When my mother was fighting cancer, I recall the many disappointments of finding research shrinking tumours in animal models, only to find out the human research showing it didnt work in humans. This was in the 2010s, before llms, but when google search actually searched the web. Then, once you found something that seemed to work in humans, you were hit with the realization that ‘cancer’ is an umbrella term, and you need to account for cell type, and its numerous mutations.. I think the best approach is to collect a sample of the cancer, genotype it, test it against all known anticancer compounds, similar to how you’d deal with a bacterial infection sample, and then hope that the compound that worked for that specific cancer cell will work inside the human