Susanna Fletcher Greer, Chief Scientific Officer of the V Foundation, shared a post with an article on LinkedIn:
“What if one of the most promising advances in cancer detection works better for some patients than others not because of biology, but because of the data behind it? New work from Dr. Jasmine Zhou shows how we can build blood-based cancer tests that are not just more powerful, but more equitable for every patient.
Tuning Out the Static: A Fairer Way to Detect Cancer Early
What if the tools we are building to detect cancer early work better for some patients than others? Not because of the cancer itself, but because of the data we used to build the tool. A new paper in Genome Biology from the V Foundation grantee Dr. Jasmine Zhou and team UCLA takes on this problem and offers a way forward.
Today, we can look for cancer using a simple blood draw. Tumors shed tiny fragments of DNA into the bloodstream, and researchers are learning how to read those signals to detect cancer earlier than ever before. It’s honestly one of the most exciting areas in cancer research. But here is the challenge: these tools learn by studying large sets of patient data. And those datasets do not always reflect the full diversity of patients. When that happens, the tool gets very good at recognizing patterns from the majority group and less accurate for others.
A simple way to think about it is this: imagine teaching someone to recognize faces, but most of the faces they see during training look similar. They will get very good at recognizing those faces, but struggle when they see someone who looks different. That is what can happen with cancer detection tools. The Zhou team approached the problem in a different way. Instead of trying to fix individual pieces of data, they asked whether they could remove the part of the signal that reflects background differences between groups while keeping the part that reflects cancer.
Their approach is like tuning a radio. There is static and there is music. If you can filter out the static without losing the song, you hear things much more clearly. Using blood-based cancer detection data, they were able to do just that. They removed the “static” linked to demographic differences and kept the “signal” linked to cancer. And when they did, something important happened: the test became more accurate for patients who had been underrepresented in the original data. In other words, it started working better for more people.
And that to me is exactly the goal: not just smarter technology, but fairer technology.
Early detection really only matters if it works for everyone sitting in the clinic, not just the patients who were most represented in the data. What I love about this work is that its not about chasing the next innovation. It’s about making sure the tools we are already building are actually ready for real patients. This is where cancer research is headed. Yes, we are making incredible advances. But we are also asking harder questions about who those advances serve and how we make them better. To me this is work worth investing in.”
Other articles featuring Susanna Fletcher Greer on OncoDaily.