Cancer is a disease of the genome. Mutations occur over time in the DNA of every cell. Most of these mutations are benign; they have no affect on the normal operation of the cell. Some mutations or combinations of mutations can be deleterious. If the mutation causes the cell to die, the result is that a single cell dies and there is likely little affect on the organism as a whole. Some mutations, however, may affect the cell cycle (cell replication) in such a way that the cell replicates uncontrollably. Since all of the uncontrollably replicating cell’s progeny have the same mutations, they and their progeny also replicate uncontrollably. Since these cells grow and divide much more quickly than the surrounding cells, these tumor cells quickly come to dominate the tissue in which they originated, starving other cells in the tissue of nutrients. The rapid replication of cells often leads to more mutations being accumulated until the tumor metastasizes and the cancer spreads throughout the body. Over the past few decades, cancer care has improved for many types of cancer as early detection has increased and various new chemotherapeutics have been developed. More recently, drugs that target specific cell processes, or pathways, whose increase or decrease in function due to mutations play a role in cancer have been developed and used effectively in the clinic. Unfortunately, the many diseases that fall under the name cancer are widely varied and extremely complex. Breakthroughs in understanding and treating one subtype of cancer may not be applicable to other cancers or even other subtypes of the same cancer. Indeed, even within a tumor there can be heterogeneity; all cells will have the mutations that led to tumorigenesis but different subpopulations within the tumor may appear, each accumulating its own specific mutations. Thus we need a better molecular, DNA-level understanding of the many diseases collectively called cancer. In other words, for each tumor we need to understand what mutations lead to tumorigenesis, how those mutations affected cell replication, and how to reverse that change of function.

To begin to discover more about the molecular nature of cancer, several years ago we began the Tumor Sequencing Project. In TSP, we use both array-based technologies and DNA sequencing to study tumor and normal tissues from a cohort of about 200 patients with lung adenocarcinoma, a form of lung cancer that affects a disproportionate amount of never-smokers. SNP arrays were used to determine copy-number variation (CNV). The CNV work was published last year in Nature, Characterizing the cancer genome in lung adenocarcinoma. While the CNV work provided valuable information about large-scale differences between the tumor and normal tissues over the entire genome, it and other array-based approaches cannot provided detailed information about specific mutations. To get a single base level resolution of the differences between an individual’s tumor and normal genomes, you have to sequence the DNA. At the time TSP started, the cost of sequencing an entire genome was much greater than running a SNP array. Therefore, the low-resolution information from our SNP array investigations and other, similar investigations in the literature were used to guide the selection of specific genes thought to play a role in cancer in general and lung adenocarcinoma in particular. For each patient, we sequenced these genes in both their tumor and normal tissues and looked for differences between the two sequences. Once you have those differences, genes annotations (definitions) can be used to determine if the mutation in the tumor tissue would lead to a change in the amino acid sequence of the protein that gene encodes. If the protein would be changed, other programs can be used to predict whether the change would lead to a change in protein conformation and a possible loss or gain of function. The cellular pathways in which the mutated protein participates can also be researched to see if any of the pathways control cell division or other important signaling paths in the cell cycle. Once the gene and its pathways are known, they can be correlated with known oncogenes and oncogene families. Once that is established, further experiments, e.g., induction of found mutations in disease model organisms, can be done to determine if the mutations and predicted changes do contribute to carcinogenesis. It is also interesting to see how changes that span several patients correlate with phenotypic/patient information, e.g., cancer subtype, patient outcome, age, sex, and smoking status. The initial sequencing results of the TSP will be in tomorrow’s issue of Nature, Somatic mutations affect key pathways in lung adenocarcinoma.

The Cancer Genome Atlas pilot project, which is studying brain, ovarian, and lung cancer, has followed much the same path as TSP, albeit on a larger scale. Compared to TSP, a wider variety of array platforms, more patients, and DNA sequencing of more genes have been employed. Initial gene lists for sequencing were obtained through a review of the literature with subsequent lists obtained through the results of the array-based studies. The initial glioblastoma paper from TCGA was published online by Nature last month and, like the TSP paper, will be in tomorrow’s issue, Comprehensive genomic characterization defines human glioblastoma genes and core pathways.

With both TSP and TCGA, we are able to look at the whole genome in a coarse-grained way and specific parts of the genome (genes) in a fine-grained way. So the question becomes, what are we missing by not studying the entire genome at single base resolution? The advent of next-generation/massively parallel sequencing has allowed us to begin to answer this question. About a year and a half ago, we began whole-genome sequencing of a single AML patient’s tumor and normal genomes using the Illumina/Solexa sequencing platform. Generating such a complete picture of a single human genome raised significant privacy issues that needed to be addressed before publishing and data release. The massive amount of data generated for each of these genomes presented significant challenges to our informatics infrastructure. All of the data had to be combed through to find the small variants (single nucleotide variations and small insertions and deletions of sequence) between the tumor and normal genomes (at the time, it was not possible to detect larger variations, e.g., structural rearrangements, with the Illumina/Solexa technology). With this approach, you are not only able to find mutations in genes that have not previously been implicated in cancer, but you are also able to find mutations in sequences conserved across species, microRNAs, regulatory regions, etc.; basically anything annotated in Ensembl or UCSC. Of course, the effect of mutations in non-genic regions are more difficult to interpret than those in genic regions that alter proteins, but the efforts of the ENCODE project, disease model organism studies, and sequencing more cancer genomes will greatly aid in the interpretation of these mutations.

In summary, these cancer sequencing efforts, especially whole-genome cancer sequencing, are increasing our understanding of cancer at a very rapid pace, laying the groundwork for more individualized approaches to treatment. As our knowledge of tumorigenesis increases, our ability to detect cancers early and treat cancers effectively will also increase. Ultimately, these sorts of studies will make the goal of ending suffering and death from cancer achievable.

Update: Check out my colleague’s post on this subject: TCGA: The billion-dollar cancer project.

Here is another, more detailed article I was interviewed for in Bio-IT World titled SNPing Away at Genome-Wide Disease Association Studies. It is a good overview of whole-genome array studies and how they will soon give way to whole genome sequencing using next-generation sequencers.

Despite some unjustified grumblings about “big science”, it is a good, very important project that will contribute greatly to the NCI‘s goal of ending suffering and death from cancer by the middle of the next decade. Unfortunately, one of the biggest positives of this project is also one of its biggest challenges. Bringing together all these disparate data sources is monumentally challenging. Even comparing different platforms that ostensibly do the same thing, e.g., CNV using either Affymetrix or Illumina platforms, can be hard to normalize and cross compare. Add to that clinical, methylation, expression, segmental duplication, rearrangements, and sequencing data and you have a real data integration problem. Then after you integrate all the data, you have to make sense of it all. Oh, and you have to do it very reliably at high throughput.

Getting back to the Data Portal Use Case Workshop, it was apparent that there will be a large and very diverse audience that will want to access these data in a large number of very different ways. Some people will want to start from the patient samples and see what results correlate with those groupings. Some will want to looks at specific patient clinical information and see if longevity correlates with anything. Some researchers will want to see if their favorite gene is sequenced or if any of the genes in the pathway they study have mutations. Some will want to access the data from a genomic or chromosomal standpoint, finding areas of interest and drilling down to see why they are interesting. And so on and so on. Regardless of where they start, researchers and clinicians will want to slice and dice the data as many ways as they can to find correlations and insights.

So how do you design a data portal that addresses all these needs? How do you design a data model that ties all this data together and allows each of the different use cases described above (and many more that were not thought of during the workshop) to be pursued? How do you display this highly multidimensional data, allowing the user to zoom in and out, and layering on more information without overwhelming the user? It is a huge challenge worthy of a research project in and of itself. Unfortunately, the people at the workshop seemed to be falling back to what they know: the UCSC browser, GenePattern, the Cancer Genome Workbench, etc. Not that these aren’t good tools, but the problem and the audience are much bigger now. So is the price of failure. We need to think bigger.