Using the Compute data cube for hGSuite tool we will start by inputting the preprocessed hGSuite created in the previous step, and ensure that the correct genome assembly is selected.
We can then select one of four statistics, here we choose the ‘Coverage (Counts of elements)’.
We then need to select across which columns the statistic will be calculated, here we select both the genotype and the attribute respectively. We then leave the remainder of parameters as their default setting. Click here to view an example, el. 2 Once the data cube is completed, we can view it using the ‘eye’ button. Here we have several options. For the purpose of this example, where we seek to observe the differences in mutation type and quantity across genotype, we do the following: Set the ‘How to treat raw’ option as ‘Aggregate across this dimension’ using the ‘average’ method.
We now have a table and associated bar plot showing us the coverage of each mutation type within each genotype.

From here we can start doing some basic result interpretation.
The first observation that can be made is that there is little to no difference between the WT and the SMUG1 KO. We can also see that the UNG KO seems to generate the largest amount of mutations, and the CT-GA mutation is by far the dominant one. Finally, we notice that the UNG-SMUG1 KO has fewer mutations than the UNG KO alone.
From a biological perspective, we know that UNG is a DNA glycosylase which is responsible for the repairing of these mutations. It is therefore expected to see a large amount of C to T mutations when UNG is not present. It also appears that SMUG1 serves as a compensatory element in the absence of UNG, however it is error-prone and therefore generates further mutations.

The rainfall plot illustrates the genomic position (x axis) versus the genomic distance (y axis) where the genomic position represents the position along the genome, starting from the beginning of chromosome 1 (left hand side) to the end of chromosome y (right hand side).
The genomic distance is the distance between two points, with each point being a point mutation.
In practice, the rainfall plot allows for the visualization of the distribution and density of the mutations found within the data.
The ‘single’ parameter overlaps every track within the GSuite/hGSuite. As soon as more than 3-4 tracks are within the dataset, this plotting method has a tendency of being quite overwhelming.
The ‘multi’ method instead shows each track in it’s own area of the plot. Where each track has its own segment on the x axis. The multi method is intended to be used as exploratory while the single method is intended to be used as in detail analysis of a few tracks at a time. In order to better illustrate this, we go through the following example.
Note that the submitted GSuite or hGSuite must be in primary format.
We use the Generate rainfall plot tool to first generate a 'multi' rainfall plot

In this rainfall we see that the six groups to the left seem to be of larger interest than the remainder of the data. Within HyperBrowser and the full image the legend is made available, though it was too large to include in this figure. The six groups of interest from left to right are the following:
UNG AT-TA
UNG AC-TG
UNG AG-TC
UNG CA-GT
UNG CT-GA
UNG CG-GC
We therefore seek to create a ‘single’ rainfall plot for these six groups. In order to do so, we must first create a GSuite containing only these six groups, which in effect are all groups of the UNG knock-out.

We use the filter hGSuite according to metadata tool to split our hGSuite per genotype.
We start by submitting our primary hGSuite to this tool, select the division to be ‘by column’ and the selected column to be ‘genotype’ and execute the tool. This creates 5 separate hGSuite subsets, each specific to one genotype of our initial hGSuite.

We then use the hGSuite subset for the UNG genotype in the rainfall plot tool, this time selecting 'single' as the plot type

We see a large dominance of CT-GA mutations (green), of which these are found in high density within the 13th chromosome as well as two different locations within the 5th chromosome.
The rainfall plot within the HyperBrowser allows us to drag the cursor across the screen to select a zoom area. Below we show these zoomed areas respectively.

We then did another filter, this time extracting all CT-GA mutations from the tracks for all genotypes. We then submit this to another 'single' rainfall plot. We observed that the UNG KO was the only contributor to the CT-GA mutations found in peaks of interest in chromosomes 5 and 13.

Considering that we have now identified chromosomal locations of interest, we seek to know if there is any overlap between these significant regions and certain genomic regions such as exons or CpGs.
Note that the method illustrated below can be used as is (exons and CpG islands), or it can be adapted to include many other regions such as introns, promoters, or even specific genes.

1) Download tracks from UCSC

Here we download tracks from UCSC for exon regions and CpG island regions. However, if one were to want to make their own track, the format is quite simple. It must be a bed file containing three columns in the following order:
chromosome name (i.g. chr1)
start region (i.g. 36511763)
end region (i.g. 36513006)
These bed files are then uploaded to hGSuite using the ‘upload file’ tool.

2) Convert tracks to a GSuite

Here we simply create a GSuite using the Create a GSuite from datasets in your history tool.

3) Preprocess the gSuite

Done using the Convert GSuite tracks from primary to preprocessed tool .

4) Compute the data cube

We once again use the data cube, but this time the tool is named Compute data cube for relations between hGSuites. This version of the data cube allows us to compare two GSuites. We enter our original hGSuite (the one with the KOs and mutations) in the first slot and our freshly created genomic location GSuite in the second slot. We select the statistic method to be overlap.
It is important to leave everything to its default setting, as specifying columns will check for similar column name in both GSuites. In our case, we will simply be comparing tracks, hence the un-specification of the column names.
Viewing these results shows us a table and a bar plot quantifying each KO+mutation and the number of mutations which overlap with exons and CpG islands respectively.

5) Chromosomal frequency view

Using the plot frequency of mutation along chromosome tool, we can generate frequency plots for each individual chromosome. We can do so with a single GSuite, or use two GSuites.
The first track inputted will be shown in the bottom of the plot while the second track will be illustrated at the top. In order to input both GSuites we must select ‘Show results on one plot’ as the way of plotting. To reproduce the below image, ‘count point’ must be selected as the statistic.

Have chromo freq above
Here we observe that our peaks of interest in chromosome 5 primarily coincide with exon (blue line) and not CpG islands (orange). This was the expected outcome as we utilized whole-exome sequencing for this dataset.

Using the Compute data cube for hGSuite tool example we input our hGSuite and select the ‘Coverage (counts of elements)’ statistic. We then specify the three columns of interest, these are 'patientid','attribute', and 'primarytumorlocation'.

By aggregating 'raw' we can observe the mutation distribution per patient, that is shown in the plot above. We can also aggregate across hmfpatientid to obtain the mutation distribution across cancer types.

Due to the large amount of information, we first filter/divide our hGSuite (primary format) by cancer type in order to not overwhelm the rainfall plots. This is done using the filter hGSuite according to metadata tool and selecting the 'primarytumorlocation' as the filter column.
We then submit all three hGSuites to the Generate Rainfall plot tool and use multi as the plotting option.

Here each color represents a specific mutation type in a specific breast cancer patient. For example, the tallest peak (blue-ish in line with the 50G on the x axis) represents patient_006 and only the C>T;G>A mutations for that patient.
Using this type of plot we are able to look for possible abnormal mutation loads in patients. The applications of such a tool are great. For the purpose of the example, let’s focus our attention on that patient for now.

We use the filter hGSuite according to metadata tool and select 'hmfpatientid' as the column.
We then input the patient of interest (patient_006) to the rainfall plot and select the 'single' parameter.

What we can notice here is that there is a very large area of chromosome 6 that appears to be mutated. We also know that this is preferentially the CT-GA mutation.
An interesting question to ask would be if these mutations map to specific DNA repair pathways.

2a. Do the mutations occur more often in coding regions or non-coding regions of the genome?

We upload bed files with genomic positions for coding and non-coding regions for the hg19. The uploaded bed files were converted to GSuite using the tool Create or modify hierarchy of hGSuite tool hGSuite.
Further this data is preprocessed using the tool Preprocess a GSuite for analysis toolpreprocessed GSuite
To get the results and the plot we upload both the hGSuites, one for the coding and non-coding regions and the other containing the patient data along with the metadata that was preprocessed in the above step (for reference check out element 39 and 84). Compute data cube for relations between hGSuites tool Relational data cube This cube element gives you a general overview of introns vs exons in the data. But we can several question with this cube using the pivot functionality. Below is a set of questions and their answers that can be obtained using one cube. To do this we use the tool Compute data cube for relations between hGSuites toolRelational data cube After we upload the cube we can set the hierarchy based on the metadata. The figure below shows the selections that has to be made in order to explore the data in different dimensions SCREENSHOT OF THE SELECTIONS OF ELEMENT 98 1. Which mutation type across different cancer types occur most frequently in the coding regions? We can still use the previous element Relational data cube and transpose the table and plot to get the results for this question. The screenshot below on the left shows the results along with selections that was made to get the results. On the right is the transposed plot that appears when you click on the transpose option.

From the results shown below we can observe the mutation CT-GA occurs more often in general and also in the coding regions. This is as expected because mutation signature 2 which occurs most commonly in cancer patient is caused by the APOBEC cytidine deaminases that are predominantly characterized by C>T mutations.

The above images can also to transposed to plot the occurrence of different mutation types across cancer.