Maureen O'Connor of the National Research Council co-authored a recent study on predicting cancer recurrence. Her team examined public data on more than 1,000 patients' breast cancer histories. They knew which genes were being expressed abnormally and which patients got cancer again years later.
“The problem up until now is a statistical problem,” says O'Connor. “It becomes a statistical nightmare to go and find the most important genes in all of that data.”
O'Connor and her team created a computer program to filter the data and find the most telling genes. They made an algorithm — a set of computations — to predict for any new patient whether her cancer is likely to recur after surgery by looking at fewer than 50 relevant genes.
“Not all women need to get chemotherapy,” says O'Connor. “This could be a guide for those who don't really need the full treatment.”
O'Connor's team tested the algorithm with hundreds of patient histories and found that when the algorithm recommends against treatment, it's correct between 87 and 100 per cent of the time, depending on the sub-type of cancer.
This is the only responsible way to go, isn't it? Test the tumor. Don't trial and error the chemo.