The various challenges associated with AI-driven genetic testing

The Human Genome Project, launched in 1990 and concluded in 2003, enabled the documentation and analysis of the building blocks of human DNA, revealing patterns that could indicate the likelihood of developing rare diseases or carrying genetic abnormalities. Initially, predictive tests were limited to exceptional cases due to their high cost. For instance, expecting mothers could undergo tests to detect fetal cells with chromosomal defects potentially associated with Down’s Syndrome or cancer patients could have their tumours examined to determine the most effective treatment options.

How has AI changed genomics?

As genetic information can be processed faster with Artificial Intelligence (AI), the amount of data going under the scanner has also increased.

In March 2024, John Hopkins University researchers revealed how genetic code sequences, which were previously ignored, were found to be associated with tumours. Using machine learning techniques, they were able to analyse more than 1,200 distinct types of “junk DNA” elements from tumours to deepen research around new therapies, new diagnostics and new screening methods for cancer.

Bootstrapped healthcare startup, Gene Box, leverages AI primarily to process vast amounts of genetic data swiftly and accurately, uncovering patterns and insights that would be nearly impossible to identify manually.

Their AI-driven algorithm helps predict genetic predispositions, interpret complex gene-environment interactions, and provide personalised recommendations based on individual genetic profiles. Moreover, their systems are designed to incorporate inputs from published, highly rated scientific evidence.

“By continuously updating our models with the latest research findings, we ensure that our analyses are not only accurate but also grounded in the most current and reliable scientific knowledge,” said Mr. Pranav Anam, founder, and CEO of the healthcare startup.

What are the potential pitfalls?

When it comes to predicting whether a child will get through an elite school or how likely is she in landing a high-paying job, these testing tools can hardly offer a straightforward answer. “These aren’t as simple as a ‘yes’ or ‘no’ question. It is a well-known fact that genetics is only 30% of the contributing factor,” Mr. Anam said.

There are other caveats. For instance, a diagnosis could change in the future or may not present the full picture, often falling under the grey zone of significance. “These are known as ‘variations of unknown significance.’ Sometimes, other members of the family could be required to be tested to draw conclusive results,” said Ms. Gayatri Iyer, a scientist at the Tata Institute for Genetics and Society.

For example, research has linked about 80 genes to Alzheimer’s. While genetic testing can identify the risky genes, that doesn’t mean the patient is bound to get it. The tests simply assess the genetic risk by identifying certain genetic variations that are associated with Alzheimer’s. Consequently, a patient can also develop Alzheimer’s without having any of the associated genes. “There are a bunch of questions we need to deal with. What if the tests found something that the patient wasn’t looking for at all? How do you handle predictions of mental health diseases,” Ms. Iyer noted.

“When dealing with sensitive cases, such as determining genetic predispositions for conditions like depression or heart disease, we adhere to strict ethical guidelines,” Mr. Anam said. “Our goal is to provide insights that can guide proactive health measures, not to make definitive diagnoses.”

Apart from genes, factors like environment, diet and schooling play equally important roles in a child’s development.

Why should users remain cautious?

Nucleus, a genetic testing startup founded in 2020 by 23-year-old college dropout Kian Sadeghi has promised to analyse cancer, depression, longevity, or how extroverted a person might be, among many other diseases and traits, better than any other firm of its kind. The startup received $14 million in funding from some of the most eminent names in Silicon Valley such as VC firms led by PayPal co-founder Peter Thiel and Reddit co-founder Alexis Ohanian.

In a promotional video posted on X, Sadeghi said its only “an inevitability that every single person has their entire human genome on their iPhone.” If that sounds like a data security risk waiting to happen, that’s because there’s a similar case that is fresh in public memory.

The once popular consumer genetic testing company, 23andMe was at the receiving end of a data security breach in October 2023, after hackers gained access to personal data including family trees, birth years and geographic locations, of 6.9 million users and sold it on the dark web. The company was directed to pay a $30 million fine in September and provide three years of security monitoring to settle the lawsuit. It’s been a dramatic fall for a company that was considered as pioneering, and valued at around $6 billion in 2021.

In the following months, most of the company’s board resigned en masse leaving co-founder Anne Wojcicki as the sole member who has been contemplating delisting the company from Nasdaq. In November, it fired 40% of its workforce.

As the company implodes, the looming question around, “What happens to the DNA data of its 15 million customers?” becomes a pushing worry. Customers have scrambled to delete or withdraw their genetic data before the firm goes bust or ownership changes hands, but there are a couple of caveats. If users opted in to have their genetic data analysed, the information can’t be withdrawn (A company spokesperson said about 80% had opted in).

Additionally, companies like 23andMe don’t fall under Health Insurance Portability and Accountability Act (HIPAA) which has federal laws protecting consumer data. Although Ms. Wojcicki has promised users that the firm will stand guard over their data, there’s little that inspires confidence right now.

But since the AI cat is out of the bag, the fledgling sector of AI-infused genomics has been drawing more interest. “Investors recognise the long-term value and impact of these technologies, particularly as the demand for personalised healthcare solutions rises,” Mr. Anam said.