A list of the modied genes carrying a non-synonymous variant

in the tumor can be more deeply interrogated by clicking on

the gene name to view the detected mutations in a genome
browser window, together with annotations from databases
such as COSMIC. The gene list can be viewed in table format
and sorted in terms of the number of molecular pathways they
potentially constitute. Drugs that potentially target the protein
products of mutated genes in the sample tumor are also
presented, with links to DrugBank.
Molecular pathway data allows large numbers of mutated genes
to be viewed within the context of synergistic relationships.
Melanoma Proler identies molecular pathways constituted
by genes modied in the tumor, together with any statistically
signicant pathway enrichment within the list of genes and
targeting drugs.
Melanoma Proler generates diagrams showing tumor variant
genes in the context of molecular signaling pathways, using
KEGG pathways included in the R graphite program. In the
KEGG pathway named melanoma, 13 genes (highlighted in red)
are modied in the sample tumor, in agreement with the known
importance of the signaling events represented in this pathway to
melanoma formation and progression.
The mutations in the molecular pathway gene list may be viewed
in waterfall plots and heat maps that show the individual tumor
of the particular patient in relationship to other tumors in the
MelanomaDB database. These plots are also annotated with
information such as relationships between RNA expression and
patient survival, which mutations have been identied as driver
mutations in melanoma, DrugBank drug targets, genes that
encode proteins which are targets of drugs that are in current
melanoma clinical trials, and genes that are predicted to encode
druggable proteins.
The MelanomaDB study allowed us to look across more than
300 melanoma tumors at the distribution of the number of
genes altered across tumors. Melanoma Proler begins by
contextualizing the data from the tumor sample in comparison
with these >300 patients with melanoma analyzed by exome
sequencing.
MelanomaDB
Scatter Chart
Variant Analysis
Pathway Analysis
Pathway Visualization
Waterfall Plots and
Heat Maps
Introduction
Clinical practice and research using genomics now increasingly require the interpretation of tumors from single patients. While
pipelines for identifying variants from DNA sequence reads are maturing, interpreting the consequences of these variants
remains a difcult and time-consuming task. Compounding this difculty, most available software and expertise for interpreting
these variants focuses on populations in large research studies, rather than individual patients in the clinic.
We have developed the Melanoma Proler web tool to facilitate interpretation of the biological, clinical and pharmacological
consequences of tumor variants for individual patients. Melanoma Proler provides a framework to understand the
consequences of tumor somatic mutations in several ways:
1) Categorize the mutations based on the biological functions and molecular pathways constituted by the proteins encoded by
the modied genes.
2) Generate associations between gene expression and patient outcome.
3) Generate associations between molecular pathways affected by modied genes and drug targets.
4) Allow clinicians to visualize where their patients tumor sits in the context of a large curated database of other melanoma
tumors.
Methods
Melanoma Proler is based on the MelanomaDB database, curated by researchers at the University of Auckland, New
Zealand and their international colleagues [1]. MelanomaDB brings together exome sequence and clinical data from over 300
melanomas and gene set data such as DrugBank drug ID targets [2], known melanoma drugs, drugs in clinical trial, druggability
indices from in silico calculations [3], and various literature links to melanoma and RNA expression in tumors associated with
patient outcome [4]. Features of MelanomaDB most pertinent to N=1 analysis of melanomas from individual patients were
translated into the Melanoma Proler web application. For more details on methods, see reference [1].
Melanoma Proler integrates with Illuminas BaseSpace computational platform, taking as its input the somatic variant analysis
(VCF) le produced by exome sequencing analysis of a tumor-normal pair.
Both MelanomaDB and Melanoma Proler use open source resources provided by the research community, including signaling
pathway information contained within the R graphite package [5]. The database and web application are in turn freely available
for researchers to use at:
MelanomaDB: http://genesetdb.auckland.ac.nz/melanomadb
Melanoma Proler: https://apps.biomatters.com/melanoma-proler
Results
A typical workow for interpreting the exome tumor data of an individual patient using Melanoma Proler to identify clinically
important genetic modications is shown at the right. 13 of the 50 genes modied by non-synonymous mutations in the
sample tumor are members of the melanoma molecular pathway found in the KEGG database. Comparison of the mutations
in the sample tumor with the mutation proles of other tumors in MelanomaDB reveals signicant features of the sample tumor,
most notably the occurrence of BRAF and NRAS mutations which, while collectively occurring in 90% of melanomas, tend to be
mutually exclusive and rarely co-occur in the same tumor. Other frequently occurring driver mutations found in the sample tumor
include TP53 and CDKN2A.
When signicant numbers of genes within a single pathway carry mutations, there may be synergies or complementarity
between mutations, which may in turn inform choice of pathwaytargeted drug therapies. In this case, despite the presence
of a BRAF somatic variant (which may be veried as a V600E mutation by interrogating the VCF le, and in some cases by
additional Sanger sequencing), resistance to the BRAF V600E targeting drug Vemurafenib may potentially be more likely due
to the presence of somatic variants in genes encoding other proteins in this signaling pathway. After they have been identied
using Melanoma Proler, the capacity for additional mutations in this pathway to contribute to Vemurafenib resistance, such
as by constitutively activating a signaling molecule downstream of BRAF, can be carefully examined. For example, this tumor
contains a mutation in the AKT gene, an integrating protein also playing a role in over 30 other molecular pathways. Activating
AKT mutations have previously been identied, and high RNA expression of AKT is associated with poor patient prognosis [4].
AKT is predicted to be druggable [3], and there are entries in the DrugBank database for experimental compounds targeting
AKT [2]. This highlights a new type of patient in which an experimental AKT inhibitor may usefully be explored.
Conclusions
Melanoma Proler provides a useful set of integrated tools to facilitate interpretation of exome sequence data for N=1 individual
tumors. Being able to categorize and visualize the patterns of somatic mutations in a particular tumor in reference to 300
other melanomas provides a context to understand the complex mutational information in an individual tumor, including the co-
occurrence of mutations in the same molecular pathway and information relating to drugs, druggability, and expression-survival
relationships.
To further extend its utility, future development of Melanoma Proler will aim to include predictions of the effects of specic
somatic mutations in the tumor (e.g. loss of function, altered function, or activation of the encoded protein), and to add
information about other types of genetic modications in tumors such as structural variations, methylation, non-coding RNA
and expression data. Expanded datasets of tumors will be added to MelanomaDB as they become available, including the
incorporation of whole genome data.
Melanoma Proler Web Tool for Integrative Analysis of Melanoma
Melissa Pentecost, Steven Stones-Havas, Christopher Duran, Brett Ammundsen
Biomatters Limited, Auckland, New Zealand
Alexander Trevarton, Cristin Print
Department of Molecular Medicine and Pathology, School of Medical Sciences, University of Auckland, Auckland, New Zealand
Contact: melissa@biomatters.com
1. Trevarton et al., Frontiers in Cancer Genetics (2013)
2. Knox et al., Nucleic Acids Research (2011)
3. Blake et al. The Cancer Journal (2011)
4. Bogunovic et al., Proceedings of the National Academy of Sciences USA (2009)
5. Sales et al., BMC Bioinformatics (2012)