Skip to content
Permalink
Fetching contributors…
Cannot retrieve contributors at this time
219 lines (175 sloc) 10.9 KB
layout title zenodo_link questions objectives time_estimation key_points contributors
tutorial_hands_on
Small Non-coding RNA Clustering using BlockClust
What do the read profiles of small non-coding RNAs represent?
How to cluster the read profiles based on some intrinsic features?
Difference between mRNA-seq and smallRNA-seq
Unsupervised grouping of the adjacent reads into read profiles
Look and learn what the shapes of the read profiles represent
Clustering of the read profiles by machine learning algorithm
1H
Small non-coding RNA read profiles are often associate with function
Clusters of read profiles correlate with the ncRNA classes
pavanvidem

Introduction

{:.no_toc} Small Non-coding RNAs (ncRNAs) play a vital role in many cellular processes such as RNA splicing, translation, gene regulation. The small RNA-seq is a type of RNA-seq in which RNA fragments are size selected to capture only short RNAs. One of the most common applications of the small RNA-seq is discovering novel small ncRNAs. Mapping the small RNA-seq data reveals interesting patterns that represent the traces of the small RNA processing.

For example, consider the miRNA biogenesis. The primary miRNA transcripts are processed by Drosha-complexes and results in hairpin precursor miRNAs. Then after they transported to the cytoplasm, Dicer slices off the hairpin. One of the RNA strands bound by Argonaute proteins regulates the target mRNA while the other strand is degraded.

The following figure (bottom part of the figure is taken from DOI: 10.1038/nrg3162) represents one of the small RNA processing patterns. The vertical block bars represent a mapped reads on reference genome. The height of the bars represents the number of reads mapped at that location. Reads in each bubble represent a read profile. In this case, they both are from miRNAs. From the small RNA-seq, we often see the two processed miRNA strands after mapping. The miRNA strand which targets the mRNA is expressed and we see more reads compared to the degraded strand. The gap between those two piles of reads represents the missing hairpin.

read profiles

In this tutorial, we will learn how to use BlockClust to cluster similar processing patterns together.

Agenda

In this tutorial, we will cover:

  1. TOC {:toc}

{: .agenda}

Preprocessing of the data

The data is we use is from human cell line MCF-7 sequenced in single-end on Illumina Genome Analyzer II. Reads are 36nt long and should be sufficiently long enough for this type of data. We start with a BAM file as an input for this tutorial. This BAM file is a result of mapping the reads on human reference genome build hg19 using segemehl. If you want to use your own raw sequencing data, please clip the adapters and map to the reference genome.

In the first step, we convert the BAM file into BED file. This is not a plain file conversion. Hence we cannot use any kind of BAM to BED conversion tools. We use BlockClust tool in the pre-processing mode for this purpose. The resulting BED file contains tags (a tag is a unique read sequence in a deep-sequencing library) and their normalized expression (in column 5), i.e. the ratio of the read count per tag to the number of mappings on the reference genome.

{: .hands_on}

Get data

{% icon hands_on %} Hands-on: Data upload

  1. Create a new history for this tutorial

    {% include snippets/create_new_history.md %}

  2. Import the files from Zenodo

    https://zenodo.org/record/2172221/files/GSM769512.bam
    

    {% include snippets/import_via_link.md %} {: .hands_on}

{% icon question %} Questions

Why is it mandatory to clip the adapters from the small RNA-seq data?

{% icon solution %} Solution

Generally, the processed RNA fragments from the small RNAs are about 18-30nt long. If the sequenced read length is longer than the fragment (which is usually the case), the sequencer reads into the 3' adapter.

{: .solution}

{: .question}

Sort BAM file

Before continue to BAM to BED conversion we need to sort the alignments in the input BAM file by their positions.

{% icon hands_on %} Hands-on: Sort BAM

  1. Samtools sort {% icon tool %} with the following parameters:
    • {% icon param-file %} "BAM File": GSM769512.bam

{: .hands_on}

BAM to BED of tags

Now it is time to do the actual conversion.

{% icon hands_on %} Hands-on: BlockClust preprocessing

  1. BlockClust {% icon tool %} with the following parameters:
    • "Select mode of operation": Pre-processing
    • {% icon param-file %} "BAM file containing alignments": output of Samtools sort {% icon tool %}

{: .hands_on}

Grouping of the adjacent reads into read profiles

Now we group the adjacent reads into so-called blocks and blockgroups using blockbuster tool. In general, each blockgroup should represent a single ncRNA read profile.

blocks of a blockgroup

{% icon details %} More details about the theory

The idea is to perform peak detection on the signal obtained by counting the number of reads per nucleotide. This signal, spanning adjacent loci, is then modeled with a mixture of Gaussians. An iterative greedy procedure is then used to collect reads that belong to the same block, starting from the largest Gaussian component, and removing them in successive iterations. The tool further assembles a sequence of adjacent blocks into a blockgroup if the blocks are either overlapping or are at a distance smaller than a user-defined threshold. A more detailed explanation is at the blockbuster website

{: .details}

Sort the BED file of tags

In order to run the blockbuster successfully the input BED file need to be sorted by chromosome, strand, start position and then end positions.

{% icon hands_on %} Hands-on: Sort the BED file

  1. Sort {% icon tool %} with the following parameters:
    • {% icon param-file %} "Sort Query": output of BlockClust {% icon tool %}
    • In "Column selections":
      • "on column": Column: 1
      • "Flavor": Alphabetical sort
    • Click on "Insert Column selections":
    • In "Column selections":
      • "on column": Column: 6
      • "Flavor": Alphabetical sort
    • Click on "Insert Column selections":
    • In "Column selections":
      • "on column": Column: 2
      • "Flavor": Fast numeric sort
    • Click on "Insert Column selections":
    • In "Column selections":
      • "on column": Column: 3
      • "Flavor": Fast numeric sort
    • "Ignore case": Yes

{: .hands_on}

Group reads into blocks and blockgroups using blockbuster

{% icon hands_on %} Hands-on: blockbuster

  1. blockbuster {% icon tool %} with the following parameters:
    • {% icon param-file %} "BED file containing read expressions": output of Sort {% icon tool %}
    • "Type of output": reads (blockbuster format)

{: .hands_on}

Clustering with BlockClust

Here we use BlockClust in clustering mode. All you need here is the output of the blockbuster.

Apart from clustering, BlockClust has built-in class specific discriminative models for C/D box snoRNA, H/ACA box snoRNA, miRNA, rRNA, snRNA, tRNA and Y_RNA. So it can also be used to predict if a read profile might belong to one of the known ncRNA class.

{% icon hands_on %} Hands-on: clustering

  1. BlockClust {% icon tool %} with the following parameters:
    • "Select mode of operation": Clustering and classification
      • {% icon param-file %} "Input blockgroups file": output of blockbuster {% icon tool %}
      • "Select reference genome": Human (hg19)
      • "Would you like to perform classification?": Yes
      • "Mode of classification": Model based

{: .hands_on}

The tool produces the following four output files.

The BlockClust: BED of predicted clusters file is the result of Markov cluster algorithm. Each entry in this file represents a blockgroup. The 4th column contains is in the form of annotation:blockgroup_id:cluster_id. The cluster_id represents which cluster the blockgroup belongs to.

The BlockClust: Model based predictions BED file is the result of classification. Each entry in this file represents a blockgroup. The 4th column contains the annotation of the ncRNA which it overlaps. For the blockgroups which are not overlapped with any known ncRNAs, the prefix predicted_ is added indicating that it is predicted by the BlockClust classification models.

The BlockClust: Hierarchical clustering plot file shows a dendrogram construced based on average linkage clustering.

The BlockClust: Pairwise similarities file contains the pairwise similarities of all input blockgroups.

Next, we will visualize the read profiles (from the BAM file) and the predictions (BED from BlockClust) together.

{% icon hands_on %} Hands-on: visualization

  1. Install IGV (if not already installed)
  2. Start IGV locally
  3. Expand the {% icon param-file %} GSM769512.bam file
  4. Click on the local in display with IGV to load the reads into the IGV browser
  5. Expand the {% icon param-file %} BlockClust: Model based predictions BED file
  6. Click on the local in display with IGV to load the reads into the IGV browser
  7. Go to the location chr4:90653059-90653141 in IGV.

{: .hands_on}

{% icon question %} Questions

  1. Do you see any already annotated transcript in that location?
  2. What does the read profile resemble?
  3. What did BlockClust predict?

{% icon solution %} Solution

  1. There is no annotated gene from RefSeq annotation on the same strand.
  2. It resembles the miRNA profiles mentioned in the introduction section. One of the read piles looks like mature miRNA and the other like miRNA*.
  3. Predicted as a miRNA.

{: .solution}

{: .question}

{% icon comment %} Comments

In order for this step to work, you will need to have either IGV or Java web start installed on your machine. However, the questions in this section can also be answered by inspecting the IGV screenshots below.

Check the IGV documentation for more information.

{: .comment}

Conclusion

{:.no_toc}

In this tutorial, we learned how to use BlockClust to cluster real small RNA sequencing read profiles based on their similarity. We also learned that the BlockClust can also be used to classify the read profiles based on pre-built classification models.

You can’t perform that action at this time.