Biologists, scientists and clinicians might tend to ignore all the hype and discussion around “big data’ assuming that it does not concern their profession and is of no importance in their work.  In fact, the conversations about ‘predictive analysis’ in the context of consumer behavior on the web, mobile device, and retail might reinforce that thinking.  However, while this consumer transactional data is commonly unstructured in millions, perhaps billions of records in the 100’s of bytes each, it is no where close to the ~350 gigabytes in raw form for the full genome of a single person. It is no surprise then, that at a recent Intel webinar conference, focused on discussing the challenges, solutions, and best practices of big data, NextBio’s expertise in using big data technologies for genomic data was showcased alongside PayPal and Forester Research’s ability to work with consumer data.

Genomic data certainly fits the key measures of big data- ‘volume’, ‘velocity’, and ‘variety’. As genomic scientist know, raw genomic data is represented in several formats and requires alignment, normalization, reduction to a reference, and compression for consumption and interpretation. The challenge then is to build and provide an infrastructure that can not only consume, process, and report on this genomic data but also retain it for years.

A full genome consists of 3.2 billion base pairs of information, with a 99.9% commonality to a reference. A single person can be represented in 10 million rows of information. That extrapolates out to 1 trillion rows of information for 100,000 people!  A few years ago, it was not economically feasible to store and process data of any sort at this level.  Now, Open Source big data solutions like Hadoop and Hbase provide truly scalable and reliable storage and distributed computing capabilities that solve important infrastructure problems.  This combined with powerful commodity computers has revolutionized genomic analysis.

In both research and clinical settings ‘real time’ capabilities to present correlations and filter massive amounts of clinical and genomic data are requirements.  The correlations enable the interpretation of the genome, biomarker discovery, and generation and testing of research hypothesis. The filtering of the data enables selection for clinical trials, and the ability to compare subgroups of a population of patients through meta-analysis. All of these abilities are immensely important to clinicians and researchers and are made possible in NextBio Clinical only through the use of big data technologies.

NextBio’s management of big data rests on three main pillars – Hadoop , Hbase and sharded databases. Hadoop itself was designed as a ‘batch’ processing system – provide a large amount of data and some time later results are available. Hadoop is the storage system and a compute engine for calculating correlations between genomic data and drug studies.  Hbase is the access method by which billions of gene data rows are stored and selectively retrieved. Sharded databases contribute to part of the solution mix as the repository for correlations and summary results to enable real time filtering.

Each technology component excels at a particular function and scales at its own rate. Careful thought is required to select the right technology for the data use case and response time, and there are other elements that require design.  Among them, Network infrastructure, fail-over of all components at all tiers, testing environments, and even monitoring systems for service quality and capacity planning.  All must be designed for scale.  Finally, given the nature of genomic data, access control and auditability “features” for data governance require investment as well.

Now that the 3 V’s of big data platforms, sophisticated analysis algorithms, and affordable genomic sequencing are at an intersection, we can expect to see huge leaps in medicine. A biologist, scientist, and clinician can now see the benefits of analysis and correlation of genomic data using big data technologies as the key to the future of scientific discovery and advancement in clinical care.

A newly diagnosed cancer patient and their family might have barely moved from the first stage to the next of the five stages of grief but, regardless of the course of their emotional journey, their cancer vocabulary begins to grow from day one. Biopsy, staging, chemo, neutropenia,… words that had meant nothing to them before, take on a very tangible meaning invoking thoughts of hospital beds and days spent being sick. As the battle with cancer continuous, their vocabulary continues to evolve as well. Remission, transplant, … relapse, metastasis, morphine, …hospice; relief alternating with despair, a roller coaster ride with too many ups and downs.

In the last few years, a newer set of words  are beginning to make their way into the cancer lexicon- whole genome sequencing (WGS), targeted therapies, biomarkers… words that are becoming associated with some recent successes and cautious optimism as we relentlessly search for a cure to cancer. These partial successes, the understanding that cancer is a genetic disease, and the decreasing cost of whole genome sequencing raise important questions about making tumor sequencing an integral part of cancer treatment.

Central to this discussion are several different scientific and social issues. On the scientific side, intratumor heterogeneity, challenges with data interpretation and management, and physician training in genomics dominate the conversation. In the social area, cost and insurance coverage, and ethical issues remain center stage.

Intratumor heterogeniety refers to the existence of genetically divergent tumor cell clones within a tumor as well as evolutionary divergence between primary tumors and metastatic outgrowths. If greater that 60% of all somatic mutations are not detectable across every tumor region as a New England Journal of Medicine study showed, then even finding an adequate sample on biopsy is a challenge. Issues of quality and quantity are also important as tumor heterogeneity analysis relies on genome amplification methods that produce a representation not a replica of the tumor genome. With respect to distant metastasis, which are responsible for the majority of cancer related deaths, radically different patterns of allelic loss and genetic divergence from the primary tumor have been documented in the metastasis. This further complicates the process of obtaining a biopsy and identifying therapeutic targets in these patients.

Management of WGS data for clinical applications implies the ability to detect actionable variants from the WGS data, the ability to store genomic data in the electronic medical record and the ability to integrate the use of this genomic data into patient care. While companies such as NextBio use Big Data technologies to scale their platform for storing and interpreting genomic data, and have passed HIPAA audits to enable secure integration into the clinical workflow, the discovery and understanding of actionable variants of clinical significance is still ongoing within the scientific community.

My fellow physician colleagues and I are familiar with targeted genomic tests in oncology, but the use of WGS continues to be a measure of last resort in the majority of cases. Presenting the option of whole genome sequencing to our patients as a part of routine oncology care and providing them with information of sufficient depth to allow them to make an informed decision will require a paradigm shift.

The social issues such as reimbursement of WGS remain unresolved. Dr. Jeffery Roche, Coverage and Analysis officer at CMS announced at a Capitol Hill briefing last October that WGS is unlikely to be covered by CMS anytime soon. Ethical issues around WGS for clinical use are being actively addressed and the Presidential Commissions final report to be presented this fall might provide guidance on at least a few.

The ENCODE project, designed to interpret the human genome, recently made headlines as it infused the findings of the Human Genome Project with meaning. The  discovery that a full 80% of the human genome is associated with biochemical function caught our attention.  In the same news announcements, this quote by renowned researcher Michael Snyder “In my mind, it is a no-brainer to get your genome sequenced if you have cancer” continued to emphasize the role of genomics in the treatment of cancer.

Perhaps, the time has come to make the inclusion of WGS in cancer care a clinical, scientific, and social priority by, investigating options for adaptive therapy, fast tracking the discovery of actionable clinical variants, reimbursing WGS where it is appropriate, and focusing on resolving the ethical issues. Clinicians can play an especially important role by helping their patients make sense of the confusing landscape, educating them, discussing pros and cons and helping them make good informed decisions. This may be our chance to finally introduce a new word into the cancer vocabulary , “Cure”.

Andy Warhol talked about the fifteen minutes of fame of the future in 1968 and this quote captured everyone’s attention coming back again and again in a variety of forms and fashions. Those of us that work in health care and genomics can be just as  captivated by the value of anonymity and privacy.

De-identification of data and safeguarding of Protected Health Information (PHI) is important to maintaining trust and security in the health care setting. HIPAA, or the Health Information Portability and Accountability Act, enacted by the United States Congress and signed by President Clinton in 1996 defines the policies, procedures and guidelines for maintaining this privacy and security. For most research purposes, de-identification of data is accomplished by stripping the data of 18 specific identifiers such as names, phone numbers, photographs, Social Security numbers, health insurance information, etc.

That this comprehensive process of de-identification is not sufficient for genomic data became painfully clear when Nils Homer and his colleagues published their landmark study in PLoS Genetics showing that it is possible for re-identification of an individuals data in a large set of pooled genetic data. This publication led to the understanding that genomic data itself is an identifier, caused a significant shift in the approach to publicly available data and, questioned the anonymity of de-identified data. This also prompted the NIH as well as other institutions to change their data sharing policies.

The answer to the conundrum of using genomic data appears to be in enhanced security measures validated by HIPAA as well as the process of informed consent. The White House Bioethics Commissions is currently considering this issue as it moves forward in drafting its genomic privacy policy aimed at addressing the individual privacy and personal choice concerns tied to the use of personal genomic data in research and medicine. The focus of this draft is on the use of whole genome sequencing technologies in research and clinical practice. A special focus of the Commission is the consent process as it relates to whole genome sequencing.

As a physician, the informed consent process is a cornerstone of my clinical research and practice, whether the discussion is around end of life care issues with a cancer patient or around recruiting a patient for a clinical study. Consent for obtaining or using an individual’s genomic data is no different. It needs to be broad enough to further research and discovery while also allowing the individual to gather a good understanding of who would be able to access their data and for what purposes. Patient preferences on use of their genomic data should be elicited and complied with.

Having an informed consent process is however not enough. Meeting HIPAA requirements will add an additional layer of security as de-identification is no longer sufficient to ensure anonymity. NextBio’s announcement of successfully passing an audit to meet HIPAA requirement is based on this new understanding that all systems that work with genomic data need to meet these higher security standards. Meeting these standards would infuse the era of genomic medicine with a higher level of trust and enable accelerated scientific and medical discovery.

Most of us probably associate being sick with the entire body- a fever, aches, chills and other broad symptoms. When it comes to a disease like cancer, we might take an organizational step or two down to think of a specific organ or tissue: breast, lung or brain cancer.

But increasingly, patient’s stories point clearly toward a finer resolution of cancer diagnosis, down to the level of a single gene. A report in the New York Times last week describes how a team of researchers worked to identify the genetic aberration underlying a colleague’s cancer, and helped treat his leukemia with an off-label drug currently used to treat kidney cancers.

Dr. Wartman’s success story highlights an increasingly dominant view of cancer- not as a single disease, nor as a disease of a particular tissue or organ, but a constellation of mutations unique to a cancer cell. The genetic changes that cause one patient’s leukemia may be completely different from those underlying another patient’s cancer. And some of these mutations may bear greater resemblance to renal or lung tumors than to any kinds of leukemia.

Knowing precisely which mutations drive a patient’s tumor, physicians can potentially choose a therapy targeting a precise genetic aberration rather than resorting to a generic chemotherapeutic regimen for leukemia. But questions of how to identify these mutations and who pays for them abound.

As a cancer researcher himself, Dr. Wartman was particularly fortunate with access to the latest research. His colleagues at Washington University included him in a research study, with the instruments, resources and manpower to sequence his entire genome and analyze changes in gene expression to find one potential genetic driver, an over-expression of the FLT3 gene. Increased expression of FLT3 is linked to cell growth and proliferation in many forms of leukemia and other cancers, and one drug on the market, Sunitinib (Pfizer), targets FLT3 over-expression in advanced kidney cancers.

When his insurance company refused to cover his off-label use of the drug and Pfizer turned him down from their compassionate care program, he used his own funds to cover the cost of a week’s medication (at the rate of $ 330 each day). His colleagues chipped in to cover another month of treatment.

After the treatments, multiple tests show his bone marrow free of cancer cells. Time will tell whether the month-long treatment was fully effective. For now, Dr. Wartman and his colleagues are planning a clinical trial to test the success of Sunitinib in treating FLT3-overexpressing leukemia in other patients.

His story highlights both the price and the value of such genetic gambles in cancer treatment. Pending clear direction from insurance companies and healthcare providers, accessing the technology and researchers who can identify genetic aberrations in cancers remains restricted to a small fraction of patients with the means to reach these scientists.

The effectiveness of molecular-targeted medicines for long-term remission is still being understood. In some cases, patients may respond remarkably well to such drugs initially, only to have their cancer rebound within a few months. Two studies in Nature this week suggest that one reason for this recurrence may be in the tissues surrounding tumors rather than the tumors themselves. Secondary mutations that evolve in tumors over time could also contribute to cancer recurrence and drug resistance evolved over treatment. Most of all, only a handful of the mutations known to drive different cancers can be effectively targeted by molecular therapies.  As of 2011, less than 100 drugs had approved genetic indications.

Several translational research initiatives are attempting multi-pronged approaches to address these concerns. Collaborations between academic researchers and pharmaceutical developers hope to integrate tumor biology research with drug development processes to speed up the process of identifying the best molecular therapeutics for specific mutations. While drug manufacturers frequently form academic collaborations with laboratories studying the basic genetic pathways underlying cancer progression, a recent NIH initiative also offers academics the opportunity to find new uses for ‘abandoned’ pipeline molecules.

Collectively, these researchers seem clearly set to target cancer not as a disease of a tissue or organ system, but a disorder of the genes. Efforts by Dr. Wartman’s colleagues and other scientists hope to establish a roadmap of cancer genomics, complete with pathways, genetic drivers, tumor markers and the drugs that target them with increasing precision.

As Dr. Ley, Director of the Washington University genome center, phrases it in the recent New York Times article, “For the past 40 years, we have been sending generals into battle without a map of the battlefield. What we are doing now is building the map.”

When all four children in a single family began to develop hearing, speech and intellectual defects as infants, doctors performed routine medical and genetic tests to diagnose their condition. The tests revealed little, and the children were diagnosed with a recessive form of intellectual disability, with no apparent heritable cause. Intellectual disability, like most neurodevelopmental disorders, can be particularly tricky to identify because of the lack of precise diagnostic tests and difficulty of obtaining tissue biopsies. Researchers probing the genetics underlying such conditions recently sequenced the exomes of this family and over a hundred others with affected children, implicating several novel genes in the development of these disorders. Their research, published in Science Translational Medicine last month, also hints at ways to adopt exome sequencing in clinical practice. 

To identify the best candidates for exome analysis, the researchers chose families with at least two affected children, as well as a history of marrying amongst relatives. In such cases, the probability of a causative genetic variant being inherited in a recessive Mendelian pattern is significantly higher, narrowing the list of potential genetic culprits and making them easier to identify.

Extracting DNA from blood samples, the researchers began by testing for SNPs already linked to these diseases. Of the 188 affected individuals, forty turned out to have such mutations and were excluded from further genetic analysis. As the efficiency of exome sequencing and analysis improves, it may prove simpler to perform than such individual tests at multiple loci.

In thirty other families, the scientists found only one significant variant that followed the inheritance patterns expected for a potentially disease-causing mutation. In all cases, this variant was a novel SNP so far unlinked to neurodevelopmental disorders. Sequencing samples from the remaining 118 patients, the researchers identified 2-3 candidate loci per individual that were likely to be associated with their conditions. Ten of these families had mutations in known genes that were missed by the earlier genetic tests. In all these cases, the mutation identified explained all the clinical symptoms of the children, and helped to correct diagnosis or prescribe better treatment plans.

The family with four intellectually disabled children turned out to be one of these ten. Though the children had tested negative for mutations in VPSB13B, a gene linked to poor intellectual development, their exome sequences revealed a mutation in the MAN2B1 gene that produced a truncated, ineffective protein. Disruptive MAN2B1 mutations cause α-mannosidosis, a lysosomal storage disorder resulting from an inability to digest glycoproteins inside cells.   Re-evaluating the children’s clinical profiles, the team found that all their clinical symptoms could be explained by this one mutation.  Unlike recessive intellectual disability, α-mannosidosis can be treated with hematopoeitic stem cell transplants that produce the missing enzyme, which have proved effective in some patients with lysosomal storage disorders.

Variants identified in the other nine families also helped to correct or clarify diagnoses for the affected children. In all these cases, the mutations identified in exome analysis helped the researchers better understand the course of the disease or improve the therapeutic regimens prescribed to the patients. The several novel variants identified in the remaining families are still being studied for their relevance to neurodevelopmental diseases. They may also prove to be relevant in broader populations, eventually building towards a database of genetic variants associated with neurodevelopmental disorders.

Perhaps more importantly however, the results of this study hint at ways to address the larger issue of the relevance of whole exome sequencing in the clinic. The diagnostic value of exome sequencing is still emerging from a cloud of concerns about data quality, cost of interpretation, ownership of an individual’s data and when to best apply such genome-wide analysis. The filtering used in this study to identify the patients most likely benefit from exome sequencing could address some of these concerns.

Tracy J. Dixon-Salazar1,, Jennifer L. Silhavy1,, Nitin Udpa2,, Jana Schroth1,, Stephanie Bielas1,, Ashleigh E. Schaffer1,, Jesus Olvera1,, Vineet Bafna2,, Maha S. Zaki3,, Ghada H. Abdel-Salam3,, Lobna A. Mansour4,, Laila Selim4,, Sawsan Abdel-Hadi4,, Naima Marzouki5,, Tawfeg Ben-Omran6,, Nouriya A. Al-Saana7,, F. Müjgan Sonmez8,, Figen Celep9,, Matloob Azam10,, Kiley J. Hill1,, Adrienne Collazo1,, Ali G. Fenstermaker1,, Gaia Novarino1,, Naiara Akizu1,, Kiran V. Garimella11,, Carrie Sougnez11,, Carsten Russ11,, Stacey B. Gabriel11,* and, & Joseph G. Gleeson1,*,† (2012). Exome Sequencing Can Improve Diagnosis and Alter Patient Management Science Translational Medicine DOI: 10.1126/scitranslmed.3003544

I am currently pursuing a post-doctoral project at Instituto Gulbenkian de Ciência and Instituto Nacional de Saúde Dr. Ricardo Jorge (Portugal) working on the analysis of GWAS carried out by the Autism Genome Project (AGP), a large international consortium for autism genetics. My research aims to develop a network-based approach for GWAS data analysis by combining association results with protein-protein interaction data, and characterize potentially pathogenic CNVs identified in the AGP whole genome CNV analysis. 

Key advancements in genotyping, sequencing technologies and analysis methods, and the formation of consortia made large genomic screenings feasible. So far, more than 1000 GWAS studies have been conducted, spanning over 200 phenotypes. However, despite many promising achievements, most of the genome-wide association studies for complex disorders identified common risk variants that only affected disease risks to small extents. These disappointing results highlight the need for alternative approaches that shift the focus from individual markers towards a broader view of affected pathways and biological processes. Additionally, the unprecedented pace at which huge amounts of genome wide data (GWAS, CNV or expression studies) are produced creates the need for tools to integrate these multiple types and sources of data.

237 traits linked to genomic loci by 1449 GWAS studies
(Source: www.genome.gov/GWAStudies)

The main objective of my network-based approach to autism GWAS is to identify sub-networks implicated in autism and to prioritize candidate genes for follow-up analysis. NextBio Genetic Markers is a very simple and quick way to collect a list of known genes previously associated with autism that can be used as a gold standard to validate candidate genes identified in my approach. The integration of multiple types of curated content provides a source of high confidence data. After the validation of my approach and the identification of protein sub-networks possibly associated with autism, I used several NextBio apps to prioritize genes for further study. Individual gene queries helped me glance at a wide variety of information on the gene, such as the most correlated compounds, tissues and diseases. I used these to identify genes expressed in the nervous system and previously implicated in autism or other neuropsychiatric disorders. The Meta-analysis app also proved very useful to prioritize candidates by providing me data on genes with significant combined evidence from multiple studies implicating them in autism.

In my research project I have also been characterizing potentially pathogenic CNVs identified in a whole genome CNV analysis for autism susceptibility. Searches on the genes intersected by rare CNVs have helped us to select relevant CNVs to follow up, giving complete information about relevant correlations with those genes. One of the most interesting CNVs I have been working on is a small in tandem duplication encompassing the ANXA1 gene that is present in 12/3063 patients and absent in 8000 control individuals. In this case, the information retrieved from Body Atlas is highly relevant to the expression studies performed for this gene. Information from several apps, especially Knockdown Atlas have elucidated possible mechanisms by which this gene can be implicated in autism, most of them required analysis of public datasets and are not fully described in the papers, helping us in the design of follow up studies.

I become truly amazed when I look back on my career and see how a student with undergraduate degrees in Computer Science and Electronics turned into a biology researcher. In the very first week of graduate school, I received an article on microRNAs (miRNAs), a class of short non-coding regulators of protein-coding genes, from my supervisor. Ever since that first read, I have been hooked on these endogeneous RNAs for four years and counting. 

My primary focus of research covers the identification of oncogenic miRNAs. Present cancer research mainly focuses at the molecular level on the regulation of genes for studying cancer pathways. However, recent investigations highlight that there is a significant association of miRNAs in the progression of different types of cancer. We focus on exploring the involvement of miRNAs in some tumor-based and non-tumorous cancers. The Disease Atlas feature of NextBio helps me to validate the results of my analysis.

In an earlier study, we identified with in silico approaches that two miRNAs, hsa-miR-186 and hsa-miR-154, are oncogenic for leukemia and prostate. There were no earlier databases validating this hypothesis. NextBio with its large base of curated information provides more robust support. NextBio Research seems to integrate various information and crosslink them. This is why information about a gene can be obtained at diverse levels.

The Disease Atlas reveals the association of hsa-miR-186 with lymphoid leukemias along with several other significant cancer types. The hsa-miR-154 is also found to be related to prostate cancer from NextBio repository. In addition to diseases, the Atlas also reports about various other traits and conditions. Queries can also be created at the SNP level, which is likely to be very important in identifying somatic mutations. It is really a strong support for disease analysis.

Statistics is a significant tool for modern research. No research outcome is persuasive unless it is statistically significant. Most of the existing analytical tools provide results without any statistical validity. NextBio provides the results in Disease Atlas together with p-values, fold changes, copy number changes and z-scores. The ranking by statistical significance make the results very practical to researchers. It strengthens the results and makes it more justifiable. These features are unique toward various applications.

As I work on human miRNAs, the genome browser facility of NextBio also helps me to look at potential targets and analyze CpG islands in binding sites. It is just as good as other genome browsers but with a cool interface. The same goes for the Literature app. It provides a unique platform through its registered account where you have your own Inbox and can put together your data, do meta-analysis, create collaborations and much more. As a whole, NextBio is helping my research almost like an extra pair of hands. My sincere gratitude to NextBio!

Something like the image above is probably what comes to mind first when you hear ‘chromatin structure’. In reality, DNA spends only a small portion of time in these tightly coiled chromosomes. Most chromatin within cells lies in diffuse strands within nuclei. Mapping the locations of genes on these strands, scientists have found that chromatin containing more actively expressed genes tends to cluster together, as do chromatin strands containing inactive genes. Now, two recent papers show that these specific arrangements of chromatin may hold clues to developing targeted drugs for diseases like cancer.

A PNAS paper by Rickman et al. describes the impact of increasing the expression of ERG, an oncogenic transcriptional regulator on the 3-dimensional clusters formed by chromatin in cancer cells. Fusions of the ERG gene resulting in increased expression are very common in prostate cancer and associated with changes in the activity of hundreds of genes across the tumor genome. The scientists transformed the prostate cancer cell line RWPE-1 to overexpress the ERG protein, and mapped changes in interactions of chromatin throughout the genome.

They found that cells over-expressing ERG showed specific, consistent changes in the 3D organization of chromatin. These changes were associated with different expression of genes in the re-organized regions. For example, ERG-overexpressing cells showed a loss of interaction for two genes, TFF3 and TMPRSS2 relative to a GFP-overexpressing cell line. Adding the ERG protein externally to this control cell line also induced a loss of interaction of these two genes, both of which are ERG targets. TMPRSS2 is also one of the most common fusion partners for ERG in prostate cancer. Throughout the genome, ERG-binding sites were the ones with differences in chromatin loop interactions between the control (GFP) and ERG over-expressing cells.

In both cell lines, the most highly expressed genes tended to group into chromatin clusters the authors describe as ‘transcriptional hubs’. Comparing the functions of genes in these high activity regions in the two, they found that the chromatin clusters in the ERG-expressing cell line had more genes associated with cell motility, invasion and de-differentiation of prostate-specific cells. Not all these changes lasted consistently through all ERG-overexpressing cells, however. Some of these chromatin interactions depended on what phase of cell division cells were going through- dividing cells in the G2/M phases had different interactions than cells in the G1 growth phase. When the researchers repeated these experiments in DU-145, a different cell line, only some of the chromatin interactions were the same as those in RWPE-1.

Organizing chromatin loops based on activity levels into highly active (or repressed) clusters is an efficient way for cells to modify the functions of groups of genes. Tumors with ERG gene fusion events behave differently from other tumors, enough to be considered a distinct sub-type of prostate cancer. This study shows that the gene expression signatures in cancers with such gene fusions are also associated with large-scale chromatin restructuring. As the authors also point out, the link between chromatin structure and gene expression is not necessarily a cause and effect relationship. However, a related study in Cell last week adds some perspective on this aspect. In the Cell paper, authors Deng et al. used a zinc-finger binding protein to force distant regions of chromatin into looping together. They found that these forced re-arrangements of chromatin could increase or decrease expression of genes within the restructured loops.

Designing molecules that loop or un-loop specific regions of DNA could be a quick way to increase or decrease the activity of clusters of tumor-associated genes. Many recent studies sequencing tumor genomes highlight how each tumor is unique. In some cases, even cells within a single tumor may have distinctly different genomic changes. The links between these sequence-level changes and larger scale changes in chromosome structure are still murky. By studying the 3D maps of cancer genomes described in these papers, scientists may be able to find common chromosome-scale patterns that could correlate to sequence-level changes like the ERG gene fusions.

Skin cancer remains one of the most common forms of the disease in the U.S. Though melanomas account for less than 5% of all skin cancers, they cause the large majority of deaths from the disease. Some parts of the skin (like the palms and soles) develop cancers much less frequently, and the rare ‘acral melanomas’ that originate from these types of skin have less in common with other skin cancers.

Understanding the genomic differences in these sub-types of melanoma can help to identify drugs that target cancer-specific genetic changes in patients. As part of this larger goal, an open-access paper in Nature earlier this month describes the exome sequences of twenty-five metastatic melanoma samples. Michael Berger and his colleagues compared the sequences from these tumors to matched blood samples from the same patients to identify differences between the cancer sample and normal tissue. Two of these twenty-five samples were acral melanomas, and one of the cutaneous tumors studied originated from a patient with a long history of chronic sun exposure.

Their analysis identified distinct similarities and differences in the genomic changes in each of these tumor types. The tumor with long-term UV exposure had the highest frequency of point mutations, or single base changes like transversions and substitutions, the most common type of DNA damage associated with UV irradiation, whereas the acral melanomas had the lowest frequency of such changes.

On the other hand, genomic re-arrangements, or mutations where chunks of chromosomes are randomly broken off and spliced, were common in both acral and cutaneous melanomas. Many of these re-arranged chromosomal regions contained large genes implicated in cancer-related processes, such as the tumor suppressors CSMD1 and PTEN and MAGI2, which binds and stabilizes PTEN. Other genes included FHIT, MACROD2 and A2BP1. The well-studied BRAF^V600E mutation was found in 16 of the 25 tumors, the rest had mutations in the NRAS gene. BRAF and NRAS mutations occurred in a mutually exclusive fashion in all tumors except the hyper-mutated sample. The researchers also identified PREX2 mutations in many of the tumor samples sequenced; one acral melanoma sample had nine different mutations in the gene. To see whether this novel gene could be a significant biomarker for melanoma, the researchers screened an additional set of 107 samples for PREX2 mutations, and found them in a significant fraction (14%) of these tumors.

The mutational signatures of sub-types of melanoma identified in these samples are similar to those described in previous studies. In particular, this research confirms the prevalence of genomic re-arrangements in acral melanomas, and the increased mutation frequency seen with greater exposure to UV-rays and sunlight. (It even validates a song a little further.)

These results are important both for choosing the best treatments for patients, and for researchers working on more precise ways to treat melanoma. For example, the recently approved drug vemurafenib specifically targets the BRAF^V600E mutation found in many of these tumors. Could the PREX2 mutations identified in this study be used to develop similar drugs? Given the diverse PREX2 mutations even in this small cohort, it seems unlikely that a single mutation in the gene plays a critical role in driving melanoma development. The different mutations of PREX2 seem to converge on a single property, where cells with the mutant gene grow into tumors faster and make more of the PREX2 protein. Rather than targeting specific SNPs, researchers may be able to use such data to identify metabolic pathways or regulatory mechanisms that are good therapeutic targets to develop drugs.