My daughter has been learning about T cells in her biology lessons at school. She calls them "killer cells" which go round in our blood and attack virus-infected cells and cancer cells. Today I have been learning about machine learning technology which takes in a gene sequence from a T cell and predicts what that T cell is currently trying to attack, or has attacked in the past.
Predictions like this from a sample of my T cells can potentially be used to suggest possible diagnosis of cancer, infections and autoimmune diseases. It would be a simple, fast and cost effective blood test and demonstrates the power of machine learning to improve lives. In my view, the predictions would need to be confirmed using other methods of diagnosis because machine learning predictions do not tell us about cause and effect, by their nature are uncertain, and can contain errors such as in the case of adversarial examples.
An alternative to using machine learning would be to manually find the mapping from T cell gene sequences to antigens. But it would take a lot of work!
The content above was originally posted on CMS DigitalBytes - CMS lawyers sharing comment and commentary on all things tech.