Computers Vs. Cancer

By Joshua Cullinan

Pancreatic cancer is one of the cancers that oncologists truly fear. Bad outcomes, few treatments and short life expectancies are rampant in this disease. With 1 in 10 patients surviving five years after diagnosis, pancreatic cancer provides a true challenge to doctors and scientists alike [1]. Sounds scary, right? Well, even scarier is the idea that we can’t even classify pancreatic cancer. Currently there is no consensus on a robust classification system. Some studies have suggested that there are up to five types of pancreatic cancers, whilst others have suggested two and still others three and four [2]. How does one start the development of medications, never mind begin treatment, when one can’t even tell what type of cancer a patient has?

This is the problem that researchers at the University of Cape Town set out to solve. Armed with a combination of genetic and protein information from an array of pancreatic cancer cells the researchers successfully used several artificial intelligence strategies to group the cancer cells into clusters. The cancer cells that clustered together were more similar to each other than they were to the other types of cells. Figure 1, below, shows this in action. Each colour indicates a subtype of pancreatic cancer. More importantly, these groups aren’t just based on the cell’s genetics, but rather a combination of all the information that makes the cancer cell what it is. This makes this classification very compelling [2].

The researchers found that there were two subtypes of pancreatic cancer and that, based on patient records, one of them had worse outcomes than the other. They also found that this worse type of cancer had a mutation that the other subtype did not. This opens up a target for drug discovery and hopefully, one day, a cure for this type of cancer.

As we generate more and more data around diseases, it becomes harder and harder to use traditional methods to interpret it all. Artificial intelligence and this type of research could be the gateway into a future where diseases of the past are simply that. In the past.


  1. Howlader N, Noone AM, Krapcho M, Miller D, Brest A, Yu M, Ruhl J, Tatalovich Z, Mariotto A, Lewis DR, Chen HS, Feuer EJ, Cronin KA (eds). SEER Cancer Statistics Review, 1975-2017, National Cancer Institute. Bethesda, MD,, based on November 2019 SEER data submission, posted to the SEER web site, April 2020.
  2. Sinkala, M., Mulder, N. & Martin, D. Machine Learning and Network Analyses Reveal Disease. Subtypes of Pancreatic Cancer and their Molecular Characteristics. Sci Rep 10, 1212 (2020). 

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s