In 2013, the National Cancer Institute of the USA announced a new program, to catalog “exceptional responders” in cancer trials. This program aims to identify a small number of well-responding patients to new treatments, with the hope that in the future, patients with cancer could be treated with the most effective drugs (personalized therapy). In this paper, I extrapolate on this idea and propose to also catalog cancer biomarkers that only work in a minor proportion of patients, and are currently ignored as clinically useless. Such biomarkers could be used to select optimal treatments, optimal monitoring or for assessing prognosis. The informative biomarkers for these rare patients may also provide the opportunity to identify molecular networks that are altered in cancer and explain why these markers are elevated in these few patients. I provide an example of two kallikreins (KLK6 and KLK10), which are highly elevated in serum of 3%-5% of pancreatic cancer patients at 100% specificity.