Computational Cancer

The need for computational approaches in cancer research

Although tremendous progress in cancer research has been made, allowing many patients with potentially fatal cancers to enjoy a prolonged life, and often to experience a cure, many others afflicted with this disease continue to suffer and die as a result of their cancer. Our challenge is to cure the substantial number of patients whose cancers either cannot be prevented, evade early detection, or resist conventional therapy.

At a molecular level, cancer is not several diseases, but many thousands of different diseases. Most cancers arise from the progeny of a single cell. For malignancy to occur, a number of mistakes must become established in the DNA of the descendants of a single cell. Some of these genetic errors “jam the accelerator” of cell growth, whereas others “loose the brakes” on cell division, or prevent the normal “programmed death” of cells that are damaged or no longer needed. Still other genetic mistakes prevent proofreading and correction of errors that occur in DNA replication during cell division. The mutations required to create a cancerous tumor can occur in any of the large number of genes that regulate cell growth, programmed cell death and DNA repair. Consequently, the number of possible combinations and permutations of cancer causing mutations is enormous.

The cancer research community is poised to make substantial strides in understanding carcinogenesis and in applying that knowledge to more effective prevention, detection and treatment of cancer. The sequencing of the human genome has initiated an explosion of research into the detailed proteomic pathways and metabolic processes of the cell. Advances in high-throughput genetic, proteomic, and metabolomic analysis technologies now enable the assembly of vast databases of detailed genomic and proteomic profiles obtained from cancer patients not only during the course of their disease, from precancerous to metastasis, but also during treatment. Analyses of these profiles promise to identify the specific mutations that cause different cancers and the specific cellular faults they create, to create accurate robust tests for each possible variation, and consequently to enable the design of patient-specific treatments.

A plethora of computationally intensive techniques in bioinformatics, biomathematics, and biostatistics promise to increase our detailed understanding of the molecular and genetic mechanisms of cancer and, consequently, the huge variation in clinical outcome observed in different patients, where a cancer that can prove lethal within months in one patient may remain dormant in another for a decade. This molecular heterogeneity also underlies the sensitivity of some cancers and the resistance of others to any single chemotherapeutic drug. Knowing the sequence of all human genes and having the ability to determine which of these 30,000 or so genes are actually turned on, turned off or altered in cancer should permit identifying which genes are actually altered in a given patient’s cancer. Computationally intensive techniques will be essential to enabling such advances.

Computationally intensive techniques also promise to enable effective responses to the heterogeneity of different cancers by enabling treatments to be devised that are specific to the particular variant of cancer and the individual in which it occurs. For instance, the availability of drugs that repair or bypass specific genetic defects, such as Gleevec in chronic myelogenous leukemia, should permit physicians to link molecular diagnostics with molecular therapeutics and to write specific prescriptions for each patient that would correct or bypass those defects that were actually present in their cancer cells. Such treatment could be more effective and less toxic than current chemotherapy. In the past, clinical investigators have had to devise the best treatment for the “average” patient, knowing full well that cancers from different patients differ in their clinical behavior and susceptibility to treatment and that patients also differ in their susceptibility to the side effects of a given drug. Computationally intensive techniques will accelerate the rate at which the need for specific drugs can be identified, the rate at which candidate drugs can be identified or synthesized, and the rate at which they can be evaluated for safety and efficacy in the targeted population.

For these promises to be realized, advanced computation is critical. Merely identifying the enormous number of possible combinations and permutations of cancer-causing mutations is a significant challenge, requiring computationally intensive analyses of large databases. For developing effective treatments, manual methods are impossibly slow. High-level computational approaches and advances are essential.

Development of the software required to realize this vision of “making cancer history” is exceedingly difficult. The rapid advance of the enabling technologies has created a huge number of opportunities for understanding carcinogenesis and its treatment, but each of these opportunities requires the development of custom, high-performance software. Not only must these software capabilities be developed, but the software must also be made to run efficiently on modern, high-performance computer architectures. These are challenging tasks requiring specialist skills. Consequently, the new applications needed for achieving the dramatic advances enabled by advanced computational technologies can only be created at a very limited rate. The research fostered by the GC4R aims to eliminate these impediments and increase the pace at which novel computational tools can be created and applied to the prevention, detection, and treatment of cancer.