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

Gunes Gundem and Nuria Lopez-Bigas. Sample level enrichment analysis (SLEA) unravels shared stress phenotypes among multiple cancer types. Genome Medicine. 4:28 doi:10.1186/gm327

In this manuscript we introduce SLEA, which we have described earlier in this blog, and we use it to explore the interrelation of different stress phenotypes in multiple cancer types. We also ask if these phenotypes could be used to explain prognostic differences among tumor samples.

First we do SLEA using Gitools with the set of genes related to Chromosome Instability (CIN genes) in a breast cancer dataset (Ivishina et al., 2006). Next we use the result of SLEA to stratify the tumors, and find that tumors with upregulation of CIN genes have worse prognosis than the others (see figure in the left).

Heatmap of tumor samples as columns and gene modules related to stress response phenotypes as rows. Enrichment is shown with colors from blue (down-regulation) to red (up-regulation) while gray values indicate no significant deviation from the expected median value.

We then ask what is the relation between the expression of CIN genes and other genes related to stress phenotypes. For this we perform SLEA with gene sets related to several stress phenotypes and we find that the tumors with over-expression of CIN genes displayed a transcriptional program pointing to evasion of the senescence barrier and particular stress phenotypes, indicating strong interdependencies between these different pathways (see figure in the right). We corroborate this relationships in 11 different datasets of different cancer types. The results of the 11 datasets analyzed can be browsed in the supplementary web browser.

We also perform a robustness analysis of the SLEA method to find that SLEA results are robust for datasets that include more than 80 samples. Overall this article shows that SLEA enables the identification of gene sets in correlation with clinical characteristics such as survival, as well as the identification of biological pathways/processes that underlie the pathology of different cancer subgroups. To learn more about SLEA you can visit http://bg.upf.edu/slea.

Michael P. Schroeder and Abel Gonzalez-Perez

This year our lab will participate again in the Gulbenkian Training Program in Bioinformatics (GTPB). Abel and Michael will give a 3 days course on Bioinformatics for Integrative Genomics – BIG12, from 7th to 9th May. The course is now open for Inscriptions, and a tentative program can be found here.

GTPB program is very well known for its practical courses on bioinformatics, which have been running since 1999. The courses are eminently practical, including few lectures and many hands on exercises. The groups are small, usually up to 20 attendees, which allows a very intense experience with the topic, the instructors and the rest of attendees.

BIG12 course will focus on the analysis of multidimensional and heterogeneous data in biomedical genomics. The use of microarray and next generation sequencing technologies yields complex, multidimensional data sets that describe in detail the myriad of changes that occur within individual cells in complex diseases such as cancer, and how these changes differ between patients, cells or conditions. Bioinformatic skills are required to analyse this data and extract useful knowledge out of it. The course will address questions like: Which of the long list of mutations detected are likely to affect the function of the protein and which ones are probably neutral?; Which of the genes affected by those mutations are already known to be involved in cancer or other diseases?; Which pathways or biological processes are affected by the transcriptomic alterations detected in my experiment? How can we integrate diverse data types from a cohort of patients (including clinical data, transcriptomic data, mutations etc.)?

The course will teach how to answer these questions using bioinformatics approaches. In particular we will extensively use three tools developed in our lab: IntOGen, Gitools and Condel.

Some of the topics that will be addressed are describe in the following posts. Take a look at them and if you’re interested come join us in the course!

Exploring multiple cancer genomics alterations with Gitools

Visualizing mutually exclusive alteration patterns in cancer with Gitools

Exploring the effect of cancer genomic alteration on expression with Gitools

Sample Level Enrichment Analysis (SLEA) in Gitools to assess the transcriptional status of pathways per tumor

Three questions you can answer with IntOGen

IntOGen Biomart portal

Condel in the world of web services

The making of Condel (CONsensus DELeteriousness Score)

Basic and intuitive analysis of microarray datasets using Gitools (Part 1)

Basic and intuitive analysis of microarray datasets using Gitools (Part 2)

One important thing to note is, that saved color scales have to be created again. But now there should be no problem with further modifying loaded color scales.

Furthermore, we adapted the Heatmap Properties Panel to the left, so that it does not occupy too much space and is resizeable for OS X and Xfce users.

A list of all bug fixes can be found here.

The effect of genomic alterations can be observed in the expression values. Note for example that samples with loss of CDKN2A shown lower expression values than samples without this alteration. This effect is also evident for the alteration of the other genes.

Today we present to you a way to explore the effect of CNAs on gene expression on a cohort of cancer samples with the help of Gitools. To do this you first need to prepare a multi-dimensional genomic data matrix containing both CNA and expression values. We have described before how it can be loaded into Gitools – see our case study six: Studying multi-dimensional cancer data with Gitools, and our previous blog post: Exploring multiple cancer genomic alterations with Gitools.

Once you have the genomic alterations and expression data in Gitools you can see if the samples with a particular type of genomic alteration (eg. gain) have different expression values than those without this alteration. In some cases this can be easily viewed in the heatmaps (see figure above), however if we want to corroborate this observation statistically we can also use the new Group Comparisons analysis, which we have added in our 1.6.0 release of Gitools, to test if the expression values in the samples with the genomic alteration are different than in those samples without this alteration.

The guide on how to explore the effect of alteration on expression can be found as a new chapter of this case study where it is explained step by step – and it also contains the following video:

The heatmap in the left shows copy number alterations of TCGA Glioblastoma project in the KEGG TP53-signalling pathway. If sorted properly we can observe that the upstream genes show a mutual-exclusive alteration pattern, but not PTEN and CDK4. Loss in blue, gain in red.

Since this feature is so striking we thought it was useful to incorporate a new sorting option in Gitools to automatically sort a heatmap by mutually exclusive patterns. This makes it easy to find mutually exclusive altered genes in a heatmap of genomic alterations in several tumors. With the new version of Gitools we released recently (Gitools 1.6.0) this features is now available to everyone.

Coupled with this post we have added a further tutorial (Finding and visualizing mutually exclusive genes) to our latest Case Study to show how to we take advantage of this feature using a clear example from the TCGA Glioblastoma data. Please see the figure and read the legend for details and watch the video tutorial to understand how it is done.

Cancer genomics data that is produced creates multi-dimensional data sets. Gitools lets you browse all that data at once.

A typical cancer genomics project nowadays screens the cancer genome, epigenome and transcriptome of a cohort of patients and identifies various types of alterations: Copy Number changes, Somatic Mutations, Gene Expression changes and others. This is the case of projects framed within The Cancer Genome Atlas or the International Cancer Genomics Consortium, as well as many others. Each of these types of alterations is represented in different data formats and it remains a challenge to integrate them to get a unified view of the process of alterations that leads to tumorigenesis. In Gitools it is possible to explore and analyze multi-value matrices in the form of interactive heatmaps, making it possible to work with various data dimensions at once. The interactivity of heatmaps in Gitools aids in the exploration of those large multi-dimensional data sets effectively, and the thoroughly customizable visualization options allow to display each alteration type in the most convenient way.

Rows and columns can be annotated by colors and text. In this example the columns are annotated by vital status and Glioblastoma subtype (top-down) and the rows are annotated by gene symbol, chromosome and band (left-right).

One of the advantages of having different dimensions in the same heatmap, like for example Copy Number Alterations (CNA) and Expression in the same file, is that after filtering or sorting on one dimension (e.g. CNA) it is possible to observe the effect of this filtering or sorting on another (gene expression changes).

Another additional layer of information in cancer genomics data is the clinical features related to the donors of the tumor samples (eg. tumor subtype, age, sex). Similarly, genes may also be linked to various sets of annotations (eg. chromosomal location, pathway), and it is very useful to visualize those annotations along with the heatmap. In Gitools a multitude of annotations can be added in the form of text or color bars (see image in the right).

We are preparing documentation in the Gitools help site to better describe these capabilities of the program. Take a look at the section “How to Browse Multi-dimensional data”. We are also preparing a new Case Study (Studying multi-dimensional cancer data with Gitools) with some video tutorials. The first one in this series (embedded below) shows how you can browse multi-dimensional data.

Multi-value data matrix (TDM)

TDM file format is a tab delimited file that has contains multiple values per row (eg. gene) and column (eg. sample). The first line is a header line following a line for each cell.

In this following example we see a .tdm-file that contains three columns and two rows.

The change from the 1.5.x to 1.6.x series promise new features, and so it is! In general, Gitools got polished to make it easier to work with multi-dimensional data. Check the list below to see what’s new

Usability:

• Color scales can be saved and loaded!
• Display options for cell value/color scale are kept: when you define the display options of a value those are saved, so that when you switch back to see that value again the display options are kept.

Data analysis:

• New Analysis: Group Comparison (Mann-Whitney-Wilcoxon), to compare the distribution of values between two sets of columns or rows
• New sorting method: Mutual exclusive sorting

Plus several bugfixes.

Expect some follow-up posts explaining some of the novelties in detail.

From an expression profile of a set of tumor samples, in Gitools you can perform SLEA to assess the transcriptional status of modules (ie. pathways) per sample.

The identification of molecular biomarkers from expression data is a major objective in cancer research. It is clear that there is a benefit in pathway biomarkers (ie. measuring the activity of the pathway instead of individual genes). One easy way to analyze the transcriptional status of pathways (or other gene sets) is using Sample Level Enrichment Analysis (SLEA) in Gitools. This way you can assess the status of each pathway in each sample. This can be used to identify tumor subtypes and to correlate molecular features with clinical features.

It is easier to explain it with an example:

Using a dataset of 156 lung tumors and adjacent normal lung tissue samples (from Hou et al 2010), I did a SLEA with Gitools to find pathways that are significantly up or down-regulated in different samples. The result is a big heatmap with samples as columns and pathways as rows. Each cell contains the result of the enrichment analysis for a particular pathway in a sample. The interactive capabilities of the Gitools heatmap viewer helps to intuitively interpret the results. For example, in the following figure samples are ordered according to their clinical annotation and we can see very clear differences between normal and tumor samples for a selected list of pathways. For instance, apoptosis genes and genes involved in MAPK signaling pathway tend to have lower expression values compared to normal samples, while cell cycle genes tend to have higher expression values in the tumor samples represented in the dataset.

If you want to learn more and even try it yourself, you can follow this Gitools tutorial.

Heatmap of tumor samples as columns and pathways as rows. Enrichment is shown with colors from blue (down-regulation) to red (up-regulation) while gray values indicate no significant deviation form expected median value. Samples are ordered according to clinical annotation as follows: squamous cell carcinoma (SCC), adenocarcinoma (ADC), large cell carcinoma (LCC) and normal lung tissue samples (Normal)

As of now you can download the new version of Gitools – version 1.5.10 – which keeps up with the newest changes of the biological databases KEGG and Ensembl.

As some may have read or heard, KEGG has gone through some changes last year and so has the accessibility of its data. This change broke the functionality of the KEGG importer in Gitools. This is now fixed!

Additionally we also made an update for the Ensembl importer and added the new versions in the archive to make sure all data can be comfortably downloaded through Gitools.

That’s it for now, expect more news to come!