1. Finding causes in Neuroblastoma genomics data

Introduction to bioinformatics tools using the web-based genomics analysis and visualization platform R2 and neuroblastoma data

This resource is located online at http://r2-training-courses.readthedocs.io

1.1. Introduction

1.1.1. Neuroblastoma

Neuroblastoma is a childhood tumor of the peripheral sympathetic system. Primary tumors typically occur in the adrenal glands, which are located on top of the kidneys, but they can also manifest in other areas of the body. Most patients are diagnosed between the age of 0 to 4 years.
Neuroblastoma patients are classified by The International Neuroblastoma Staging System, the INSS. The increment in stages does not reflect progression of disease, as is the case for colon cancer, but represents different characteristics of the disease. Stage 1, 2 and 3 neuroblastomas have a very good prognosis. Stage 4 neuroblastomas usually go into complete remission upon therapy but often relapse as therapy-resistant disease. About 40% of stage 4 neuroblastoma have an amplification of the MYCN oncogene. This implies that instead of two DNA copies, each neuroblastoma cell has 30 to 100 copies of the MYCN gene. In addition to stage 1, 2, 3, and 4 neuroblastomas, there is the unusual stage 4S neuroblastoma, which is metastasized but goes into spontaneous regression. Over the last twenty years, the outcome for stage 4 neuroblastoma patients has not substantially improved.

Figure 1: Low stage versus high stage neuroblastoma

Figure 1: Low stage versus high stage neuroblastoma

Several cancer types, such as pancreatic cancer and colon cancer, follow a multi-step process involving a series of genetic alterations, similar to Vogelstein’s multi-step genetic mutation framework as shown below. These mutations accumulate over time and lead to the transformation of normal cells into cancerous cells.

Figure 2: Mutation paths during cancer progression.

Figure 2: Mutation paths during cancer progression

Extensive research into possible mutation mechanisms in neuroblastoma has been done (also in the AMC Oncogenomics group). Unfortunately, such a mechanism has not been found for this often deadly childhood tumor.

1.1.2. Finding differences and biological processes

One of the technologies that can be used to study a disease or biological process is gene expression profiling. With this high throughput technology, we determine the mRNA expression of nearly all genes known in a single experiment.
We will look for different subgroups in our data, find genes that make a difference and find cellular pathways that are activated in neuroblastoma patients with an unfavorable prognosis.

1.1.3. Research questions

During this practical course, we will use the R2 bioinformatics platform to study two research questions:

  • Which genes make a difference between neuroblastoma subgroups?
  • Which molecular pathways are activated in neuroblastoma patients with an unfavorable prognosis?

1.1.4. Go to the R2 platform

You’re now on the R2 main page. This web based molecular biology data analysis platform contains a wealth of data and methods to analyze the datasets. Step by step, researchers are guided through a web of options for data analysis. R2’s main page shows this principle: step through each of the numbered boxes to develop your analysis of choice.

The in this document will open up a Google form, one per section, with which you can submit your answers.

1.2. Finding prognostic factors in your data

Data used:

  • 88 Human Neuroblastoma samples (Tumor Neuroblastoma public - Versteeg - 88 - MAS5.0 - u133p2)

Techniques used:

  • mRNA Microarray expression

Analysis used

  • Find Correlating Genes with a single gene
  • Finding Differentially expressed genes
  • Kaplan Meier by annotated parameter


1.2.1. Selecting datasets

Let’s first make sure that the correct dataset is selected.

  • On the main page, you find a menu in the middle of the page that consists of several boxes.
  • Verify that in box 2 the following dataset is selected: Tumor Neuroblastoma public - Versteeg - 88 - MAS5.0 - u133p2. If not, follow the steps below.

If you see a different dataset selected, you can change the dataset as follows:

  • Click on the selected dataset, which will make a grid popup, showing all the datasets that are available to you within the platform. Each row is one dataset and each column shows particular characteristics of that dataset. You can filter for specific datasets with the headers of the columns.
  • In the textfield of column Author type Versteeg and in the column of number of samples N type 88.
  • Click on the row of the correct dataset and then on the blue button to Confirm Selection

1.2.2. Investigating single genes / Expression of key genes

  • The button below brings you to the form in which you can submit your answers for section 1.2.



It is known that the amplification of the MYCN gene is associated with a poor prognosis. Now we will analyze the mRNA expression of MYCN in the selected dataset. We will use the R2 analysis module ‘One Gene View’.

  • On the main page in field 3 select View a gene, which is also the selected analysis by default. Then click Next.
  • For this analysis, R2 needs to know which gene/ reporter to use. Fill in the text field Search by Gene the gene name of our interest: mycn. And select with a mouse click the mycn gene from the dropdown in order to select the correct reporter.
  • The rest of the settings we leave as is. Click Submit

The scatter plot in the central diagram shows the mRNA expression values of MYCN for all tumors of the 88 patients, next to each other. The samples are ordered according to increasing MYCN expression.
  • The expression is given on a log scale by default. To switch to a linear scale, scroll down to the Adjustable Settings box under the plot. In the field Transformation, select None and click Submit.

In R2 the samples of a dataset can be annotated with meta-information, e.g. clinical data or molecular biology parameters. Each group of annotated data is called a “Track” in R2. These tracks can be informative / useful for a large variety of additional analysis and visualization functions. To name a few, tracks can be used to filter datasets, to compare groups of samples, to color scatter plots of samples with meta information, or to correlate genomics patterns in your data to e.g. different phenotypes or demographic characteristics.

  • Directly underneath the plot the tracks of this dataset are shown with their color coded values. You can see the tracks for e.g. agegroup, INSS stage and MYCN amp(lification). Hover with your mouse over the different blocks.

The track mycn_amp shows which samples have MYCN amplification.


_images/R2d2_logo.pngWhat is the relation between the track MYCN amplification and MYCN expression?

_images/R2d2_logo.pngWhat do you notice about inss staging versus mycn amplification when you look at the annotation underneath the graph?


We can show such relationships more clearly by grouping the tumors in the graph according to a property in a track. Go to the Adjustable Settings box underneath the plot. At the top of that box, you can adjust the Analysis type with a dropdown.

  • Change this setting from single gene to gene vs track.
  • In the Adjustable Settings box you now have to choose a Track by which you want to separate the samples. In the drop down menu, select mycn_amp (2cat). And click Submit. Check the plot.
  • In this view, the samples are not ordered by their MYCN expression value by default. If you would want to adjust this, you could use the Extra Graph Option and choose the value Track and Gene sort. Always click Submit to see the results of adaptations in the settings.

    Additional options to get familiar with settings in R2:
  • Try it out with a different track: select Track inss(2 cat) and click Submit.
  • Also try out the Graphics options. For instance, try Graph type Box and Add Scatter True to see the data points in the box plots. You can use Color mode (groups) Color by Track to add the color of the track categories to your boxes. Don’t forget to Submit.
  • Lastly, when you click on the gear icon in the left top corner of the graph, you can either save your plot, copy it to clipboard. You can alsouse the General tab to find options to adapt the graphic parameters such as font size, axis-with etc. Most of these options are directly responsive, so you don’t have to hit a Submit button. If you don’t see changes, or the plot needs a bit of space reorganisation, you can hit the redraw plot button. Feel free to try them.


Now we will use links on this page that lead to additional analyses.
  • To learn basic properties of the gene, look at the one line table above the Adjustable Settings box.
  • Click on the link in the ‘Gene ID’ column. This brings us to the National Center for Bioinformatics database in Bethesda, USA.


_images/R2d2_logo.pngWhat is MYCN?


In the top right corner of the R2 mycn expression plot page, links bring you to other recourses.

  • Click on the hyperlinked MYCN below Pubsniffer. In a new tab, Pubsniffer shows how many papers in PubMed mention MYCN in their abstracts or the words ‘MYCN’ and ‘neuroblastoma’. Click on the number to see these papers.
  • Go back to the open tab with the grouped plot for MYCN expression of our dataset.

All the tumor samples in our dataset were analyzed by making sections of frozen tumor tissue and selecting sections with more than 80% tumor cells under the microscope. The sections, from which the RNA used for this microarray analysis was isolated, can be seen in the Sample overview.
  • Under the header Data set in the top right corner, click on Sample overview. All samples of the dataset are available via the dropdown.
  • Select ITCC0001 and click on View Sample. Play with the magnifications in the top left corner of the sample image (2x, 10x, 40x)
  • Go back to the tab with the grouped plot of the MYCN expression (N.B. you may close the other tabs)
  • Below the graph you can click on an arrow to View additional details. For the MYCN gene, the table ‘Alternative Reporters’ shows that this gene is represented by 5 sets of reporters (probe sets) on the Affymetrix U133 Plus 2.0 microarray. Take a look at the signal intensities for the different reporters (the red numbers in brackets). It’s good to realize that genes could have more than one reporter for a given platform in this case the Affymetrix platform. Also in RNAseq datasets, if transcripts (isoforms) are annotated, this can be very relevant. By default, R2 chooses the reporter with the highest signal which is in ~99.9% of the cases the most representive for a gene.
  • For many platforms which use reporters the genome location is also added. Click on the MYCN link in the ProbesetVerification box leads you to the genome browser where the exact location of the reporters can be investigated.


_images/R2d2_logo.png Do you think it is a wise idea to average the signal for MYCN over all the reporters and why?


1.2.3. Finding Correlating genes

Many approaches have been conducted to target the MYCN gene. Historically as a transcription factor MYCN has been regarded as “undrugable”. A way to identify downstream targets of MYCN which may be potentially drugable, is to identify genes which show a similar expression pattern. In R2 these can be identified by using the ‘correlating genes with a single gene’ analysis module.

  • Go back to the main page using the link Main in the upper left corner of the page.
  • In the main menu select in box 3 Find Correlated Genes with a single Gene and click next.
  • Provide the MYCN gene in the Search by Gene field and make sure to click on the reporter in the dropdown.
  • In the ‘Corr. p <= cut-off’ field , change 0.05 to 0.01 (0,01 if your Windows system is set the Dutch) and click Submit. In the next screen a set of tables is generated; one table for negative and one for positive correlating genes.

_images/R2d2_logo.png Approximately, how many genes were found by the test?


  • All identified genes in the table are linked to a detailed view. Choose a gene and first hover your mouse over the magnify glass symbols in the View column and then click on the icon. Do this with one gene for both tables to get an impression of the graphs for (inverse) correlated with the MYCN gene.

While leaving the tab open with the tables of (inverse)correlating genes, you can close the detailed plot tab(s) again after you have looked at the graph(s).

Inspecting genes one by one quickly becomes a dull task. We can also analyze the complete results with several of the provided analysis options on the right. Let's have a look at the chromosomal locations of our identified genes for this section. For this type of question, R2 has the 'Chromosome Map'.
  • Click on the Chromosome map in the right menu and investigate the result table. If you realize that MYCN is located on chromosome 2, did you expect to see the result you obtained (which chromosome has the most significant p-value)?
  • One of the nice features in R2, is that you can easily explore results further. Go back to the correlation analysis page and scroll to the bottom. Here you can make adaptations to the analysis.
  • To gain more insight in what might be going on, in the adjustable settings menu of the correlation analysis result page, change the correlation direction to only negative and click submit. Perform the chromosome map analysis again.

_images/R2d2_logo.png Where are overrepresented genes primarily located with respect to their chromosome location?


In neuroblastoma, at the DNA level, MYCN amplification and loss of 1 copy of the chromosome 1p arm is a well established connection. It is described in literature that a number of tumor suppressor genes are located on chromosome 1p. 1p loss of heterozygosity (LOH) is frequently observed in MYCN amplified tumors. Interestingly, we can even ‘see’ this loss in the mRNA profiles, since at least a proportion of the genes show reduced expression in patients with elevated MYCN expression.

1.2.4. Finding differentially expressed genes

We have seen that MYCN expression has a clear preference for some chromosomal regions in the previous analysis. Next to looking for patterns that resemble the expression of a gene, you can also investigate the expression patterns between groups of patients in a differential expression analysis.

Can we find biological processes by looking at differentially expressed genes between groups? For example, which genes are differentially expressed between the ‘alive’ or ‘dead’ group? Let’s have a look:

  • In the main page menu, select in box 3 section Differential expression, Differential Expression between two groups and click next. In the next screen, use the T-test which is selected by default and click alive (2 cat) in the pulldown menu at Group by and click submit.
  • In the next adjustable settings panel all kinds of settings can be adjusted but for now select no and yes for subsequently group 1 and 2 and click Submit.


_images/R2d2_logo.png How many genes are differentially expressed between the alive “no” and “yes” group?


In the generated gene list you can find marker genes or genes playing a key role in neuroblastoma. However, it would be of interest to find out if certain genes are overrepresented in e.g. biological processes.

  • In the right menu next to the genelist, click on the button Geneset analysis.
  • In the next screen select KEGG pathways in the Gene set collection pull down and click next.
  • In the next screen you find a collection of annotated KEGG genesets in which the genes from the analysis are overrepresented.

_images/R2d2_logo.png Which KEGG pathway is the most significant?


Click on the blue A which leads the user to an annotated KEGG map where upregulated genes are colored green and downregulated genes are colored red.
Unlimited DNA replication is one of the characteristics of malignant cells. Therefore, some DNA replication proteins are considered as promising cancer biomarkers.


_images/R2d2_logo.png Can you find the function of the MCM2-7 complex in the picture? For which group of our analysis are these genes upregulated? (Look at the color coding on top of the table)


  • Also try the other blue H.

The heatmap visualization produces a grid in which the samples are placed horizontally, and the genes from the KEGG gene set that you clicked on are placed vertically. One little colored square represents the expression value (transformed into a z-score) of the respective sample for the respective gene. A high z-score is here colored in red, a low z-score in blue, as you can see from the color bar underneath the heatmap. This way we can see which samples show relative low expression of the gene and which ones a relative high expression of the gene.You can see from the colorscheme which samples show relative low expression of the gene and which ones a relative high expression of the gene. You can also see that samples group together that show similar expression for certain genes. A hierarchical clustering algorithm looks at the z-score profiles of the samples and calculates which samples show similar profiles. The heatmap is reordered, such that the samples that show similar expression profiles cluster together and the ones that show different expression are positioned further away from each other. The same holds true for the genes. The clustering algorithm shows the genes together that show similar behavior in the cells.

  • Look at the high vs low expression groups, and their values of the inss and mycn_amp annotation above the graph. See if the clusters that you find in the heatmap correspond to these tracks.

1.2.5. Using annotation and the Kaplan Meier curve

We have seen that annotation tracks in R2 can be used to group the samples in a dataset for comparative analysis of gene expression. Another valuable use of tracks is to evaluate their prognostic value with a Kaplan Meier curve. We will now take a look at the above mentioned dataset that consists of 88 human neuroblastoma samples.
This dataset is annotated for a number of clinical and molecular parameters. We will analyze the prognostic value of stage, age at diagnosis and amplification of the MYCN oncogene.

  • On the main page, use the dropdown in box 3 to select the correct Kaplan Meier analysis: Kaplan Meier by annotated parameter. Click Next.
  • We are going to separate the patients based on the INSS staging system, that was explained in the introduction of neuroblastoma in the beginning. Choose for Type of Survival the value overall-c1103. For the Track setting, select the value inss (cat 5). Click Next.
  • The Kaplan Meier curves appear.


_images/R2d2_logo.pngWhat does a drop in the curves mean? And the little verticle tick-mark on the horizontal parts of the curves?
Scroll over the drops and the tick-marks of the curves to see clinical details of patients.



Below the graph, you can change several settings in the Adjustable Settings box. If you change settings, don’t forget to click the button Redraw Graph.

  • Now select agegroup as prognostic value under de setting Track, click Redraw Graph.
  • Do the same for MYCN amplification with the track mycn_amp (cat2).

_images/R2d2_logo.pngDo you observe a significant difference between the groups?


1.2.6. Kaplan Meier: Validating prognostic factors such as gene expression

Go back to the main page. We will now investigate if MYCN expression can be used to segregate our patient cohort in 2 groups that coincide with a difference in survival percentage.

  • In box 3, select Kaplan Meier by Gene Expression and click next.
  • Enter mycn and select the mycn gene in the Search by Gene field.
  • Choose for Type of Survival the value overall-c1103 and click Next
  • Read above the graph which cut-off value is used to obtain two groups of high and low MYCN expression.

The Kaplan Meier Scanner is a powerful tool in R2. This tool analyzes the prognostic value of the expression level of an individual gene. In contrast to e.g. staging or MYCN amplification, expression data are not discrete (yes / no amplification), but a continuum of values.

The question then is, which expression value to take as a cut-off point to separate groups with ‘high’ and ‘low’ expression. In many publications, the median value is taken as a separation between high and low expression. This however does not take into account that potentially a sub-group of the patients drives the separation. Within R2, we can make use of the to called KaplanScanner. Within this analysis, every expression value in a series is used as cut-off point (scanned) after which the value that gives the most significant discrimination in a good and bad prognosis group is chosen.


_images/R2d2_logo.png What is the survival chance for high and low MYCN expression, as assessed by the kaplan scan (use the extreme right values)?


Now let’s compare this survival analysis to the survival analysis with a median or average expression value cut-off point. Underneath the graph you will find the Adjustable Settings box, where you can adjust your analysis.

  • Switch to median in the Cutoff mode dropdown menu. Click Redraw Graph
  • Now analyze survival when we use average MYCN expression as Cutoff mode. Don’t forget to redraw the graph.

_images/R2d2_logo.png Which method gives the clearest prognostic groups? Is there a consequence for the statistics and P-value of the scanning method?



1.2.7. Using tracks as result of an analysis

If you are ‘logged in’ with an account in R2, then you can also create personal grouping variables, that can be used later on. In this way you can extend R2 with information that is useful for yourself.

  • Switch back to the scan mode and Redraw, for the analysis below.
  • Another approach to find possible regulating genes is to use the groups based on the mycn expression cut-off value for further analysis. Below the Kaplan Meier graph, click on the “store as track” button.

In the next screen all the individual samples are listed each assiged to the “low” or “high” group. At the bottom you can store the two groups based on the Kaplan Meier. In this example we will store this track as Temporary (24hrs) but you can also store this track permanent.

  • Click on the Build set button. Now the track is stored for further usage.

Let’s find genes that differ between the two groups that we just found with the Kaplan Scan:

  • Go back to the main page. Select the Differential Expression between two groups.
  • On the next page select the grouping variable that you just stored from the Kaplan Meier Scanner. If you can’t remember the name but didn’t change the naming you will find the track underneath the header temp_24hrs and it will be called kaplanscan-mycn (2cat) and click submit.
  • In next the screen select the low and high grouping variables for Group 1 and Group 2 and click Submit.
    Now a list of differentially expressed genes has been found based on the Kaplan Meier most prognostic MYCN values cut-off.


    Again, this result can be taken for further analysis:
  • On the right side of the page, click on **Plot all genes (xy, volcano etc)
  • Change plot type to Vulcano plot
  • Hover over a few of the dots with the highest -log10 pvalue and the highest and lowest logfold.

  • Notice how, if you click on the dots with the left mouse button, they get marked, and a menu on the left appears with which you can change their colors.
  • Notice how, if you click on the dots with the right mouse button, you get the detailed gene expression values plot, grouped by the chosen track.
  • Go back to the result page of the “Differential Expression between two groups” analysis and in the Adjustable settings menu set Max number of results to 100 to get the top 100 differing genes.
  • After the new result list has come in, hit the button Store result as custom gene set.
  • Type in a name, e.g. kaplanscan_mycn_deg and click Save gene set.

Tracks and gene sets that are generated as a result of an analysis can be stored and used throughout the many R2-analysis modules in R2. You can see in them in overview and edit them in User Options:

  • On the main page, go to User Options from the menu on the left, and choose Custom gene sets. The gene set that you just made, should be visible under _TEMPORARY.
  • Go to User Options again and choose Tracks and then Manage Temp Tracks. Since tracks are dataset dependent, you choose te dataset that we have been working with, and click Next. You should find the kaplanscan_mycn track listed.

1.3. Different expression patterns between subgroups and the underlying biology

We have seen several characteristics of neuroblastoma patients that act as prognostic factors of survival chances, such as MYCN amplification, age and INSS staging. The patient group with the aggressive subtype of neuroblastoma, INSS stage 4, shows a high percentage of relapses after treatment. Initially responsive to therapy, there is a seemingly complete remission of the tumor. Unfortunately this is often followed by a relapse that is resistant to therapy. Children developing a relapse almost always die.

Figure 3: Bright field image of isogenic cell line pairs.

Figure 3: Bright field image of isogenic cell line pairs.

The big question now is: if and why a few neuroblastoma cells are able to escape the treatment.

New neuroblastoma patient material showed interesting morphological features in the tumor tissue. From several neuroblastoma patients multiple cell lines were obtained from the same biopsy. These cell lines share genetic defects and are therefore called isogenic cell line pairs.

Remarkably, in each patient two morphologically differing cell types were observed. For three patients (identified by the numbers 619, 717 and 700) a microscopy image of each pair is provided below.

Figure 4: Bright field image of isogenic cell line pairs.

Figure 4: Bright field image of isogenic cell line pairs.

  • The button below brings you to the form in which you can submit your answers for section 1.3


Two images on a row belong to one patient (e.g. 619-MES and 619-ADRN). As you can see, the first image of the first patient shows a strong resemblance to the first image of the other two patients. Same for the second image.


_images/R2d2_logo.pngWhat do you note about the morphology of the cell lines?


Most neuroblastomas are located in the abdomen and are associated with adrenal glands or sympathetic ganglia. Cells of the sympatho-adrenal lineage develops from the neural crest, undergoing an Epithelial-to-Mesenchymal Transition as shown below.

Figure 5: Development of the sympatho-adrenal lineage from the neural crest

Figure 5: Development of the sympatho-adrenal lineage from the neural crest

We profiled the mRNA expression of genes using Affymetrix mRNA chips of these pairs and of a previously established neuroblastoma cell line that after culturing gave rise to two very divergent phenotypes. We are now going to take a look at the differentially expressed genes between the morphologically differing cell types.

Data used:

  • Cell lines recently derived from three different patients. Two morphologically different looking cells were taken per patient. This dataset is combined with two classical Neuroblastoma cell lines that showed this different morphology as well.

Techniques used:

  • mRNA Microarray expression

Analysis used

  • Toplister: unsupervised gene selection
  • Unsupervised hierarchical clustering
  • Heatmap visualization

For this analysis we will use one of the analysis tools of R2: Toplister. The toplister is a handy tool for a first start to analyze a given dataset. Particularly, if you have poor dataset clinical annotation. The Toplister can assess which genes show the most different expression values throughout a dataset. It does so by selecting the genes whose expression values have the largest standard deviation within a given set of samples. This provides an unbiased view of the differences in gene expression.

  • On the main page, click in box 2 on the selected dataset. The dataset grid pops up, showing all the datasets that are available to you within the platform. Let’s use the headers of the columns to filter for the dataset that we need in the next section.
  • In the textfield of column Author type Versteeg and in the column of number of samples N type 8.
  • Click on the correct dataset. Optionally read the dataset details in the information box underneath the grid
  • Select the dataset for further analysis by a click on the blue button Confirm selection
  • Confirm that in box 2 the dataset ‘Mixed Neuroblastoma (MES-ADRN) - Versteeg - 8 - MAS5.0 - u133p2’ is selected, containing 6 recently patient derived cell lines (2 per patient) plus the 2 classical Neuroblastoma cell lines.


1.3.1. Finding and visualizing genes that behave differently within a dataset

Genes that have a large variation in gene expression within one dataset are possibly interesting to look into. They might be able to explain different phenotypes within the dataset.

  • Select Toplister (Gene filter stdev) as the type of analysis in box 3 from the dropdown menu (- scroll almost all the way down. You will find it listed under the header Meta analyses). Click Next
  • With the default settings, Toplister will find the 100 genes that have the largest variation in gene expression among these 8 cell lines. Leave the settings as is, and click Next again. A list of genes appears.

_images/R2d2_logo.pngDo you recognize any genes from figure 5 when you scroll down through the list? I.e. genes that come into play in the development of the sympatho-adrenal lineage from the neural crest?


On the right side of the page you can find several buttons that allow you to perform further analyses with the list of genes that you just obtained with Toplister.

  • Click on Heatmap(zscore)

The 100 genes of the Toplister result list are shown on teh vertical axis, the samples on the horizontal axis. The unsupervised clustering algorithm reordered the samples and the genes again according to their z-score values.


_images/R2d2_logo.pngRoughly how many groups of samples do you see in the heatmap, showing similar expression profiles within that group? Is this what you expected?


Above the heatmap you can see two tracks of this dataset: cell_type and pairs. The cell_type track shows for each sample to which of the two morphological subtypes the sample belongs. The pairs track shows to which patient the sample belongs.


_images/R2d2_logo.pngWhen you hover your mouse over the squares of the tracks, you can see the detailed information of the sample. What feature determines the clustering of the samples?

(N.b. the color choices of the tracks are independent of the color scheme that is used for the heatmap grid. Check what the blue color and the orange color signifies in the cell_type track)


1.3.2. Which genes make a difference? Creating signatures

We have identified two different types of cells that occur within the same patient. Neuroblastoma apparently has a heterogenous nature. What genes determine the difference between the two types? We’ll use RNA expression data again but now we will use a predefined, supervised classification in groups to search for genes that characterize this classification best, or in other words, that are differentially expressed between these two groups.

Data used:

  • Mixed Neuroblastoma (MES-ADRN) - Versteeg - 8 - MAS5.0 - u133p2 (same as above)
  • Gene Ontology
  • Broad curated hallmark datasets

Techniques used:

  • mRNA arrays

Analysis used

  • Differential Expression: supervised gene selection
  • Gene Ontology Analysis: overrepresentation calculation


  • Go to the main page of R2 and confirm that in box 2 still the dataset ‘Mixed Neuroblastoma (MES-ADRN) - Versteeg - 8 - MAS5.0 - u133p2’ is selected.


  • In Field 3 choose Differential expression between two groups and click Next

This dataset has been annotated with ‘cell type’ information. Each sample was assigned to either the MESenchymal or the ADReNergic cell type. The information is stored in R2 in a track.

  • Choose the proper track in the Group by dropdown and click Submit. An additional adjustable settings menu pops up.
  • Choose one of the types for Group 1 and the other for Group 2.
  • Since we have only 8 samples make sure that the Corr. voor multiple testing is set to No correction. (More information on Correction for Multiple Testing can be found here) and click Submit.
  • A list of differentially expressed genes appears with correlation p-value < 0.01 in this dataset. Click on the magnifying glass icon in front of a gene of your choice to see its expression in the sample set.
  • Go back to the tab with result page of the differentially expressed genes. This is still open in one of your browser tabs. Try an oppositely correlating gene as well.
  • Again go back to the result page.
  • Click on the Heatmap(zscore) button in the right menu panel; a heatmap shows the expression of the differentially expressed genes for each sample.


_images/R2d2_logo.pngHow is this figure different from the former?



For future use, this list of genes has been stored for you in R2 as saved genesets. The list has been split into two genesets: one set of genes that is highly expressed in the MES type of samples (r2_mesadrn_mes) and one set of genes highly expressed in the ADRN type of samples (r2_mesadrn_adrn).

On the result page of the differential expression analysis, from the right panel of menu buttons, R2 provides several additional analyses for the list of genes that we just generated.
As a next analysis step, we can check a data resource called the Gene Ontology that provides a tree of systematically ordered biological terms that is used to formally describe the biological role of each gene.
The Gene Ontology Analysis tool in R2 calculates for each of the thousands of groups of genes that are annotated with a specific biological term whether your set of choice is over-represented in them.
  • On the page with the differentially expressed genes, select the Gene Ontology Analysis button in the menu on the right




_images/R2d2_logo.pngWhat can you say about the function of the differentially expressed genes?



  • Now scroll down to the end of the page (or click the filter button in the left upper corner of the page) and adapt the settings such that only the genes that are higher expressed in the MES type of cells are selected (check the adrn < mes). Click Redo analysis.


_images/R2d2_logo.pngWhat can you say about the function of the group of genes that are upregulated in the MES type of cells?



In R2 there are many more sets of genes that have been found to be implemented in specific processes. The Broad Institute has compiled quite some of these sets of genes that characterize hallmark biological processes.

  • Go back to the result page with the differentially expressed genes.
  • Select the Gene set analysis option from the right menu
  • Select the Custom collection and then for Gene set Broad 2020 09 h hallmark, Confirm selection and click Next


_images/R2d2_logo.pngWhich hallmark category of genes pops up as most important? Can you explain this?



As mentioned above, the lists of differentially expressed genes between the MES and ARDN groups are also stored in the gene sets databases of R2. We can use these genes as a proxy for mesenchymal- or adrenergic-ness. Let’s first look at these 2 groups of genes in a heatmap, when these are applied to a dataset where a number of neuroblastoma cell lines are represented next to the 8 samples that we have looked at above.

  • Go to the main page and click on the dataset that is currently selected.
  • A dataset selection grid pops up in which you can use the filters on top of the columns to find datasets of interest. Select ‘Mixed Neuroblastoma (MES-ADRN-CREST) - Versteeg/Etchevers - 34 - MAS5.0 - u133p2’ as a dataset to explore (e.g. fill in the grid crest in the textfield of the Tissue/Tumor column and click on the correct row and apply your selection with the Confirm selection button)
  • Now select View Geneset (Heatmap) from option box 3 and press next.
  • We click on the field behind Gene set, such that a grid opens where we can select a gene set.
  • Click on the arrow in front of r2 provided gene lists then the arrow of oncogenomics_groningen_NatGen_2017, next of functional_genesignature and then check r2_mesadrn_adrn as well as r2_mesadrn_mes. Press Comfirm selection and Submit.

A heatmap is now shown on your screen. You should see a clear separation in the genes, that correspond to the 2 different gene sets.
The vertically aligned grey/red stripes on the side correspond to the gene set a gene was coming from. In addition, you see a clear separation in the samples too.


_images/R2d2_logo.pngWhat cell_type of the samples are assocciated with the 2 groups?


The current dataset also contains neural crest cells, which are naive (and still undifferentiated) cells from which neuroblasts develop.


_images/R2d2_logo.pngIn which cluster are the neural crest cells positioned, and does that make sense?


Using a heatmap as we have just generated can be very helpful in determining how a list of genes is behaving within a dataset. However, it is not very scalable and requires detailed manual inspection to interpret. Condensing the information from all genes in a single value, opens some new possibilities. Within R2, we define a gene signature as the average value of what you see in the columns for every sample (effectively the average of all zscores, within every single gene set). These are automatically calculated during heatmap analyses, and are represented in the bottom of the pictures. We can store these signature scores as tracks in R2, and subsequently use them in R2 as if they are a gene (a meta-gene). Within the current dataset, the scores for MES and ADRN are already provided as separate tracks. Under normal conditions, you would store the signatures as tracks in your personal account.

  • Go back to the main page. We will now look at the samples in this dataset as signature scores. To do this, we select relate 2 tracks and press next.
  • As track for X, we select s_mesadr_adrn, and as track for Y, s_mesardn_mes. Click Submit.

You should now see an XY plot where the signature values are used to position the samples.


_images/R2d2_logo.pngWhat is the relationship between MES and ADRN scores?


  • The samples can also be colored by a track or the expression of a gene. Set the Color mode dropdown to ‘Color by a Track’ and then select cell type (4 cat) as track. Click the Submit button.

The resulting graph is a very insightful representation that summarizes the 800 genes that we were looking at in the previous heatmap.

Within this dataset, also the 4 pairs are represented from the earlier analyses. We can make these stand out a bit by drawing lines between the samples from every pair. For this, R2 has sample_paths, which are comma separated samples that will be connected by a line. These can have a thickness, and a color associated, separated by a :. A new line can be started by using a ; .

  • within the Sample paths field paste gsm2413241,gsm2413246:#eeeeee;gsm2413239,gsm2413243:#eeeeee;gsm2413242, gsm2413245:#eeeeee;gsm2413240,gsm2413244:#eeeeee; and press Submit.
  • We can also make the samples stand out a bit better by marking them using the ‘samples to mark’. Paste the following into this field gsm2413241,gsm2413246,gsm2413239,gsm2413243,gsm2413242,gsm2413245,gsm2413240, gsm2413244:#222222; and press Submit.

We have now seen the very basic usage of signature scores.

In the signature plot we can again see that the MES cell lines are mingled with the neural crest cells, which are the cells from which neuroblasts can develop.

In the next analysis we are going to make use of a single cell dataset that describes the human adrenal medulla, a tissue that can generate neuroblasts. Single cell data is frequently depicted in 2 dimensional representations called UMAP , or tSNE. Both are dimensionality reduction algorithms that try to group cells together based on resemblance of their expression profiles. Within R2, such representations are available in the ‘sample maps’, that can be accessed from the main page in the left set of menu items.

  • On the main page, click on the Sample maps (UMAP/tSNE) and subsequently select the following resource: ‘Normal Adrenal Medulla - Westermann - 10739 - cp10k - 10x300hg19’ and make sure to select the Manuscript version.
  • The cells are not yet colored. So let’s stain them with the setting Color by a Track and Color track cell_type (11 cat).

_images/R2d2_logo.pngCan you see the route(s) of development in the representation?


  • There are also tracks defined for the MES and ADRN signatures in this dataset. Try both.

_images/R2d2_logo.pngDo the signatures ‘light up’ the expected regions in the UMAP ?


Resources:


R2 has many more options, that you can explore by yourself if you want. You can try to find a publicly available dataset that corresponds to your line of research. Try and see if some of the analysis pipelines from this practical could be applied to that dataset.

This concludes our series of tasks for today. If you would like to use R2 for your research in the future, then just visit http://r2.amc.nl and get started.