Pictured: Artificial Intelligence, Digital Technology Concept/iStock, Phongsak Sangkhamanee
Drug discovery may undergo an astonishing transformation in the coming decades thanks to artificial intelligence. But experts say it is now important to deal with potential ethical and social justice landmines that may be encountered on the road into the new era.
“One of the keys to AI-based drug development is to dramatically improve our ability to find potential targets,” says computer scientist Suchi Saria, director of the Machine Learning and Healthcare Laboratory at Johns Hopkins University. But she told us, This brave new world will be ‘full of moral questions’ biospace. “The cost of designing a drug and bringing it to market is very high, so our trials tend to be homogeneous. We select simple, narrow patient groups. And the system needs to be more open to reflect real-world populations.”
talk with biospaceExperts raised questions, including: How can data be truly anonymized to protect patient privacy? How strict is the current consent system?How scientists and clinicians can develop robust patient groups that don’t exclude members of underrepresented groups Or a vulnerable group in society?
First, the good news: AI has the potential to help researchers fine-tune who is likely to respond well to specific drugs, thereby improving fairness and equity. it will be more likely “People who may have a reaction to the drug will actually be covered,” said Kim Branson, global head of artificial intelligence and machine learning at GSK.
Branson told us that there are often many options for treating a disease, and insurance companies may not be willing to pay for the most expensive drugs without evidence that they are best for patients. biospace. Understanding which patients will respond to a certain drug would benefit socioeconomically disadvantaged groups — 49% of whom said their insurance company refused to cover at least one drug in a given year, according to a 2020 report. NPR. This is also critical for pharmaceutical companies themselves. “In some cases, we simply can’t produce enough drugs for everyone,” Branson said.
I cited a recent Phase II trial of GSK’s bepirovirsen in hepatitis B, in which about 10% of patients responded well and experienced functional cure (viral markers below the lower detection limit). The study relied on millions of data points, including clinical data, lab tests and viral genotyping, and the company plans to continue sifting through the data through machine learning algorithms to better understand who might benefit. “There is strong evidence that drug targets that are genetically validated and responsive to biomarkers are more likely to succeed, and artificial intelligence plays a key role in the discovery of complex biomarkers,” Branson said.
Another benefit could be cutting the staggering costs of drug discovery, making it easier to bring more drugs to market at a lower cost. Currently, the cost of developing a new drug and bringing it to market is approximately $2.3 billion. Every successful drug must bear the cost of many failures. Society as a whole bears this cost, and AI may help alleviate the burden.
“Improving those odds could open the door to more affordable drug pricing,” said Daphne Koller, a computer scientist and former Stanford professor who is CEO and founder of Insitro, a machine learning-powered drug discovery company. , is working with companies such as Gilead and Bristol-Myers Squibb.
Protecting privacy: a key issue
Legal protections such as the Health Insurance Portability and Accountability Act (HIPAA) in the United States and the General Data Protection Regulation (GDPR) in Europe prevent foreign access to sensitive patient data.Even so, medical data breaches are still a huge problem, and according to HIPAA Magazinewhich has been on an upward trend over the past fourteen years.
A 2021 study BMC Medical Ethics pointed out that individuals can be identified in data repositories of public and private institutions even if the data has been deleted and anonymized. In a 2018 study, algorithms were able to re-identify more than 85% of 4,720 adults and nearly 70% of 2,427 children in the cohort, even when protected health information was removed. In 2015, in the UK, Google DeepMind signed an agreement with the Royal Free Hospital in London to copy 1.6 million medical records and feed them into DeepMind AI. Eventually, the deal was struck and Google backed down.
Mihaela van der Schaar, a computer engineer who heads the University of Cambridge lab, writes that data sharing and patient privacy are not necessarily incompatible. Her lab is developing synthetic patient records that simulate real-world records and can be used for machine learning. This novel framework, called “Anonymization by Data Synthesis Using Generative Adversarial Networks,” generates artificial data but preserves the properties of real datasets. To remove any inherent bias in real data, including the historical underrepresentation of marginalized groups, and transfer it to synthetic data, the lab is developing DECAF, a synthetic data generator that eliminates artificial data deviation edge in .
Even if patients agree to share their data, they may not understand all the implications. “Patient consent should be provided to all patients before any type of screening is performed,” Kohler told biospace. “These forms should clearly outline how patients’ data will be used and provide them with the option to opt out of data sharing if they wish.” Kohler said it’s important for health care providers to make sure “patients are signing these consent forms It is important to fully understand what they are agreeing to.”
How to be fair and represent everyone
Abdoul Jalil Djiberou Mahamadou, a postdoctoral researcher in artificial intelligence and biomedical ethics at Stanford University and GlaxoSmithKline, said the FDA study found that about 10% of drugs approved between 2014 and 2019 showed racial/ethnic differences in exposure and /or response differences. White people were overrepresented in clinical trials of drugs approved during this period, while blacks, Asians and other ethnic groups were significantly underrepresented. “If you have biased data, you’re going to get biased results,” he told biospace.
One of the most trusted and popular databases comes from the UK Biobank, which has 500,000 residents voluntarily participating in the database, but Mahamadou said, “That’s a high-income country.” 94% of the bank’s data comes from white people , 2.3% are from Asians and 1.5% are from black or black British people. “If (you) are using artificial intelligence to develop new drugs, you need representative data, but also data from low- and middle-income countries,” he said.
One solution is to collect new data, “but that could take years and be expensive,” Mahamadou said. “You also need to understand that the data collected reflects culture and society.” For example, as recently as 2021, Sudan had no legislation on privacy and ethics, according to the United Nations Conference on Trade and Development. “So you might only want to collect data in countries that have strict licensing,” Mahamadou continued. Or, he said, ethicists could provide ethical data collection principles for countries without legislation. The bottom line, Muhammad said, is that we can’t just rely on data scientists; we have to rely on data scientists. We must bring in clinicians and ethicists who understand the country and culture in which the data are being collected.
Database is deeply aware of these issues and is working to resolve them. At UK Biobank, a project called Our Future of Health is recruiting data as broadly as possible, Kohler said. “They have recruitment centers in urban pharmacies. The All of Us project in the U.S. has done similar efforts to diversify their patient populations.” Some of the data is already diverse, she said, because that’s Histopathological data for patients with solid tumors are part of the standard of care in almost every region of the world. At GlaxoSmithKline, “We’re trying to buy data on a global scale,” Branson said. “We want to make medicines for all of humanity. “
Jill Neimark is a freelance science writer from Macon, Georgia. Contact her at jillneimark.com.