If one was to track every cell from birth do adult in 100 kg of biomass (corresponding to a 220 pound person) one would have to make an entry for each of approximately 100 trillion cells. The table that tracks specific instances of cells is much taller than it is wide. 100 trillion entries tall, 32,000 ± wide. A large table, 3.2 million trillion entries, or 320 thousand gigabytes, to first order.

In a fine needle biopsy of a suspected tumor, a micro-coring penetrates consecutive layers. These consecutive layers consist of a several tissue types, only one of which may be of concern. It is necessary to microdissect those cells from the coring that are histologically significant from a tumorigenesis point of view. Even after microdissection any subsequent gene expression analysis is typically made from an ensemble of cells, not necessarily from the specific cancer clone that gave rise to the tumor. This implies that the gene expression studies may be diluted with the expression of normal cells whose mRNA's are inadvertently thrown into the batch with those of the cancer clone.

One of the problems of cancer cells, is that they are, immunologically speaking, cells of self. Thus it is difficult to devise immuno based strategies that preferentially attack the cancer cells, that do not also cause the problems in the patient's normal tissues. There are tumor markers, that are are unique to cancer cells, but these are not universally expressed in all cancers or at all times. Cancer is nothing more (or less!) than DNA damage. Further it is "nothing more" than six to ten accumulated mutations in growth regulators, cell cycle checkpoints, oncogenes and the like. This makes the problem of specifically targeting the cancer cell peculiarly difficult even in the face of growing understanding of human genome sequence. Your cancer belongs to you and no one else. Remember that.

I had held out the hope that cell simulation would open the door to understanding cancer to the point of treating it more effectively. The cancer clone that gives rise to a specific line of cancer cells. Simulating the cancer clone is akin to simulating one of the patient's own cells, with a few important differences. Cancer cells are a little different than the patients normal cells, but not enough different to be easily recognized. Cancer cells typically divide rapidly without going through the DNA quality control checkpoints (apologies for the anthropomorphic POV) and without triggering the growth arrest triggers via apoptotic/ caspase pathways. These cells are slightly "weaker". They incur progressively more damage as they multiply, else therapies which exploit their weakness would not succeed. Cancer cells are more primal and express a broader range of gene products. But alas these gene products are most often products that you would find in the normal patient, albeit in overexpressed quantities.

Without being excessively pessimistic my concern is that, even if one could completely simulate the cancerous clone, one would simply have recreated the tragic portrait of a broken system, without providing the critically important handle on fixing it. Fixes, if they do exist, depend on exploiting very specific mutations in the cancer clone that would belong to the individual. Cancer is personal. Thus, you will not find me buying shares in any new "cancer wonder drug" since any plausible drug would have to be patient specific.

This tells us that for near term, practical simulations, we want to identify those cells who are known to conspire in the cancer process, such as fibroblasts and ductile epithelial cells in breast cancer. The direct simulation of the normal and abnormal gene expression states of these selected cells will produce the highest benefit per unit of cost, whatever that benefit might be.