Algorithm spots cancer drug sensitivity
Researchers have developed an algorithm that could identify new groups of cancer patients who would benefit from existing drugs.
A type of drug known as PARP inhibitors are used to treat some cancers, but the current genetic tests used by clinics to identify who to give them to don’t catch everybody who would respond to them.
Rather than require new genetic tests, the algorithm, developed at Harvard Medical School, can use the data from the existing tests to find the patients who would otherwise be missed.
“You can’t give the drugs to everybody. You need a way of identifying which sub-group would benefit,” explains Dr Peter Park, lead author of the study. “The wonderful thing about this algorithm is that the testing is already being done, and all we need to do is just add this analysis component to the data that are already being generated.”
PARP inhibitors are effective against cancers that have faults in a DNA repair pathway called homologous recombination, or HR. The PARP protein acts in an alternate DNA repair pathway, and so if PARP is blocked and HR isn’t working, there isn’t a way for cells to fix DNA damage.
HR-deficient tumours that are given PARP inhibitors therefore collect DNA damage quicker than healthy cells that have working HR, meaning they also die quicker.
A pair of genes called the BRCA genes are important parts of the HR pathway and BRCA mutations are often the cause of faulty HR. Patients are tested for BRCA mutations to see if PARP inhibitors are suitable, particularly if they have breast or ovarian cancer.
However, there are more parts to the HR pathway than just BRCA. Patients with mutations in these other genes could also benefit from PARP inhibitors, but as many of these genes aren’t yet known, they can’t be included in tests.
Even without PARP inhibitors, when the HR repair pathway isn’t working DNA damage can’t be completely fixed. This is how the algorithm can spot tumours with faulty HR that might respond well to PARP inhibitors, without needing to know the specific genes involved.
“Even though you don’t know all the players, these players leave common signatures in the DNA, they tend to generate certain types of mutations,” says Dr Park. “So we still wouldn’t know exactly what generated the signature, but we know that it’s some player in that same pathway.”
The Harvard researchers trained their algorithm to look for the telltale signature of disrupted HR in the genetic data already being collected by clinics.
The algorithm was originally developed using full genome sequencing of all 20,000 genes, but the team adapted it to pick out HR-deficient tumours using just the sets of 200-400 genes that clinics routinely test in cancers.
Now that they know how to apply the algorithm to a smaller set of data, Dr Park says they are considering what other pathways they could look at.
“We think that a similar kind of approach could be used to identify cases that may respond better to immunotherapy,” which is a developing form of cancer treatment that helps the body’s own immune system to attack cancerous cells.
They hope that the current algorithm can be routinely applied to every patient, whatever cancer type they have, to find as many patients as possible that might benefit from taking PARP inhibitors.