MelanomaDB Help Page
Despite on-going research, metastatic melanoma five year survival rates remain low and treatment options
limited. Researchers can now access a rapidly growing amount of molecular and clinical information about
melanoma. This information is becoming increasingly difficult to assemble and interpret due to its
dispersed nature, yet as it grows it becomes increasingly valuable for understanding melanoma biology
and clinical progression. Integration of this information into a comprehensive resource to aid rational
experimental design and patient stratification is needed. We have assembled a web-accessible melanoma-
focused database that incorporates clinical and molecular data from several sources, which will be
regularly updated as new information becomes available. The way this database is constructed, information
about melanoma is incorporated as gene sets (e.g. one gene set could be those RNAs for which expression
level in metastatic melanomas is significantly associated with patient survival). This melanoma-
associated gene set database allows complex links to be drawn between: the genetic changes in individual
melanomas revealed by DNA sequencing, associations between gene expression and patient survival, data
concerning drug targets, biomarkers, druggability and clinical trials, as well as statistical analysis of
relationships between molecular pathways and clinical parameters using these data sets. Several drug and
biomarker gene sets are included in particular to allow the identification of novel druggable molecules
associated with melanoma development or progression. The database is freely available. Detailed
instructions for this database's' use can be found in the Help link at the top of this page. Please go to
either of the links below to start using the database. This database was last updated on 20 May 2013.
Current version 2.0.
http://genesetdb.auckland.ac.nz/melanomadb
Go to the link below if you simply want to see whether your own gene list has statistically significant
enrichment for any melanoma-associated gene sets.
http://genesetdb.auckland.ac.nz/melanomadb/haeremai.php
If you need help please don't hesitate to contact us (bioinformatics@auckland.ac.nz).
General information about the gene sets.
-
Trans Membrane (4530 genes)
This set consists of genes identified in Affymetrix’s annotation files as encoding transmembrane proteins. This is based on data from three databases:
- Ensembl (e.g: www.ensembl.org/id/ENSP00000378496) entries that contain transmembrane elements in their protein summary.
- Pubmed protein (e.g. http://www.ncbi.nlm.nih.gov/protein/NP_001171536) entries containing at least one ‘Site‘ field containing "/site_type="transmembrane region""
- GenScan (e.g. GENSCAN00000008560), which uses a Markov model to predict protein features, according to the original paper
(http://ai.stanford.edu/~serafim/cs262/Papers/GENSCAN.pdf).
-
DrugBank Drug ID (2068 genes) up to date 20 May 2013
http://www.drugbank.ca/
This gene set consists of genes encoding proteins that are drug targets listed in the DrugBank database.
-
Known Melanoma Drugs (18 genes)
http://www.cancer.gov/cancertopics/druginfo/melanoma
This set consists of genes that encode targets of clinically available drugs for melanoma.
-
Therapeutics Targets Database melanoma drug targets (72 genes)
http://bidd.nus.edu.sg/group/cjttd/
This gene set consists of genes encoding a protein (or mRNA) that is a drug target associated with melanoma, including both FDA approved drugs and drugs that have been trialed for the treatment of melanoma, according to the Therapeutic Targets Database.
-
ClinicalTrials.gov melanoma clinical trial (314 genes)
http://clinicaltrials.gov/
This set consists of genes encoding proteins that are targets of drugs that have been trialed for the treatment of melanoma in a trial listed at clinicaltrials.gov.
-
Melanoma biomarkers (85 genes)
http://www.ncbi.nlm.nih.gov/pubmed/17221215
RNA or encoded protein is a recognized or putative biomarker according to:
- KEGG BRITE
- Utikal et al (2007) (Serologic and Immunohistochemical markers)
- A review by Mehta et al (2010).
-
REMARK study melanoma biomarkers (40 genes)
http://www.ncbi.nlm.nih.gov/pubmed/21659462
This gene set contains genes encoding the putative biomarkers significantly associated with melanoma specific mortality or disease-free survival according to:
- The REMARK meta-studies by Gould Rothberg et al (2009)
- Schramm, Mann (2011).
-
Melanomagenesis Drivers (19 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22475929
This set consists of the oncogenes and tumour suppressor genes that are thought to be drivers of melanomagenesis according to Flaherty, Hodi, Fisher (2012).
-
Druggability: Li and Lai (1332 genes)
http://www.ncbi.nlm.nih.gov/pubmed/17883836
This gene set consists of genes predicted to be ‘druggable’ (but not yet drugged)
by Li and Lai (2007)
-
Druggability: Sophic ENSEMBL list (2031 genes)
http://www.sophicalliance.com/services%20drug%20genome.php
These genes are those on Sophic’s ‘ENSEMBL list’, genes that were included in the druggable genome described by Hopkins and Groom (2002).
-
Druggability: Sophic DrugBank list (1368 genes)
http://www.sophicalliance.com/services%20drug%20genome.php
These genes are on Sophic’s DrugBank list, which lists genes identified as druggable by DrugBank.
-
Druggability: Sophic InterPro-BLAST list (2269 genes)
http://www.sophicalliance.com/services%20drug%20genome.php
These genes are on Sophic’s InterPro-BLAST list. This list of druggable genes was obtained by the following method:
- Extract all protein sequences from Drugbank corresponding to the InterPro families.
- Extract the Swissprot protein sequences corresponding to each gene name listed in the HUGO database.
- Carry out a BLAST search for each sequence in the Drugbank against all the downloaded Swissprot protein sequences. This method provides an expanded list of proteins that are then mapped back to their respective HUGO gene symbols.
-
Druggability: Sophic BioLT-Drugbank list (3335 genes)
http://www.sophicalliance.com/services%20drug%20genome.php
These genes are on Sophic’s BioLT-DrugBacnk list. The BioLT-Drugbank list is the list of genes identified as druggable using the BioLT7 Text Mining Tool. BioLT is a literature mining tool that uses both a lexical and natural language processing approach to efficiently mine information from literature sources. This tool was used to mine all gene names in the literature (Pubmed) that co-occurred with the name of a drug (as identified using Drugbank) and the term “inhibit*”. A thousand entries out of the 3,610 BioLT genes were manually checked and 77% were found to be viable “suspect” druggable genes.
-
Druggability: Sophic‘s Qiagen list (6259 genes)
http://www.sophicalliance.com/services%20drug%20genome.php
These genes are on Sophic’s Qiagen list, which are genes that have been publically identified as druggable by Qiagen Inc.
-
Druggability: Primary druggable genome v2 (DG2) (6047 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22147262
These genes are those included by the study of Tiedemann et al (2011) in their primary druggable genome v2 (DG2) for their Achilles heel siRNA screen.
-
DrugEBIlity: Positive tractable scores (2885 genes)
https://www.ebi.ac.uk/chembl/drugebility/structure
This gene set consists of the genes encoding proteins which had a positive average tractable score for at least one domain. EBI’s tractable score evaluates the suitability of the binding site for small molecules under the Lipinski‘s Rule of Five.
-
DrugEBIlity: Positive druggable scores (1548 genes)
https://www.ebi.ac.uk/chembl/drugebility/structure
This gene set consists of the genes encoding proteins which had a positive average druggable score for at least one domain. EBI’s druggable score also evaluates the suitability of the binding site for small molecules under Lipinski‘s Rule of Five, but with more stringent conditions than the tractability score.
-
DrugEBIlity: Positive ensemble scores (730 genes)
https://www.ebi.ac.uk/chembl/drugebility/structure
This set consists of the genes encoding proteins which had a positive average ensemble score for at least one domain. EBI’s ensemble score is the average of druggability score calculated under a range of different models.
-
Secreted: Uniprot (1479 genes)
http://www.uniprot.org/
The set consists of the genes encoding proteins with UniProt entries that have ‘Secreted’ under their ‘location’ field.
-
Secreted: Predicted by Secreted Protein Database (283 genes)
http://spd.cbi.pku.edu.cn/
These genes encode proteins that are predicted to be secreted by the Secreted Protein Database.
-
Protein Kinase (516 genes)
http://kinase.com/kinbase/
These genes encode proteins that are listed on the protein kinase database, KinBase.
-
IRIDESCENT: Gene link to ‘melanoma‘ (371 genes)
http://www.ncbi.nlm.nih.gov/pubmed/14734310
Wren et al (2004) developed IRIDESCENT, a computational method to identify relationships within published literature by constructing a network of tentative relationships between objects. IRIDESCENT identifies in published literature 371 genes having a relationship (direct or implicit) to the object ‘melanoma‘.
-
IRIDESCENT: Gene link to ‘metastatic melanoma‘ (1267 genes)
http://www.ncbi.nlm.nih.gov/pubmed/14734310
Wren et al (2004) developed IRIDESCENT, a computational method to identify relationships within published literature by constructing a network of tentative relationships between objects. IRIDESCENT identifies in published literature 1267 genes having a relationship (direct or implicit) to the object ‘metastatic melanoma‘. Although ‘metastatic melanoma’ is a subset of ‘melanoma’, this gene set is larger because criteria were less stringent.
-
GAMMA: Gene link to ‘melanoma‘ (2528 genes)
http://www.ncbi.nlm.nih.gov/pubmed/19447786
Wren (2009) developed GAMMA which conducts a meta-analysis of gene expression behaviour across 16,000 microarray experiments to infer function. GAMMA identifys co-expression connections between genes that are consistent across all different experiment types. If a gene has co-expression connections to genes with published associations with melanoma, then it has an inferred connection to melanoma. The strength of published associations with ‘melanoma‘ is calculated by IRIDESCENT. This set consists of the 2428 genes with inferred connection to melanoma.
-
Jonsson RNA Expression Survival Associations (284 genes)
http://www.ncbi.nlm.nih.gov/pubmed/20460471
This gene set consists of the 284 genes that our own analysis identified as having expression associated with survival in metastatic melanoma, based on the data of Jonsson et al (2010). Jonsson‘s microarray data was obtained from the Gene Expression Omnibus (Series GSE22155), loaded into R and analysed using R‘s ‘survival‘ package. A Cox proportional hazards model (Cox PH) was fitted to each probe using R‘s coxph function. After examining the distribution of p-values, a probe was considered associated with survival if the Cox PH p-value was less than 0.0005. Additionally, the tumours were divided into two groups for log rank tests, once at each decile (10-quantile). Subsequently, 9 log rank tests were performed for each probe using R‘s survdiff function. A probe was considered associated with survival if any log-rank p-value was less than 0.000001. Furthermore, a probe was also considered to be associated with survival if it had both a log-rank p-value lower than 0.0001 and a Cox Ph p-value lower than 0.005.
-
John RNA Expression Survival Associations (18 genes)
http://www.ncbi.nlm.nih.gov/pubmed/18698035
This set consists of the genes that John et al (2008) identified as having highly significant differential expression between 16 "poor prognosis" and 13 "good prognosis" patients from stage III melanoma samples.
-
Mandruzzato RNA Expression Survival Associations (66 genes)
http://www.ncbi.nlm.nih.gov/pubmed/17129373
This set consists of the genes that Mandruzzato et al (2006) identified as associated with survival in 38 patients with stage III and stage IV melanoma. Forty-three tumour tissues from 38 patients with stage III and stage IV melanoma were profiled with a 17,500 element cDNA microarray. Expression data were analysed using significance analysis of microarrays (SAM) to identify genes associated with patient survival, and supervised principal components (SPC) to determine survival prediction.
-
Journe RNA Expression Survival Associations (222 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22045183
This set consists of the genes that Journe (2011) associated with survival in 32 metastatic melanoma patients. 32 stage III skin and lymph node metastases were collected and used to generate microarray profiles. The tumours were divided into two groups, overall survival below 30 months (n = 10) and overall survival over 30 months (n = 22), and probes with a fold change greater than 2.5 and with a t-test p-value less than 0.05 were considered to be associated with survival.
-
Timar RNA Expression Metastasis Associations (184 genes)
http://www.ncbi.nlm.nih.gov/pubmed/20177751
These genes are those that the review of Timar et al (2010) identified as being associated with metastasis by two or more studies of melanoma.
-
Timar RNA Expression Metastasis Associations in 3 or 4 separate studies (17 genes)
http://www.ncbi.nlm.nih.gov/pubmed/20177751
These genes are those that the review of Timar et al (2010) identified as being associated with metastasis by three or more studies of melanoma.
-
Bogunovic RNA Expression Survival Associations (317 genes)
http://www.ncbi.nlm.nih.gov/pubmed/19915147
This set consists of the genes that our own analysis identified with expression associated with survival in metastatic melanoma. Bogunovic‘s microarray data was obtained from the Gene Expression Omnibus (Series GSE19234), loaded into R, normalised using RMA and analysed using R‘s ‘survival‘ package. A Cox proportional hazards model (Cox PH) was fitted to each probe (row) using R‘s coxph function. After examining the distribution of p-values, a row was considered associated with survival if the Cox PH p-value was less than 0.0001. Additionally, the tumours were divided into two groups for log rank tests, once at each decile (10-quantile). Subsequently, 9 log rank tests were performed for each probe using R‘s survdiff function. A probe was considered associated with survival if any log-rank p-value was less than 0.000001. Furthermore, a probe was also considered to be associated with survival if it had both a log-rank p-value lower than 0.0001 and a Cox PH p-value lower than 0.001.
-
Berger (2010) Gene Fusions (22 genes)
http://www.ncbi.nlm.nih.gov/pubmed/20179022
Berger et al (2010), in 10 patient-derived short term cultures and cell lines, found 11 novel melanoma gene fusions.
-
Berger (2010) read-through transcripts (24 genes)
http://www.ncbi.nlm.nih.gov/pubmed/20179022
Berger et al (2010), in 10 patient-derived short term cultures and cell lines, found 12 novel melanoma read-through transcripts.
-
Berger (2010) non-synonymous mutations (621 genes)
http://www.ncbi.nlm.nih.gov/pubmed/20179022
Berger et al (2010), in 10 patient-derived short term cultures and cell lines, found 721 novel, non-synonymous coding variants.
-
Berger (2012) 11 frequently mutated genes (11 genes)
http://www.ncbi.nlm.nih.gov/pubmed/20179022
Berger et al (2012), in 25 metastatic melanoma tumours, identified 11 genes significantly mutated.
-
Berger (2012) non-synonymous mutations (5712 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22622578
Berger et al (2012), in 25 metastatic melanoma tumours, identified 9276 non-synonymous somatic base-pair mutations in protein encoding regions.
-
Berger (2012) small InDels (39 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22622578
Berger et al (2012), in 25 metastatic melanoma tumours, identified 40 small InDels in protein coding regions.
-
Berger (2012) structural rearrangements (1942 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22622578
Berger et al (2012) in 25 metastatic melanoma tumours, identified 2428 somatic structural rearrangements.
-
Cancer Cell Line Encyclopedia melanoma mutations (1538 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22460905
Cancer Cell Line Encyclopedia (Barretina et al 2012) in 59 malignant melanoma cell lines identifies 7374 mutations in 1539 genes.
-
Catalogue Of Somatic Mutations In Cancer (COSMIC) missense mutations (618 genes)
http://www.sanger.ac.uk/genetics/CGP/cosmic/
From the Sanger Institute‘s Catalogue Of Somatic Mutations In Cancer (COSMIC), 5599 missense substitution mutations in 618 genes were retrieved from malignant melanoma, benign melanocytic nevus and melanocytoma samples.
-
Nikolaev copy number alterations (45 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22197931
Nikolaev et al (2012) in 7 germline-matched melanoma cell lines identified altered Copy number of 45 genes from aCGH and SNP arrays and transcript expression.
-
Nikolaev InDels Somatic Mutations (25 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22197931
Nikolaev et al (2012) in 7 germline-matched melanoma cell lines, identified 25 genes with InDels
-
Nikolaev non-synonymous mutations (1710 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22197931
Nikolaev et al (2012) in 7 germline-matched metastatic melanoma cell lines, identified 1710 genes with 2730 non-synonymous mutations.
-
Palavelli Matrix Metalloproteinase mutations (8 genes)
http://www.ncbi.nlm.nih.gov/pubmed/19330028
Palavalli et al (2009) identified 28 somatic mutations in 8 Matrix Metalloproteinase (MMP) genes in metastatic melanoma resections.
-
Prickett tyrosine kinase mutations (19 genes)
http://www.ncbi.nlm.nih.gov/pubmed/19718025
Prickett et al (2009) identified 99 somatic mutations in 19 tyrosine kinase genes in metastatic melanoma resections.
-
Prickett G protein-coupled receptor mutations (94 genes)
http://www.ncbi.nlm.nih.gov/pubmed/21946352
Prickett et al (2011) identified 106 somatic mutations in 94 G protein-coupled receptor genes in 11 metastatic melanoma resections.
-
Prickett 11 G protein-coupled receptor genes with multiple mutations (11 genes)
http://www.ncbi.nlm.nih.gov/pubmed/21946352
Prickett et al (2011) identified 106 somatic mutations in 94 G protein-coupled receptor genes in 11 metastatic melanoma resections. 11 of these 94 genes harbored at least two somatic mutations.
-
Matched Pair Cancer Cell Lines melanoma mutations (966 genes)
http://www.sanger.ac.uk/genetics/CGP/Studies/Matched/
The Sanger Institute‘s "Matched Pair Cancer Cell Lines" contains 6 malignant melanomas. The genes in this gene set are those that have mutations in any of these 6 cell lines.
-
Stark non-synonymous mutations (1724 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22197930
Stark et al (2011) performed exome sequencing of 8 melanoma cell lines and matched normal lymphoblastoid cell lines. In 1724 genes, they found 2139 mutations predicted to alter protein structure.
-
Turajlic non-synonymous mutations (41 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22183965
Turajlic et al (2012) confirmed 44 non-synonymous mutations in a primary acral melanoma and its lymph node metastasis.
-
Turajlic structural variations (614 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22183965
Turajlic et al (2012) validated 73 structural variations in a primary acral melanoma and its lymph node metastasis. The genes in this set are those involved in any of those structural rearrangements.
-
Wei 10 mutations with repeated occurrence (10 genes)
http://www.ncbi.nlm.nih.gov/pubmed/21499247
Wei et al (2011) in 14 normal-matched metastatic tumour whole-exome sequences identified 10 mutations that occurred in more than one of their samples.
-
Wei 16 highly mutated genes (16 genes)
http://www.ncbi.nlm.nih.gov/pubmed/21499247
Wei et al (2011) in 14 normal-matched metastatic tumour whole-exome sequences identified 16 highly mutated genes.
-
Wei 68 genes with elevated frequency of mutation (68 genes)
http://www.ncbi.nlm.nih.gov/pubmed/21499247
Wei et al (2011) in 14 normal-matched metastatic tumour whole-exome sequences detected 68 genes that appeared to be somatically mutated at elevated frequency. This includes the 16 genes from the highly mutated genes set.
-
Wei non-synonymous mutations (2239 genes)
http://www.ncbi.nlm.nih.gov/pubmed/21499247
Wei et al (2011) in 14 normal-matched metastatic tumour whole-exome sequences detected 2813 nonsynonymous amino acid changes (2589 missense and 175 nonsense alterations, 49 mutations at splice sites).
-
Hodis genes mutated in more than 10% of their 121 tumour samples (516 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22817889
Hodis et al (2012) performed exome sequencing on 121 paired tumour and normal genomic DNAs. They identified 516 genes containing non-synonymous mutations in more than 10% of tumours.
-
Hodis genes mutated in more than 20% of their 121 tumour samples (79 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22817889
Hodis et al (2012) performed exome sequencing on 121 paired tumour and normal genomic DNAs. They identified 79 genes containing non-synonymous mutations in more than 20% of tumours.
-
Hodis recurrent somatic coding mutations (24 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22817889
Hodis et al (2012) performed exome sequencing on 121 paired tumour and normal genomic DNAs. They identified 24 recurrent somatic coding mutations (identical mutations occurring in multiple tumours).
-
Hodis small InDels (378 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22817889
Hodis et al (2012) performed exome sequencing on 121 paired tumour and normal genomic DNAs. They identified 378 small InDels.
-
Krauthammer genes mutated in more than 5% of their 147 tumour samples (381 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22842228
Krauthammer et al (2012), in 147 melanoma tumours (both primary and metastatic tumours) found 23888 missense and 1596 nonsense mutations in 8984 genes. This set consists of the genes that were mutated in more than 5% of their samples.
-
Krauthammer genes mutated in more than 10% of their 147 tumour samples (49 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22842228
Krauthammer et al (2012), in 147 melanoma tumours (both primary and metastatic tumours) found 23888 missense and 1596 nonsense mutations in 8984 genes. This set consists of the genes that were mutated in more than 10% of their samples.
-
Krauthammer InDels (274 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22842228
Krauthammer et al (2012), in 147 melanoma tumours (both primary and metastatic tumours) found 282 InDels in 274 genes.
-
Krauthammer Splice-site mutations (368 genes)
http://www.ncbi.nlm.nih.gov/pubmed/22842228
Krauthammer et al (2012), in 147 melanoma tumours (both primary and metastatic tumours) found 399 splice site mutations in 368 genes.
-
Mutated in >5% of total metastatic tumours (1558 genes)
http://genesetdb.auckland.ac.nz/melanomadb/chooseSets.php
Across all of the given somatic mutation sequencing studies, there were 96 separate metastatic melanoma tumour samples. This set consists of the genes that were mutated in 5% or more of all metastatic tumour samples.
-
Mutated in >10% of total metastatic tumours (245 genes)
http://genesetdb.auckland.ac.nz/melanomadb/chooseSets.php
Across all of the given somatic mutation sequencing studies, there were 96 separate metastatic melanoma tumour samples. This set consists of the genes that were mutated in 10% or more of all metastatic tumour samples.
-
Mutated in >15% of total metastatic tumours (69 genes)
http://genesetdb.auckland.ac.nz/melanomadb/chooseSets.php
Across all of the given somatic mutation sequencing studies, there were 96 separate metastatic melanoma tumour samples. This set consists of the genes that were mutated in 15% or more of all metastatic tumour samples.
-
Bogunovic Cox PH Survival Associations (307 genes)
http://www.ncbi.nlm.nih.gov/pubmed/19915147
This set consists of the genes that our own analysis identified with expression associated with survival in metastatic melanoma, similar to the Bogunovic RNA Expression Survival Associations gene set, except this set is entirely determined by a Cox proportional hazards model with a higher p-value cut-off. Bogunovic‘s microarray data was obtained from the Gene Expression Omnibus (Series GSE19234), loaded into R, normalised using RMA and analysed using R‘s ‘survival‘ package. A Cox proportional hazards model (Cox PH) was fitted to each probe (row) using R‘s coxph function, and Cox PH function of the SAMR package. After examining the distribution of p-values, a row was considered associated with survival if the Cox PH p-value was less than 0.0005.
-
Bogunovic Lowest 20% RNA Expression Survival Associations (167 genes)
http://www.ncbi.nlm.nih.gov/pubmed/19915147
This set consists of the genes that our own analysis identified as having low expression associated with survival in metastatic melanoma. Bogunovic‘s microarray data was obtained from the Gene Expression Omnibus (Series GSE19234), loaded into R, normalised using RMA and analysed using R‘s ‘survival‘ package. The tumours in this data-set were divided at each probe set by level of RNA expression. For each probe set, the 20% of tumours with the lowest expression were considered as a ‘low’ expression group and the remaining tumours were considered the ‘not low’ expression group. Survival curves were constructed for both groups and compared using the log rank test. A probe set was considered associated with survival if the log rank p-value was less than 0.0005.
-
Bogunovic Highest 20% RNA Expression Survival Associations (297 genes)
http://www.ncbi.nlm.nih.gov/pubmed/19915147
This set consists of the genes that our own analysis identified as having high expression associated with survival in metastatic melanoma. Bogunovic‘s microarray data was obtained from the Gene Expression Omnibus (Series GSE19234), loaded into R, normalised using RMA and analysed using R‘s ‘survival‘ package. The tumours in this data-set were divided at each probe set by level of RNA expression. For each probe set, the 20% of tumours with the highest expression were considered as a ‘high’ expression group and the remaining tumours were considered the ‘not high’ expression group. Survival curves were constructed for both groups and compared using the log rank test. A probe set was considered associated with survival if the log rank p-value was less than 0.0005.