Hacker News new | ask | show | jobs
by cyberax 919 days ago
That's because mouse models suck for many diseases. Cancer very much included, for two reasons:

1. Mice _love_ to get cancer naturally. If you have 100 mice, it's quite likely that around 20 of them will die within one year of cancer. This makes it difficult to extract useful signals. BTW, that's why if you see a study that a "chemical X results in cancer in mice", you should take that with a grain of salt.

2. Mice are small, so tumors are necessarily small too, with several orders of magnitude fewer cells than typical human tumors. So many drugs can just cure mice of cancer entirely, by killing cancerous cells too quickly to allow them to evolve resistance.
2 comments
maratc 918 days ago
> If you have 100 mice, it's quite likely that around 20 of them will die within one year of cancer.

If you have 100 people, about 40 of them will be diagnosed with cancer over their lifetime (39/100 females and 41/100 males). Note "diagnosed with" is very different from "die" and "over lifetime" is not the same as "within one year", but the probability of people getting cancer naturally is high as well.

link
cyberax 916 days ago
Sure. But in general, cancer risk should scale with the number of cell divisions, which is fairly low for mice. And other small, short-lived mammals like weasels are not as susceptible to cancer. Naked mole rats (also rodents!) are downright cancer-proof.

This all means that mice don't have a good natural cancer-suppression machinery, and this in turn makes them somewhat awkward to use to discover new treatments.

link
glenstein 918 days ago
I'm not sure it follows that high incidences of cancer means that it's unhelpful to use them for research to understand the mechanisms of cancer. And I'm not sure that either of your observations one or two are pertinent in this case as it applies to the on/off protein switch, as it is not about guaging the statistical frequency of cancers or about killing cancer cells.
link
cyberax 917 days ago
Murine models are not unhelpful, just very tricky. Quite a few experimental data turned out to be incorrect because experiments were poorly conducted.
link