Question

What is the relationship between risk prediction and genomics, particularly within the example of predicting the risk for developing breast cancer?

Answer 1

@BioEpic Depends on the level you want to investigate. If we look at the so-called breast cancer gene, brca1, we know that it greatly increases the risk of women developing breast cancer. However, if you look at individuals the predictions get much more hazy. You should think of the "odds" of this and similar diseases as a 3-dimensional field that is influenced by genotype, age, behavior, and environment. If you look at a persons "field" you would see random valleys and hills, maybe some mountains, and any odd landscape you can think of. This is what the real model for predicting some ones risk of any disease would be.

We can have good, solid predictions in large populations and still fail at small populations is because large populations are able to minimize the effect of randomness in sampling. Meaning, if we want to find the average rate of "heads" in a coin flip, we will have more accurate results if we run 500 flips vs. 5 flips. If we have 250/500 + 4 we would still say the odds are 50/50 for heads/tails. Now, if we have only 5 flips and have 4 heads we have a 4/1 heads/tails which says head are 80% and tails are 20% and is obviously very wrong (assumed fair coin).

Answer 2

Thanks!