Many filter options. Good for small candidate miRNAs set studies.

StarScan: Database to browse the targets of miRNAs from degradome data. It has a fancy interface, and many species and data from GEO.

miRtex gives targets from literature. Good for finding validated targets to help discussion in papers or further functional experiment based on new hypothesis.

piRBase: Database for piRNA annotation and function. Published last year, for now the best I can find out there.

chimira: Web tool to analyze isomiR. It gives you a quick idea of you samples.

MicroCosm: MiRNA target database. Updated and download option.

IsomiR Bank: isomiR database from many species and tissues. For single queries is useful.

Command-lines¶

miRVaS : tools to predict the functional changed due to nt changes in the miRNA sequence.

IsomiRs¶

miR-142-3p isomiR: “We furthermore demonstrate that miRNA 5′-end variation leads to differential targeting and can thus broaden the target range of miRNAs.”

A highly expressed miR-101 isomiR is a functional silencing small RNA.

A challenge for miRNA: multiple isomiRs in miRNAomics.

miR-183-5p isomiR changes in breast cancer. Validated target regulation of new genes different from the reference miRNA.

A comprehensive survey of 3’ animal miRNA modification events and a possible role for 3’ adenylation in modulating miRNA targeting effectiveness.

PAPD5-mediated 3′ adenylation and subsequent degradation of miR-21 is disrupted in proliferative disease.

High-resolution analysis of the human retina miRNome reveals isomiR variations and novel microRNAs.

General¶

A novel piRNA mechanism in regulating gene expression in highly differentiated somatic cells.

Differential and coherent processing patterns from small RNAs to detect changes in profiles of processing small RNAs.

Targets¶

Identification of factors involved in target RNA-directed microRNA degradation.

Techonolgy¶

miRQC: work studying the accuracy and specificity of different technologies to detect miRNAs.

Important features affecting the detection of small RNA biomarkers: How the sample can affect the detection of biomarkers (like RIN value, concentration, ...)

Comparison of alignment and normalization . I will take the message that TMM and DESeq/2 normalization are the best to avoid strong bias if we consider to have a small proportion of DE miRNAs. For the alignments, here you have another comparison for miRNAs annotation: https://rawgit.com/lpantano/tools-mixer/master/mirna/mirannotation/stats.html

review of tools for detect miRNA-disease network.

review of tools for miRNA de-novo and interaction analysis