Time estimation: 1 hourLevel: Advanced AdvancedSupporting Materials:Last modification: Jan 24, 2023License: Tutorial Content is licensed under Creative Commons Attribution 4.0 International License. The GTN Framework is licensed under MITpurlPURL: https://gxy.io/GTN:T00156
In this tutorial, we will deal with:
- Processing many samples at once
- Getting data
- Creating a list of paired datasets
- Using collections
- Processing collection as a single entity
- Let’s look at what we’ve got!
- We did not fake this:
- If things don’t work…
Processing many samples at once
Here we will show Galaxy features designed to help with the analysis of large numbers of samples. When you have just a few samples - clicking through them is easy. But once you’ve got hundreds - it becomes very annoying. In Galaxy we have introduced Dataset collections that allow you to combine numerous datasets in a single entity that can be easily manipulated.
In this tutorial we assume the following:
- you already have basic understanding of how Galaxy works
- you have an account in Galaxy
warning At this time this tutorial is using Galaxy’s test server at https://test.galaxyproject.org. Once the main site is updated this tutorial will be edited.
In this history are a few datasets we will be practicing with (as always with Galaxy tutorial you can upload your own data and play with it instead of the provided datasets):
M117-bl_1- family 117, mother, 1-st (F) read from blood
M117-bl_2- family 117, mother, 2-nd (R) read from blood
M117-ch_1- family 117, mother, 1-st (F) read from cheek
M117-ch_1- family 117, mother, 2-nd (R) read from cheek
M117C1-bl_1- family 117, child, 1-st (F) read from blood
M117C1-bl_2- family 117, child, 2-nd (R) read from blood
M117C1-ch_1- family 117, child, 1-st (F) read from cheek
M117C1-ch_2- family 117, child, 2-nd (R) read from cheek
These datasets represent genomic DNA (enriched for mitochondria via a long range PCR) isolated from blood and cheek (buccal swab) of mother (
M117) and her child (
M117C1) that was sequenced on an Illumina miSeq machine as paired-read library (250-bp reads; see our 2014 manuscript for Methods).
Creating a list of paired datasets
The following wizard will appear:
In this case Galaxy automatically assigned pairs using the
_2 endings of dataset names. Let’s however pretend that this did not happen. Click on Unpair all (highlighted in red in the figure above) link and then on Clear link (highlighted in blue in the figure above). The interface will change into its virgin state:
Hopefully you remember that we have paired-end data in this scenario. Datasets containing the first (forward) and the second (reverse) read are differentiated by having
_2 in the filename. We can use this feature in dataset collection wizard to pair our datasets. Type
_1 in the left Filter this list text box and
_2 in the right:
You will see that the dataset collection wizard will automatically filter lists on each side of the interface:
Now you can either click Auto pair if pairs look good to you (proper combinations of datasets are listed in each line) or pair each forward/reverse group individually by pressing Pair these datasets button separating each pair:
Now it is time to name the collection:
and create the collection by clicking Create list. A new item will appear in the history as you can see on the panel A below. Clicking on collection will expand it to show four pairs it contains (panel B). Clicking individual pairs will expand them further to reveal forward and reverse datasets (panel C). Expanding these further will enable one to see individual datasets (panel D).
By now we see that a collection can be used to bundle a large number of items into a single history item. This means that many Galaxy tools will be able to process all datasets in a collection transparently to you. Let’s try to map these datasets to human genome using
Here is what you need to do:
- set Using reference genome to
- set Single or Paired-end reads to
Paired collection(blue outline);
M177-collectionfrom Select a paired collection dropdown (magenta outline);
- In Set read groups information select
Automatically assign ID(green outline);
- scroll down and click Execute.
You will see jobs being submitted and new datasets appearing in the history. IN particular below you can see that Galaxy has started four jobs (two yellow and two gray). This is because we have eight paired datasets with each pair being processed separately by
bwa-mem. As a result we have four
Once these jobs are finished they will disappear from the history and all results will be represented as a new collection:
Let’s look at this collection by clicking on it (panel A in the figure below). You can see that now this collection is no longer paired (compared to the collection we created in the beginning of this tutorial). This is because
bwa-mem takes forward and reverse data as input, but produces only a single BAM dataset as the output. So what we have in the result is a list of four dataset (BAM files; panels B and C).
Processing collection as a single entity
bwa-mem has finished and generated a collection of BAM datasets we can continue to analyze the entire collection as a single Galaxy ‘item’.
Ensuring consistency of BAM dataset
Let’s perform cleanup of our BAM files with
cleanSam utility from the Picard package:
If you look at the picture above carefully, you will see that the Select SAM/BAM dataset or dataset collection parameter is empty (it says
No sam or bam datasets available.). This is because we do not have single SAM or BAM datasets in the history. Instead we have a collection. So all you need to do is to click on the folder () button and you will our BAM collection selected:
Click Run Tool. As an output this tool will produce a collection contained cleaned data.
Retaining ‘proper pairs’
Now let’s clean the dataset further by only preserving truly paired reads (reads satisfying two requirements: (1) read is paired, and (2) it is mapped as a proper pair). For this we will use
Filter SAM or BAM tools from SAMTools collection:
parameters should be set as shown below. By setting mapping quality to
20 we avoid reads mapping to multiple locations and by using Filter on bitwise flag option we ensure that the resulting dataset will contain only properly paired reads. This operation will produce yet another collection containing now filtered datasets.
Merging collection into a single dataset
The beauty of BAM datasets is that they can be combined in a single entity using so called Read group (learn more about Read Groups on old wiki, which will be migrated here shortly). This allows to bundle reads from multiple experiments into a single dataset where read identity is maintained by labelling every sequence with read group tags. So let’s finally reduce this collection to a single BAM dataset. For this we will use
MergeSamFiles tool for the
Here we select the collection generated by the filtering tool described above in 3.1:
This operation will not generate a collection. Instead, it will generate a single BAM dataset containing mapped reads from our four samples (
Let’s look at what we’ve got!
So we have one BAM dataset combining everything we’ve done so far. Let’s look at the contents of this dataset using a genome browser. First, we will need to downsample the dataset to avoiding overwhelming the browser. For this we will use
Downsample SAM/BAM tool:
Set Probability (between 0 and 1) that any given read will be kept to roughly
0.05) using the slider control:
This will generate another BAM dataset containing only 5% of the original reads and much smaller as a result. Click on this dataset and you will see links to various genome browsers:
Click the Human hg38 link in the display with IGV line as highlighted above (learn more about displaying Galaxy data in IGV with this movie). Below is an example generated with IGV on these data. In this screenshot reads are colored by read group (four distinct colors). A yellow inset displays additional information about a single read. One can see that this read corresponds to read group
We did not fake this:
The two histories and the workflow described in this page are accessible directly from this page below:
From there you can import histories to make them your own.
If things don’t work…
…you need to complain. Use Galaxy’s Help Channel to do this.