AI, Genome Sequencing and New Sensors Creates Pre-Cancer Management

Steve Hsu, the cofounder of Genomic Prediction, was interviewed by Scott Adams. He talked about using AI and genome sequencing to detect and predict disease. Predictors now exist for about 20 diseases.

Steve predicts that soon…DNA tests will be required for health insurance policies.

We have been able to detect single genes that substantially increased risks for disease. Now we can detect more complex situations that increase risks. This could be dozens of genes combining to increase risks. In breast cancer, one in a thousand women have genes that increase risk for breast cancer. Those people have been required to have far more frequent mammogram tests. Now Steve Hsu’s company will be able to identify ten times as many people who have more complex disease risks.

Brian Wang of Nextbigfuture had a prediction in 2006 that insurance companies would require regular tests to track biomarkers and biomarkers would have to be maintained in a prescribed range.

Pre-Cancer Management

Cancer in its late stages becomes an exponential explosion of bad cells. Pre-cancer management would be working to defuse the bomb before it explodes. If we can precisely determine that disease will occur in x years then we will have time to pre-empt the problems.

I believe that cancer medicine will change to exact forecasts of the type and timing of pre-cancer and management of pre-cancer. We will be able to decide the best time and method of pre-cancer interventions.

How much does early detection matter. If we could catch all cancers at stage 1, stage 0 or even just before cancer develops then over 90% of those with early breast cancer would survive vs 15% for late detection and similar numbers for most other cancer.

We have several problems with diagnosing cancer. Many people do not get the frequency of tests that can find cancers early. This can be for good reasons. The tests can be a painful hassle and expensive. Cancers can be missed in testing and their can over-diagnosis of 30% in some cases.

Right now, People have better health monitoring and more frequent inspections of their car than their own bodies.

There are 4 cancer medicine breakthroughs

1) You already know about Steve Hsu’s startup company called Genomic Prediction is combining AI with genetic sequencing. They provide a report with your top ten highest genetic risks for disease. Those with 5X or 10X higher likelihood for specific cancers can choose to have a stepped-up rate of testing and a lifestyle coaching app to attempt to prevent the actual emergence of the cancer.

2) AI is improving the analysis of X-ray and other diagnosis images. AI is already as good as or better than the best human medical specialists in the detection of cancer. AI imaging and tests will become frequent and cheap. An AI from South Korea called Lunit had a 97% detection rate for lung and breast cancer in a 2016 competition.

3) MIT SAIL can predict breast cancer 5 years before the disease starts with 99% accuracy. Getting an accurate near-term pre-cancer diagnosis for the deadliest cancers like lung and pancreatic cancer would be huge to rapidly improve beyond stage 1 survival rates. Pre-diabetes indicates that 50-70% of people can make the lifestyle changes to prevent the development of diabetes. Cancer medicine can make a big advance not just from aggressively working to prevent the disease but placing monitoring to catch all cancer cases in stage 0 or stage 1. There is also the need for cheap and accurate monitoring. Some tumors need to be watched for months to make sure a tumor actually requires treatment.

Stage 0 means there’s no cancer, only abnormal cells with the potential to become cancer. This is also called carcinoma in situ.
Stage I means the cancer is small and only in one area. This is also called early-stage cancer.

4) There is a new wave of new blood, saliva and urine tests which could become like today’s body thermometers for detecting a fever. Wearable devices have been tested which continuously capture cancer cells directly from the vein, screening patient’s blood.

AI analysis combined with accurate and more comprehensive monitoring will be far better than once a year checkups.

SOURCES – Steve Hsu, Scott Adams, American Cancer Society, MIT SAIL
Written By Brian Wang, Nextbigfuture.com

8 thoughts on “AI, Genome Sequencing and New Sensors Creates Pre-Cancer Management”

  1. To hell with breast cancer. Only a fraction of all women will get breast cancer but 100% of all men will get prostate cancer if they live long enough.

  2. Requiring DNA sequencing for health insurance would run afoul of the ACA provision to cover pre-existing conditions in the US. It will be a covered service more and more often as it becomes seen as another tool that can be used to identify and stratify populations by risk. I can see, for example, the presence of the BRCA1 gene being used to recommend and authorize more frequent mammograms in order to identify and start treating breast cancer earlier in a known higher risk population.

  3. Being able to catch past criminal behavior once DNA genotyping became affordable has been obvious for decades. Not necessarily obvious to police:
    https://www.globalresearch.ca/us-court-ruled-you-can-be-too-smart-to-be-a-cop/5420630
    looks like 114 IQ max. Very sad. If you are bright, I suppose you can fake relative stupid for a test. Might be tricky to get it in the sweet spot.

    And if some allegations are true, Trump may not be eager for them to start testing every kit. 🙂

    They did not talk about using AI to figure out what someone looks like from DNA, which they can do now: https://snapshot.parabon-nanolabs.com/
    It is pretty good, not perfect. Foreheads, scull width, jaw width, and eyebrows tend to be off. They probably did not train on enough features.

    One interesting thing is that if you can do this, you can also tell what an embryo will look like as an adult. Though obviously, they would have to train on more features to be accurate enough to compare embryos with the same parents. Women getting anonymous sperm donation though, could find this useful as is.

    Another interesting thing is that you can use a tooth say of a Denisovan and see what they may have looked like. Accuracy will not be phenomenal, but it would still be very interesting. It still might be recognizable as a Denisovan to a Denisovan. And thus some Denisovan may have looked very much like that even if the one the tooth came from did not look precisely like that.

  4. These things have been obvious for years. Though these people have not really thought things through. Being able to tell if you are going to get something from your genome does not mean you have to go to single payer. It is sufficient to make it illegal for companies to ask for that info or obtain it in any other way (though you would have to be diligent and also look at the rates of disease of those they insured. if they look too lucky, they probably are). They are still going to insure some number of people and the burden will be shared. Secondly, no, they probably will not be buying the cut rate plan. They will want good care. They will pay for the higher end care…because they will still be saving money and maybe their life.

    No, the real issue is people opting out because they figured out that they have great genes and will not be getting most of the typical diseases and only start getting insured at 96. Then the insurance companies are out the premiums these people would have been paying in for decades.

    All this said, single payer would be vastly better and potentially far cheaper than what the US has. We pay vastly more than anyone else. Ours is more than twice the cost of Canada’s: https://en.wikipedia.org/wiki/List_of_countries_by_total_health_expenditure_per_capita

    Even the Swiss and the Luxembourgers can’t squander as much as we do.

  5. If AI can learn to make good chess moves, well beyond merely superhuman in just a few hours of training…well.

    Though, what is needed is a few million mammograms, maybe a few hundred million. These things need a lot of training. But if you get the data, and the data is all verified accurate (probably by waiting to see who definitely got it), it really should be possible to get very accurate diagnoses.

    If we were serious about it, and got everyone’s mammogram in the U.S. in a data set, the accuracy that could be obtained would be outstanding. Combine that with E.U., Chinese, Russian, Japanese data…and, yes, the mammogram readers would be out of a job. Getting this going with current privacy laws and such is not so easy.

    And it is like landing robots on the Moon, you sorta want people on the Moon. Similarly, even if the machines says, “yes, you have breast cancer”, you still want the human to say “yup, that is what I see too”.

    The Chinese will probably beat everyone because they can make and break any rule they want.

  6. I work in health care. I feel like I am constantly seeing overblown claims regarding what AI is capable of doing with cancer diagnostics. It’s as if reporters swallow marketing claims whole without attempting due diligence in double checking the research or in consulting practicing experts. If AI is truly better than a traditional radiologist, why does it have zero market presence? Believe me, there are no real marketplace barriers that would stop AI from taking doctor jobs. That is, if it can be done just as well for a fraction of the price, it will happen. But my radiologist colleagues are continuing to churn along day after day with no concern at all that they might be about to lose their jobs. And none of them are talking about buying an AI system to help them with day to day tasks.

    I am not saying that there will never be a place for AI in day to day medical diagnostic work, but I feel like I would know if there was anything viable on the horizon that could be of real, tangible assistance to practicing clinicians. Just my 2 cents.

  7. makes me want to buy into the company.
    The rape kit information was news to me. There is an issue that screams for funding.
    And I guess single payer health care is going to happen.
    On the whole I think the benefits outweigh the negatives, but man there are going to be some kicking and screaming folks along the way.

Comments are closed.