As a result, 11 molecular subtypes have emerged that mainly reflect the cell of origin of the malignancy, either from a cell type at a specific developmental stage or a cell type with a defined function. On detail, the largest differences were observed between cancers of epithelial and non-epithelial origins, followed by epithelial cancers from basal layer-like cells against those with functions of secretory cells. This lead to tumor samples from the same tissue to split into different subtypes as well as samples from multiple tissues to coalesce in the same group. For instance, several tumor types were grouped in the squamous-like cluster, which appears to arise from an epithelial cell type common to diverse tissues that contain environmentally exposed epithelial surfaces (e.g. oral cavity, lungs, and bladder). On the other hand, bladder cancer was the most heterogeneous disease of the malignancies included in the study, and its samples mainly distributed across three subtypes that correlated with different clinical outcomes. Interestingly, the study also found particular alteration signatures shared by histologically distinct epithelial malignancies that can be exploited by the same therapeutic regimes and confirm that the basal-like breast cancer is a totally independent entity, as distinct to other breast cancer subtypes as it is to tumors from other tissues.

In addition to the results reported in the paper, one of the most interesting contributions of this work is the generated resource of omics data for more than 5000 tumors from 12 different cancer types using multiple platforms. All this data is now accessible into a unified resource in Synapse to support integrative bioinformatics analysis (at https://www.synapse.org/#!Synapse:syn2468297). In addition the results have been made available through several portals to facilitate their navigation, including the UCSC Genome Browser, Gitools, and MD Anderson’s Next Generation Heatmaps.

Our contribution on this part of the project has been to prepare all TCGA pan-cancer-12 datasets used in this study and their subtypes information ready to be navigated with Gitools interactive heatmaps. This data is available for download at http://www.gitools.org/datasets/pancancer12 and can also be opened directly from the web with the latest version of Gitools. In this heatmap (see figure below), the columns are tumor samples, rows are genes and each cell has multiple values indicating the mutation, copy number status of the gene in the tumor sample, and the expression and methylation level. Each sample is annotated with its tissue of origin and molecular subtype (see figure below) among many more annotation values that are available from within the heatmaps. Multiple options of Gitools, such as sort, filter, zoom, search, cluster etc. allows you to navigate interactively this large amount of data to extract meaningful information.

Screenshot of TCGA pancancer 12 dataset in Gitools.

If you want to get a glimps on how to navigate TCGA data in Gitools you can look the Gitools Videos, and if you want to learn all you can do with Gitools you can follow the Gitools – From A to Z tutorial.

We hope you find this new cancer genomics resource useful, and please let us know your experience navigating these data.

Low expression data events events ordered by mutual exclusion

Mutual exclusivity and co-occurrence statistics 

Selective pressure of cancerogenesis on the alteration of certain pathways is expected to result in a mutually exclusive alterations pattern within a cohort of cancer samples. Similarly, other alterations may be cooperating. In order to be able to test a hypothesis directly with the prepared cancer genomics heatmap datastes we implemented a weighted permutation approach to test for mutual exclusivity and co-occurence of alterations across a group of genes which is currently available as an additional step when sorting the genes by mutual exclusivity.

If you have loaded alterations data in a Gitools heatmap with genes listed in rows, you can test for mutual exclusivity and co-occurrence via the menu Edit->Rows->Sort by mutual exclusion dialogue. Just activate the option at the lower part of the dialogue (see Screenshot 1 below). After the test has been performed for the selected genes, the results will be shown to you in a new dialogue (see Screenshot 2). For more information, see the “Mutual exclusion and Co-occurrence Test” page in the documentation.

Data Events

The mutual exclusivity sorter expects binary data: events and non-events. Therefore we have introduced “Data Events” which is an automatic way to decide which data points are (positive) events, and which are not. Given that you are viewing test-results with the p-value color scale, all data points that fall below the selected significance cut-off are events. For more examples, see the “Data Events” page in the documentation. If need be and the current event function does not serve for your purpose it is easily switched for another one using the drop-down menu.

The summary of the selected genomic alteration data (left) shows that 14 % of the selected cases (4 genes and 306 samples) are altered. At the left we can see that the same samples and genes are upregulated in 128 cases (10%)

The same event functions are available at the new selection box, that appears always at the left of the heatmap if some rows and/or columns are in selected state. In the above image, we show two examples of the selection box that is summarizing the selected genomic alteration data (left) and the selected expression data (right). In the expression data, which is median-centered, all values above the scale cut-off, highly expressed genes, are set to be data events.

Screenshot 1: Mutual exclusive sorting and statistical test in Gitools.

Screenshot 2: An example of a mutual exclusive test result

Changelog Gitools 2.2.0 (July 8th 2014)

* Mutual exclusion plugin (new)
* SVG image export (new)
* Hierarchical cluster heatmap header with contextual menu (new)
* Selection statistics (new)
* Data events (new)
* Import string values: importing strings to categorical values (new)
* Layer groups (new)
* Recent files (new)
* New .mtabix compressed data file format (new)
* View enrichment data via contextual menu in results (new)
* View group comparison data via contextual menu in results (new)
* Parellization: use multiple cores for many tasks
* Many performance improvements
* UI improvements

Hierarchical clustering in Gitools: The lines in the heatmap header represent the hierarchical tree (Dendrogram) splitting at different levels. The root of the tree is located at the bottom, the leafs at the top. See video at YouTube

Perform a hierarchical clustering

Once we are viewing a data heatmap on which we would like to perform a clustering, we can achieve this by selecting the menu Analysis > Clustering. A wizard dialog will pop up that asks which clustering and how we want to perform.

See a short video on hieararchical clustering at YouTube: https://www.youtube.com/watch?v=jiQurFfE408

In the first step we want to select hierarchical clustering and if we want to cluster columns or rows. Furthermore if we were not viewing the values we want to cluster, we change the “Take values from” selection.

The second step gives us the chance to modify the distance measurement. If you do not have any special requirements, go ahead and directly hit the Finish button. After a moment the clustering should be finished and the hierarchical tree visible, explained in detail in the next section.

The result: Hierarchical tree, heatmap header, heatmap order and bookmark

One challenging issue with clustering in Gitools was how to show the clustering results while maintaining the interactive capabilities of the heatmap. We solved this by showing the hierarchical organization of the columns or rows as colored bars (see figure above). In addition the Newick tree is also provided.

The result of the hierarchical clustering are four things:

• A Hierarchical tree or Dendrogram. The tree is painted as a static image in a new tab. This image can be exported to an image via File > Export Hierarchical tree as image.
• A Heatmap header. As header of the heatmap, 10 levels of the hierarchical tree are added.
• Heatmap order. The order of the heatmap is changed according to clustering. While it respects the order of the splits, we calculate the in which order the different leafs of the tree should go by a density function. This guarantees that different clusters (or leaves) which are more similar to each other are placed one besides the next avoiding creating a perception of “artificially distinct” clusters.
• A Bookmark. After after applying the clustering order we add a one-dimensional bookmark to Gitools which contains the exact order of the clustered dimension (rows or clumns). In case the heatmap order has been changed this bookmark can be applied

How to interpret the hierarchical trees as colored bars/clusters

The header that has been added to the heatmap is a summary of the hierarchical tree. In the image above a hierarchical clustering header for the columns is seen. By default 10 different levels of the hierarchical tree are being displayed. Closest to the heatmap, at the bottom, the root level of the tree is shown. Each level towards the contains more splits.

Some properties to consider for interpretation the above image:

• Broad and detailed clusters. In the image above, each horizontal line represents the split branches of the hierarchical tree at different levels. You may decide which level(s) works best for you, depending on how many clusters you want to have. The more levels you move up, the more fine-grained the clusters are.
• Outliers columns become leafs fast. That means that after very few splits they are considered as clustered and therefore receive not any more clusters in upper levels. In the heatmap header this is reflected as white space. The left-most columns in the image, have become leafs after level 5.
• Peaks represent regions of higher similarity. As outliers soon receive no clusters, the columns with many clusters assigned (horizontal bars at many levels) are the ones that are more difficult to tell apart for the algorithm. As a consequence peaks or high plateaus form. In some cases these may be useful to tell apart with great detail two columns, in other cases the similarity is high enough so this information can be overlooked.

And remember, the whole heatmap with the hiearchical clustering header can be exported to an image via Export > Save heatmap to image.

Hierarchical Clustering in Gitools

The new version of Gitools (2.1) contains many internal changes and improvements for usability and analyses. Below we list the new main features which we will try to explain in more detail in upcoming posts.

• File import: Any flat text file containing values may now be imported to Gitools.
• Data integration: Via file import it is possible to load new data layers on an existing heatmap. You have a heatmap with methylation values and now the mutations have come in? Just load them as an additional value for each cell.
• Group Comparisons may now be done much more easily and for a series of groups at once. Group by annotations, values or without constraints and check for differences between two groups of samples.
  

  Gitools Bookmarks

• Bookmarks: Gitools heatmaps can now be bookmarked. Save a constellation of visible rows and columns in their order and the value being viewed.

• Clustering: The clustering code has been improved. We tackled in particular the difficulty of combining hierarchical clustering and interactive heatmaps. Besides that a classical newick tree is being generated, the heatmap is annotated with colored clusters representing the tree and its splits. Thus the order of the heatmap may be changed without invalidating the hierarchical tree.

There are many more things to discover in the details, especially within contextual menus.

If you want to try it yourself you can follow the step-by-step Tutorial we have prepared for VIZBI at http://www.gitools.org/vizbi

Let us know if you have questions.

Therefore it is a pleasure to announce that this year we have the opportunity to lead a tutorial session about Gitools and Nuria Lopez-Bigas will have a talk about the topic Biomedical Genomics Data on 4th and 5th of March 2014 in Heidelberg.

The tutorial will cover basic first steps with Gitools to visualize genomics data, as well as doing some simple analyses and reading data for publication. Take a look at the prepared Gitools Datasets in order to get an idea how the cancer genomics data can be visualized and explored in Gitools. The session will take place on Tuesday 4th of March during the afternoon session from 14:00 to 17:00. No programming skills are needed as we are planning to teach the following points to our participants:

1. How to load data into Gitools
2. How to obtain the visualization that fits your needs.
3. How to interact with the heat-map (sort, filter, move, cluster, search)
4. How to integrate different data dimensions.
5. How to visualize column and row annotations.
6. How to perform analyses over the heat-map (e.g. sample level enrichment analysis, group comparison, correlations and others).
7. How to load predefined interactive heat-maps of interest for cancer research, including multidimensional data from The Cancer Genome Atlas, Cancer Cell Line Encyclopedia and IntOGen.
8. How to export heat-maps into a publication-ready figure.
9. How to share your data in the form of interactive heat-maps.

Important Links:

• VIZBI 2014 Registration
• VIZBI 2014 Program

Screenshot of Gitools 2.0 showing the TCGA pancancer12 drivers dataset

We have been interested for quite a while in the study of patterns of genomics alterations in cancer across tumor types. Thus a project like the TCGA Pan-Cancer provided a unique opportunity to apply our tools and expertise to a unique collection of data.

In the past few years we have developed computational methodologies to identify cancer drivers by analyzing the patterns of somatic mutations across tumors (i.e OncodriveFM and OncodriveCLUST) as well as tools to facilitate the visual exploration of multidimensional cancer genomics datasets (i.e. Gitools, IntOGen, see our review on this topic if you are interested in this), we now had the opportunity to apply those tools to TCGA Pan-Cancer data.

What does TCGA Pan-Cancer data consist of?

Integrated data set for comparing and contrasting multiple tumor types. Figure from Nature Genetics article. See details in NG.

The TCGA Pan-Cancer project assembled data from more than 3000 patients with primary tumors from different organs, covering 12 tumor types. In each of these tumors a number of omics technologies were applied to obtain complete genomics, transcriptomics, proteomics and epigenomics profiles of the tumors.

Collaboration and teleconferences

David Tamborero, Nuria Lopez-Bigas and Abel Gonzalez-Perez

The TCGA Pan-Cancer collaborative project works through regular teleconferences (usually on Thursdays at 2pm ET) with all the members of the consortium and collaborators. In each teleconference different groups present the results of their analyses. In our case, being in Barcelona, the time of the teleconference was quite inconvenient (8pm) for work-life balance, but alternating between David, Abel and myself we managed to attend most of the teleconferences and to present the progress of our work to other researchers several times.

Data and intermediate results generated by different groups are shared through the Synapse platform. There is a nice paper describing the use of Synapse for the collaborative work within TCGA Pan-Cancer (Omberg et al., Nature Genetics 45, 1125-1126).

Our Contribution

We tried, as much as we could, to use our tools and expertise to extract interesting knowledge from the valuable data generated by the TCGA consortium. In total we contributed to the project with 4 different results:

IntOGen-mutations

Authors of IntOGen-mutations. From left to right, Michael P. Schroeder, David Tamborero, Nuria Lopez-Bigas, Abel Gonzalez-Perez, Jordi Deu-Pons and Christian Perez-Llamas. Two more authors of IntOGen-mutations missing in the picture are Alba Jene-Sanz and Alberto Santos.

IntOGen-mutations is a web platform for cancer genomes interpretation. It not only analyses TCGA Pan-Cancer data but also additional datasets generated by other initiatives such as those included within the International Cancer Genome Consortium. In the current version users can retrieve driver mutations, genes and pathways acting on 4623 tumors covering 13 cancer sites. They are also able and to analyze newly sequenced tumor genomes and identify relevant mutations by putting them in the context of the accumulated knowledge. 

Probably the most interesting feature of IntOGen-mutations is that it provides a comprehensive view of cancer vulnerabilities across cancer types, which was not available before. Tumor re-sequencing projects usually report a list of cancer drivers identified with differing criteria and methodologies, which make it difficult to have a complete view of which genes are drivers in each cancer type. It is now possible to have this comprehensive view with IntOGen-mutations.

We have designed the IntOGen-mutations to be updated regularly and to be scalable to the analysis of much larger cohorts of tumors, so that we can keep up with the expected increase in the number of sequenced tumor genomes/exomes available. Thus with each update we will obtain a more complete view of cancer drivers across tumor types.

To know more about this project you can read this previous Blog Post.

Article: Abel Gonzalez-Perez, Christian Perez-Llamas, Jordi Deu-Pons, David Tamborero, Michael P Schroeder, Alba Jene-Sanz, Alberto Santos and Nuria Lopez-Bigas. IntOGen-mutations identifies cancer drivers across tumor types. Nature Methods, doi:10.1038/nmeth.2642 (2013)

TCGA data visualization using Interactive Heat-maps (Gitools)

One important challenge posed by the large and complex data generated by the TCGA Pan-Cancer project is how to provide access to researchers to explore it and extract useful knowledge from it. We have been working previously on this topic and we propose the use of Interactive Heat-maps (read more on that). We have prepared all TCGA data ready to be navigated with Gitools interactive heat-maps. See video below to learn how to use it.

Comprehensive identification of mutational cancer driver genes

One of the main advantages of analyzing the aggregated data of more than 3000 tumors across 12 tumor types is that it provides increased statistical power to distinguish driver mutations from passenger ones. In collaboration with other researchers of the Pan-Cancer project we have analyzed the mutational patterns of genes across tumors in the search of signals for positive selection that points to candidate cancer drivers. Integrating also additional data generated by the consortium has allowed us to obtain a reliable list of 291 mutational drivers acting in one or more of the 12 cancer types -accounting for 3,205 tumors. We have confirmed and extended the role of known cancer genes and we have identified novel candidates that complete the mutational landscape of these diseases. The article describing this work will be published next week. You can browse the results of this project in IntOGen (at http://www.intogen.org/tcga) and also using Gitools (at http://www.gitools.org/datasets). In addition the results are also available in Synapse (syn1962006).

Article: David Tamborero, Abel Gonzalez-Perez, Christian Perez-Llamas, Jordi Deu-Pons, Cyriac Kandoth, Jüri Reimand, Michael S. Lawrence, Gad Getz, Gary D. Bader, Li Ding i Nuria Lopez-Bigas. Comprehensive identification of mutational cancer driver genes across 12 tumor types, Nature Scientific Reports, 2n October, DOI: 10.1038/srep02650.

The mutational landscape of chromatin regulatory factors across 4623 tumor samples

Chromatin regulatory factors are emerging as important genes in cancer development and are regarded as interesting candidates for novel targets for cancer treatment. For this reason, and also due to previous interest in our group, we focused our effort to the study of the mutational landscape of these class of genes. For this we used all TCGA Pan-Cancer mutational data and additional datasets, summing up to 4623 tumors. You can read more on that in a recent post.

Article: Gonzalez-Perez A#, Jene-Sanz A# & Lopez-Bigas N. The mutational landscape of chromatin regulatory factors across 4623 tumor samples. Genome Biology 2013, 14:r106. # Equal contribution

What’s next?

The current state-of-the-art of the technology provides an unprecedented opportunity for the understanding of tumor biology. Most importantly, this should eventually lead to develop better treatments for the disease. At this moment, the bottleneck of oncogenomics is not to produce the data but to interpret it in order to retrieve useful knowledge. Initiatives as the Pan-Cancer project succesfully deal with the front-line of these analyses by sifting the important information from the huge amount of alterations that are observed in a tumor cell. Many challenges must be solved before all this information may improve the clinical management of cancer patients. Although many histories of success have been already incorporated to the clinical practice, now we are understanding the complexity of the disease and how many efforts are required to disentangle its mechanisms. It’s a long way to go, but we believe that we are walking in the good direction.

Also we want to announce that a new release of Gitools, version 1.8.4 is available. Some bugs related to GenomeSpace integration have been fixed, some memory managment improvements and the KEGG importer module has been rewriten to use their REST webservice instead of the deprecated SOAP webservice. You can browse all the solved issues here.

Since the Gitools users community is growing, we have set up a mailing list at Google Groups where you can sign up and ask any question. So we encourage to ask any question via the mailing list rather to mails to single members of our group. Also have in mind that BioStars is a good option to ask general questions about bioinformatics problems you may face.

Links and Related posts

Use case of Gitools in GenomeSpace 

How to visualize multidimensional cancer genomics data?

Gitools 1.8 for improved data managament and transfer 

Exploring multiple cancer genomics alterations with Gitools.

• Send matrices from IGV (Integrative Genomics Viewer) directly to Gitools
• Load matrices from GenomeSpace
• .GCT file support for matrix files (format used by GSEA).
• Import matrices from Excel data sheets

Go download latest version

New in Gitools 1.8.x: Support for loading data from Genome Space, importing from Excel and the .gct file format

GenomeSpace

GenomeSpace aims to be a platform that manages the researcher’s data and at the same time alleviates them from tedious tasks like converting data from and to different data formats. This approach holds a great promise to get standardizations for biological data and improved data transfer between different tools. We have enabled Gitools to log into GenomeSpace and load data. Tools that also connect to GenomeSpace are ArrayExpress, IGV, GenePattern, Cytoscape and more. See here for yourself.

IGV (Integrative Genomics Viewer)

Another great news is that the people at the Broad Institute have been working on a new feature for IGV which will let the user export loaded track data into .tdm matrix format and also load it directly in Gitools. For the latter, both Gitools and IGV need to be running so that they are able to communicate one with the other and transfer the data. Also note that sample information has to be loaded in order to let IGV know which tracks are from the same sample. This makes it possible for IGV to send matrices to Gitools that contain multidimensional genomics data as for example a matrix where each cell contains a value for expression, mutation and copy number variations for each sample and gene. As this is a new feature in IGV – and not in its definitive form – it is available through the early access versions which are preview versions presenting the features of the future releases. We see this as a great new feature that enables the user to work with two specialized and complementary visualization tools at once, which is enhanced by the link available in Gitools which navigates in IGV to a locus of interest.

Excel and .GCT

The .gct file format which is used by GSEA is now supported by Gitools for loading data to avoid the hassle of converting to .cdm. To the same end, Gitools can now import data directly from Excel data sheets. A wizard guides the user to choose which columns of the excel file correspond to the columns, rows and cells of the heatmap.

Related posts

Gitools 1.7.0: the new features

Exploring multiple cancer genomics alterations with gitools

Visualizing mutually exclusive alteration patterns in cancer with gitools

Gitools published in plos one

Gitools is an interactive heatmap viewer which can also perform various analysis over the data. Heatmaps in Gitools can be multidimensional, with various values per cell, which is very practical for cancer genomics data analysis and visualization (read more).

Let us introduce the new features step by step.

New data visualization features (screenshot)

IGV search (1 in screenshot)

Integrative Genomics Viewer (IGV) is a very powerful genomic browser that provides a complementary visualization of multidimensional genomics data to that of interactive heatmaps. Often it is useful to explore this type of data with the two tools at the same time. With that idea in mind we did a first step to integrate the two types of views. Using the IGV interface for other applications to send commands, Gitools can now locate any position within the human chromosome in your opened instance of IGV. In the screenshot we have the focus on the gene CPM (with the Ensembl id ENSG00000135678). If you click the link “Locate Id in genomic viewer (IGV)” in the left details panel, Gitools will locate the gene within IGV. If you have many genes selected, then Gitools will tell IGV to show them in split screen view.

Display name in color labels (2 in screenshot)

Adding data header from aggregated values (3 in screenshot)

A new type of header can be added now to columns or rows: Aggregated heatmap from matrix data. In the screenshot we can see at (3) the mean expression values, calculated with the mean aggregation method for each row. The aggregation of the values can be calculated for the whole row, just for some selected columns or according to column annotations, as it is the case in the screenshot. Note that even though the mean expression values are shown in the rows annotations, the main heatmap is displaying alteration events.

The aggregated values are represented by the color scale (chosen by the user) and optionally by value text labels. We can see that, in the highlighted gene, the expression mean between the classical and mesenchymal subtype of the glioblastoma brain tumour differ substantially.

Categorical scale and scale drawing (4 in screenshot)

We have added a new color scale to Gitools: the categorical scale. It is designed  to visualize categorical data, as shown in the screenshot. Each different data value is assigned to a color, which can be set by the user. The way that the color scales are drawn has been redesigned to be more intuitive.

Further features

Color scale memory: Gitools remembers to what data dimension which color scale has been selected. After having actively selected or modified a color scale, this very color scale will be selected after switching back to the data dimension.

Save analysis: When you perform an analysis “on the fly” in Gitools by selecting the menu Analysis -> Your analysis the save button on the top will be activated to let you save the analysis data and results to the hard disk.

Related posts

Exploring multiple cancer genomics alterations with gitools

Visualizing mutually exclusive alteration patterns in cancer with gitools

Gitools published in plos one

A gene-sample matrix is being converted to a gene-module matrix where module can be sets of genes like f.ex. pathways. The transcription level status can be used for stratifying and/or relating with clinical annotation

It is an easy way to stratify the samples into subgroups and/or relate the transcription level status of modules to clinical data. So this last week we have prepared a further video tutorial to show you how to perform SLEA easily with Gitools and gain more insight into your data.

Watch the video below or read the instructions in the fourth step of the Case Study: “Study multi-dimensional cancer data with Gitools”.

With this video tutorial we also release a new version of Gitools, version 1.6.2 so it is possible to have multi-value data matrices as input data for the enrichment analysis. Also we got rid of some bugs.

Download the latest version at www.gitools.org