We all carry over 9,000 genetic mutations. Most are benign, but what happens if an AI predictive algorithm tells us we carry a risky disease-causing mutation?
We typically carry around 9,000 mutations in our genomes, making us all genetic mutants to some degree. We inherited most of these mutations from our ancestors, but accumulate around 64 mutations during our own development, passing them on to our offspring for good measure.
Most of these mutations are benign and have little to no effect on our health, but some can severely disrupt protein function, and can cause diseases such as cancer, diabetes or heart conditions. Scientists are now capable of predicting the risk of developing certain diseases from genetic screens. AI algorithms are the main tools used for these predictions.
For Andrea Downing, these predictive algorithms may have saved her life. Downing tested positive for a mutation in a gene called BRCA1 when she was in her 20s. Mutations in BRCA1 and BRCA2 are known to increase the risk of cancer: between 45 and 72% of women with mutations in their genes are diagnosed with breast or ovarian cancers.
“The women in my family have a history of cancer,” Downing told DW. “My mother was diagnosed with stage three breast cancer when I was very young and had recurrences that shaped my childhood.” Downing took the decision to have a mastectomy in order to eliminate the risk of breast cancer in the future. Angelina Jolie made the same decision in 2013.
“I firmly believe that my decisions saved my life. I wouldn’t want to know what my path would have been if I hadn’t known about the mutation,” Downing said.
Predictive medicine and AI
Downing’s experience was part of an early wave of an approach called predictive medicine. The idea of the approach is to predict the likelihood of getting a disease, how the disease will progress, and which treatments people are likely to respond to.
AI-based algorithms are at its heart. They are used to analyze DNA sequences, pinpoint mutations, and predict disease risks of those mutations by comparing against reference data. Christopher Mason, a professor of genomics and biophysics at Cornell University in the US, is at the forefront of research in the field. He believes it will completely transform our approach to medicine and healthcare.
“Cancer treatment is perhaps the best example. We can use tools to know what cancer can be targeted by the best therapy. In a matter of hours or days, we can have DNA sequence information mapping the best treatment for a patient based on the data,” he said.
New healthcare AI tools seemingly appear every month. There are tools to detect strokes in patients, heart diseases, analyze medical images, manage health data, and map the likelihood of treatment success based on your genetic code, along with many others.
Mason was keen to add that AI shouldn’t replace physicians but could help to be a co-pilot and data resource when diagnosing and treating patients.
Difficult decisions
Along with genetic screens in clinical settings, more and more people are taking genetic tests from companies such as 23andMe and Ancestry.com. These tests often come with many predictions about what variances and mutations in your genome might mean for your health.
They might identify risk factors for conditions like diabetes, dementia, heart conditions, even if you don’t have any symptoms at the time. It’s this aspect of knowing your future which Downing said she wouldn’t wish on anyone.
“There’s a lot about living with knowledge of your future [after a positive BRCA1 result] which is traumatic, even though I never had cancer,” she said. It’s a drastic step to undergo a mastectomy to eliminate the risk of cancer. In her mind it was worth the risk, but the decision might not be so clear for other people with different disease risks.
In conditions such as Alzheimer’s disease or schizophrenia, people cannot make decisions like Downing did to remove body parts.
“If you found out you had a genetic risk factor that says you have a 95% chance of early onset Alzheimer’s disease, and by 45 you won’t remember much, would you just say well, I’m not going to care about my future and live a crazy life? What if a new therapy in 10 years can reverse the disease?” Mason said.
Data and privacy concerns in the US
Using AI-based algorithms in healthcare also raises other issues. “They can help to eliminate diseases better, but they can be scary. They completely change the risks around data privacy,” Mason told DW.
Downing now co-runs a patient advocacy organization called Light Collective. In 2022, she published a paper exposing Facebook’s sharing of people’s health data for marketing purposes without patient consent.
“We’re seeing that information used from a marketing perspective to target content to those people. For people in my community, it’s often chemotherapy, but also snake oil and other harmful medical misinformation,” she said.
The impact goes far beyond targeted ads. Downing said data used by predictive algorithms were now also affecting people’s access to healthcare in the US.
The Guardian highlighted a case in 2021 where an AI algorithm used by US healthcare providers resulted in a severely disabled woman with cerebral palsy having her access to care and benefits severely reduced.
“My big theme now is to increase the transparency of how health systems are using predictive algorithms in care. It’s not clear how decisions based off those algorithms are validated,” Downing said.
Predictive medicine is in its infancy. How people live with health predictions, and how health systems use those predictions to impact people’s healthcare, is only just beginning to be explored.
Mason and Downing both said there was a long way to reducing the new risks they pose. “My big hope is we have the best possible science at our fingertips to make sure we’re not being taken advantage of whilst going through being a patients and suffering that path,” Downing said.