There were countless kinds of cancer cells, and under different expression conditions, their surface antigens were extremely complex, with all sorts of markers as well.
For example, there were common carcinoembryonic antigens highly associated with gastric cancer, colon cancer, and pancreatic cancer; alpha-fetoprotein associated with testicular cancer, gastric cancer, liver cancer, and so on; and carbohydrate antigen 19-9 associated with gallbladder cancer, colon cancer, and the like.
In addition, there were all sorts of common tumor markers such as neuron-specific enolase, prostate-specific antigen, and squamous cell carcinoma antigen.
There were a great many tumor markers. The sensitivity or specificity of a single marker was often rather low, and under certain physiological conditions or with certain benign diseases, some tumor markers could also appear, so they were difficult to use as the sole basis for tumor diagnosis.
Even by the era before Murphy was reborn, automatically identifying cancer cells by non-artificial means was still extremely difficult.
After thinking it over, Murphy ultimately decided to use evolutionary algorithms and deep learning to train the viruses.
The general idea was this: take several or several dozen more invasive viruses or bacteria and form them into an experimental population, then continuously feed them all kinds of normal cells and cancer cells, while manually labeling the various cells through Seed of the Mind to complete the training.
These labels included: whether it was a cancer cell, the pH of the cellular tissue, cell morphology, and the characteristics of various known normal cells and cancer cells, among others.
The feedback mechanism was set to be extremely harsh. Any pathogenic individual that killed normal cells more than three times, or failed to kill cancer cells more than three times, would disintegrate and die.
Through this method, combined with trace amounts of Dittany Promoter to accelerate evolution, calculations