If you want to cure a cancer, identify it as soon as possible. That’s long been a truism among medical experts. But what if your techniques were so advanced that you could identify the precursor steps to cancers before they had even started to initiate a tumour in someone’s body? According to an article by Anna Bawden and Nicola Davis of the Guardian, the first steps to do just that are now feasible, as two studies suggest.
Instead of simply rehashing their excellent prose[1] upon which we urge you to click, we’ll provide a brief summary, and raise some interesting and rather hopeful observations. The first looked at 44000 samples from the UK Biobank. 618 proteins were identified, which could then be linked to 19 different types of cancer. In a different take on the same trope, a second study using a whopping 300 000 samples came up with 40 different proteins linked to 9 different types of cancer. We dare not comment, but dare to observe:
1 It’s amazing the amount of new discoveries you can make just by crunching data. As AI comes into its own, it should be able to handle bigger and bigger numbers. Think of alpha-fold, if you don’t believe us-and that quite old hat by now!
2 Talking of hats, let’s all take ours off to Cancer Research UK, whose steady, patient work down the decades has not only provided a congenial ecosystem for researchers, but also a steady stream of reliable income for the planners and the finance people. Come on, hands in pockets, please! [2]
3 We were impressed that the results were already identifying different types and subtypes of cancers. It suggests a subtlety of technique which has probably only just got going.
and, finally:
4 The bigger the database, the better. Without belittling today’s researchers and journalists, these are still relatively small numbers. Imagine an AI supercomputer tirelessly combing the biological samples of every human on the planet. And maybe their pets. Would might it not find.
Oaks and acorns times, gentle readers. Keep donating.
#database #cancer #medicine #AI #protein #gene #prediction
Learning, Science and Society 