Written by: Stephen Hsu
Primary Source: Information Processing, 01/30/2019.
As I mentioned in this earlier post, I’ll be in the UK next week for a Ditchley Foundation conference on the intersection of machine learning and genetic engineering.
I’ll present these slides at the meeting.
The slides review the rapidly evolving situation in genomic prediction, focusing on disease risk predicted using inexpensive genotyping. There are now 10-20 disease conditions for which we can identify, e.g., the top 1% outliers with 5-10x normal risk for the disease. The papers reporting these results have almost all appeared within the last year or so!
On the last slide I give a simple cost-benefit analysis of population wide genotyping and conclude that the net benefit is already positive given the tools we have. The numbers used are per capita. The UK NHS is already headed in this direction.
I use breast cancer as the example on the slide, but since the same genotype can be used for 10+ disease risks (including diabetes, atrial fibrillation, hypothyroidism, etc.) the net benefit is potentially much larger than what is obtained from breast cancer alone. The point is that G is really small compared to the potential benefit.
Details of breast cancer calculation below. I am sure one can do much better, but it provides a quick back of the envelope estimate of the numbers.
Spend $100 per person to genotype all women in the population. Identify those with top 10% risk score. About 33% of these individuals will get breast cancer. Treat the risk outliers by giving them, e.g., regular mammograms starting a decade earlier than usual (~$100 annual mammogram x 10y = $1k). In the slide I assume the average cost of the intervention / treatment is $1k and the average benefit is $12k. All of the high risk women (10%) get the intervention, but only the 33% percent that get breast cancer (or some subset of that group) benefit from early detection. This paper estimates that early detection of breast cancer saves typically tens of thousands of dollars per individual, so my numbers are conservative. If one uses multiple tens of thousands as the benefit amount, one could spend much more on early treatment and still have a positive net benefit.
Latest posts by Stephen Hsu (see all)
- Brainpower Matters: The French H-Bomb - August 16, 2019
- MSU Research Update (video) - April 17, 2019
- Interview with Genetic Engineering & Biotechnology News - April 17, 2019