ClinOmicsTrailbc 1.0
A visual analytics tool for breast cancer treatment stratification using multi-omics data
Overview of ClinOmicsTrailbc:
Step-by-step tutorial for ClinOmicsTrailbc:
Click on the single steps to unfold detailed descriptions.
ClinOmicsTrailbc analyzes and integrates (epi-)genomic, transcriptomic and clinical data, which can be uploaded on the start page.
The minimal required input for a ClinOmicsTrailbc analysis is a score file containing differential gene expression values per gene, preferably z-scores of tumor vs control samples.
A: Additional genomic and clinical data can be provided by clicking on the respective icon in the panel. This will then unfold the respective upload prompts.
B: A data set from the user's local file system can be selected by clicking Browse.
C: For an exemplary run of ClinOmicsTrailbc using example data from a CAMA1 breast cancer cell line, click Load example data and run analysis.
A: After the selection of a file from the user's local file system, the data needs to be uploaded via clicking on Upload. After upload, the respective files are validated to make sure that they fulfill all formatting requirements.
Details on supported input types can be found here.
An example file per omics type can be viewed and downloaded by clicking on the question mark next to the data type's name.
B: P-values for the computed pathway activities of the sample under investigation can be derived from permutation testing. The number of permutations to be performed and a random seed for reproducibility of the results can be specified. Please refer to the documentation for additional details on the significance assessment.
Once at least a gene expression score file was uploaded, the Analyze data button becomes active. Clicking this button will start the annotation and analysis steps. The results page of ClinOmicsTrailbc will open in a new tab.
ClinOmicsTrailbc's results page is composed of seven main tabs that are listed at the top of the page (A-G). Each of the tabs contains different analyses and visualizations, focusing on different aspects of the tumor.
A: Overview
The sunburst chart provides an overview of relevant driver genes for breast cancer across the provided
data types. (cf. Step 3)
B: Subtyping
The (intrinsic) molecular subtype of a tumor is investigated based on rule-based and clustering techniques.
(cf. Step 4)
C: Pathways
The radar chart displays pathway activities for a curated set of breast cancer-relevant pathways, also in comparison to
other primary breast tumors and cell lines. (cf. Step 5)
D: Drugs
In-depth assessment of 17 standard-of-care breast cancer drugs. (cf. Step 6)
E: Driver events
Investigation of driver mutations and related driver targeting drugs. (cf. Step 7)
F: Immunotherapy
Assessment of suitability of immunotherapy and identification of neoepitopes for cancer vaccine design. (cf. Step 8)
G: More
Assessment of clinical trial eligibility and seamless integration with DrugTargetInspector and
GeneTrail2. (cf. Step 9)
The sunburst chart provides an overview of relevant breast cancer driver genes across all considered omics data types. Genes are grouped according to the pathways they are most characteristic for. The plot is organized in rings, where the innermost ring corresponds to the considered pathways, the second 'inner' ring displays gene expression scores. Depending on the data provided by the user, information on (differential) methylation, copy number alterations and the mutation status of the contained genes is shown in the third, fourth and fifth ring, respectively. Somatic mutations are color-coded by the predicted type of mutation. Clicking on a mutation of interest yields additional details on the contained mutations, including their severity according to SIFT and PolyPhen. The second most outer ring indicates whether the corresponding gene is an oncogene or a tumor suppressor. The outermost ring shows whether a gene of interest is a molecular drug target. Clicking on a drug target yields additional information on the targeting drugs as obtained from DrugBank.
A: Clicking on the light blue button provides a summary of all provided molecular, genetic and clinical data for the tumor sample under investigation.
B: Additional information on the composition and the interpretation of the sunburst chart is shown when clicking on How to interpret the sunburst chart?
C: The plot is fully searchable and extendable. In order to highlight single genes or whole pathways of interest in the visualization, enter their name into the search field and click on the blue search button. If a gene of interest is not yet contained in the sunburst chart, it is interactively added to the chart in a user-defined category.
D: The visualization is fully interactive. Hovering with the cursor over a gene or pathway will highlight the respective section of the plot. Clicking on the pathway ring will 'zoom' into this pathway and just display the pathway's genes for a more spacious representation.
E: In order to 'zoom out' of the pathway and hence return to the overview, the icon has to be clicked.
Here, the (intrinsic) molecular subtype of the considered tumor can be investigated.
Based on user-provided clinical information on the status of (hormone) receptors and the tumor's growth rate, a rule-based assessment of the molecular subtype is performed. For the four major subtypes Luminal A, Luminal B, HER2-enriched and Basal-like (triple-negative), the table contains information on whether or not the provided information matches the respective subtype and predicts the most likely subtype, in the exemplary case on the left Luminal A.
As the subtype-specific characteristics of tumors are also reflected in a tumor's gene expression levels, we compute a clustering of the sample under investigation in comparison to more than 500 primary breast tumor samples obtained from TCGA. To this end, we use the classic Principal Component Analysis (PCA) or t-distributed stochastic neighbor embedding method (t-SNE) [1], a non-linear dimension reduction technique that captures the similarity of samples in a two-dimensional space. The tumor sample under consideration is indicated by a blue diamond-shaped symbol. The reference samples are color-coded according to their molecular subtype as assigned by the PAM50 classifier [2].
The radar chart displays pathway activities of a curated set of 20 breast cancer relevant pathways. Each axis of the radar chart corresponds to a pathway. The larger the area spanned by this axis, the more active the respective pathway is in the investigated sample. As a reference, the pathway activity patterns of more than 500 primary tumor samples from the TCGA breast cancer cohort and 45 breast cancer cell lines [3] can be interactively added to the plot by selecting the corresponding sample's checkbox. The reference samples are sorted in decreasing order of similarity to the sample under consideration.
A: Additional information on the interpretation of the radar chart is shown when clicking on How to interpret the radar chart?
B: Clicking on a reference sample's name (i.e. its button) will open an extra view with additional information on clinical markers and in the case of cell lines additionally details on drug sensitivities and growth rates.
For a set of 17 FDA-approved, standard-of-care breast cancer drugs, ClinOmicsTrailbc assesses relevant biomarkers and the genomic and transcriptomic status of respective molecular drug targets, drug-processing enzymes, resistance-promoting factors and associated pathways.
A: Additional information on the interpretation of the drug assessment is shown when clicking on How to interpret the results?
B: The indicator marks show the estimated (effect on the) suitability of the considered drug. The single assessments are summarized per category of sensitivity-influencing factors and finally per drug. Please refer to the documentation for additional details on the rules applied for drug assessment.
C: Clicking on a mutation's indicator symbol opens additional details on the contained mutation(s), their estimated effect on the protein's function, as well as links to COSMIC and dbSNP for known variants.
Tumors can potentially contain a plethora of mutations that are usually divided into driver and passenger mutations, based on their impact on disease development. Driver mutations are thereby defined as those mutations that confer a selective growth advantage to the cell [4]. ClinOmicsTrailbc uses the IntOGen database for the identification of breast cancer specific driver and passenger genes (mutations). ClinOmicsTrailbc provides a prioritized list of putative driver genes contained in the tumor (sorted by their frequency of occurrence in breast cancer).
A: Additional information on the interpretation of the driver mutations is shown when clicking on How to interpret the driver mutations?
B: The user can choose to display only driver mutations as provided by IntOGen or to show both, driver and passenger mutations.
C: Mutations contained in the sample under investigation are color-coded by the estimated impact of the contained mutation(s) on the affected gene. Clicking on the indicator symbol will open a view with additional details on the specific mutation(s).
Besides the assessment on on-label drugs only, ClinOmicsTrailbc also investigates a set of 23 'driver-targeting drugs', i.e. FDA-approved drugs that require the presence or absence of pathological markers, mutations or other genomic alterations. These drugs are approved for various cancer types, however not necessarily for breast cancer, yet they could be considered for off-label use.
The total number of somatic mutations per coding region of the genome is defined as tumor mutational burden (TMB). ClinOmicsTrailbc computes the TMB as number of somatic mutations per megabase exon. The TMB of a tumor sample under investigation (red line) is then displayed in comparison to the TCGA breast cancer cohort (blue histogram). The single bars indicate the number of TCGA samples per interval of mutation frequencies (left y-axis). The TCGA samples are sorted by increasing mutation load. The black dots depict the logarithmized number of somatic mutations per megabase exon (right y-axis). Tumors with a high mutational load have the potential to respond well to immune-system inducing therapies like adoptive cell therapy or antigen vaccination.
Also, ClinOmicsTrailbc assesses the genomic and transcriptomic state of a variety of biomarkers for checkpoint blockade, along with their corresponding inhibitors as obtained from DrugBank.
As high mutation rates are oftentimes caused by deficiencies in the DNA repair machinery, ClinOmicsTrailbc also assesses the status of a variety of repair genes, curated by the MD Anderson Cancer Center.
A: As default, only those repair genes affected by protein-coding mutations are listed. When selecting the Show all checkbox, all repair genes with their corresponding gene expression and copy number scores are listed (if provided by the user).
B: The Mutation status column contains indicator marks color-coded based on the predicted severity of the contained mutation(s) on the protein function. Clicking on the symbol will open another view with additional details on the respective mutation(s).
Besides checkpoint blockade, personalized cancer vaccines are another promising approach to cancer immunotherapy. ClinOmicsTrailbc offers functionalities to predict potential neoepitope vaccine targets based on the identified somatic mutations and HLA genotype of a patient using the immunoinformatic toolbox ImmunoNodes. ImmunoNodes provides various classes of epitope prediction methods to compute (neo-)epitopes and to assess their affinity to the patient's set of HLA alleles. The identified epitopes can then serve as basis for the synthesis of a personalized cancer vaccine.
A: In a first step, the proteins to be considered for the neoepitope prediction need to be selected.
B: For epitope lengths between 8 and 17, the user can select between up to 13 neoepitope prediction methods. Based on the supported set of HLA alleles for a chosen method, the HLA type of the considered patient needs to be provided.
C: The resulting neoepitopes and their affinities to the considered HLA alleles are presented in a color-coded manner in the results table.
In cases where standard-of-care treatment solutions are not applicable due to e.g. resistance mutations or other hindering factors, it might be of interest to see whether there are ongoing clinical trials the patient might be eligible to participate in. Based on ClinicalTrials.gov, ClinOmicsTrailbc links to phase II, III, and IV clinical trials that are recruiting in a large variety of countries.
A: Click here to show clinical trials for breast cancer in the selected country.
B: Click here to show European clinical trials for breast cancer from the European Union Clinical Trials Register.
C: ClinOmicsTrailbc also makes a first assessment of the eligibility for various classes of clinical trials listed on BreastCancerTrials.org. This stratification considers tumor characteristics like the BRCA1/2 mutation status and the tumor grade, as well as different treatment types including hormone therapy, PARP inhibitors, targeted therapy and immunotherapy. Clicking on the icon in the last column will link to the respective trials.
ClinOmicsTrailbc is seamlessly integrated with its sister projects DrugTargetInspector [5] and GeneTrail2 [6] . Uploaded data sets can directly be forwarded to the respective analysis workflows by clicking the respective Start analysis button.
A: DrugTargetInspector will, if provided, use gene expression and mutation data to investigate deregulated molecular targets and their corresponding drugs, highlight recommended drugs for breast cancer and offer several downstream-analysis to assess the importance of a drug target of interest.
B: GeneTrail2 will use the user-provided gene expression data to perform enrichment analyses on a large set of signaling pathways and biologically relevant gene sets from a variety of databases.
Biblibgraphy
- Visualizing data using t-SNE Journal of machine learning research
- Supervised risk predictor of breast cancer based on intrinsic subtypes Journal of clinical oncology American Society of Clinical Oncology
- Subtype and pathway specific responses to anticancer compounds in breast cancer Proceedings of the National Academy of Sciences National Acad Sciences
- Cancer: drivers and passengers Nature Nature Publishing Group
- DrugTargetInspector: An assistance tool for patient treatment stratification International journal of cancer Wiley Online Library
- Multi-omics Enrichment Analysis using the GeneTrail2 Web Service Bioinformatics Oxford University Press