This audio is automatically generated. feedback.
AI can reduce the cost of drug development, allowing small and medium-sized enterprises to do more with less. One 2023 study found that AI-driven R&D efforts from discovery to preclinical could: Save time and money At least 25% to 50%.
Panna Sharma, CEO of Lantern Pharma, a biotech company that is using its own AI platform to develop oncology drugs, said smaller biotech companies developing drugs can benefit from AI’s computing power and efficiency.
The company currently has three lead candidates in early-stage clinical trials and an antibody-drug conjugate program in preclinical stages.The company is also partnering with other biotech companies that want to use its AI technology platform, Response Algorithm for Drug Positioning and Rescue (RADR), which aims to predict a patient’s potential response to a drug.
Panna Sharma, CEO, Lanthanum Pharma
Used with permission from Lantern
“If you look at the new chipsets that are being developed, calculations that[three years ago]would have cost maybe $100,000 and consumed a month or two of machine time, can now be done in a day for less than $1,000 to $2,000,” Sharma says. “This completely changes the ability of individual researchers and small companies to become meaningful developers. And this is going to continue to change. It’s a one-way curve.”
As more pharmaceutical and biotech companies leverage AI platforms to develop medicines, some risks still loom large.
Future challenges for AI
As many have pointed out before, AI and machine learning models are trained on available datasets, which can leave gaps in patient demographics.
“There’s a risk that the data set will be incomplete, and that’s something we’re always concerned about,” Sharma said. “An incomplete data set will lead to incorrect or incomplete conclusions, and I think that’s one of the biggest challenges.”
Incomplete or biased data sets are under the FDA’s jurisdiction. expressed concern Last year, there was debate regarding the ethical considerations and generalizability of findings extrapolated outside of a testing environment due to incomplete data.
“The new generation of drug developers doesn’t fully understand the biological complexities of disease. You might end up with a lot of useless molecules at an early stage.”
Panna Sharma
CEO of Lantern Pharma
The data also needs to reflect the patient population, which can be difficult if the dataset reflects a smaller group of specific patients, Sharma said. Companies may also be training AI algorithms based on available data, but that data may not reflect the actual patient population.
“You have to ask yourself, is it their biology in the real world that I’m really trying to look at and target,” Sharma said.
The complexity of the disease
Another challenge in computational biology, Sharma said, is the complexity of diseases: AI models aren’t always trained for the characteristics of diseases like cancer or neuroscience, which can vary widely from patient to patient.
“One of the biggest challenges I see in talking to other AI companies is that the new generation of drug developers doesn’t fully understand the biological complexities of disease,” he says. “Let’s be honest, there’s a lot of crappy early-stage molecules out there.”
He said that relying too heavily on AI can lead to missing the forest for the trees.
“AI has tended to focus too much on software and data and not enough on the complexities of disease and biology,” Sharma said, “and that’s going to hold back a lot of people in the field of AI drug development.”
As more medicines are developed using AI technology platforms, patient skepticism could also pose a challenge and risk for companies seeking to enter clinical trials.
“There may come a time when patients will question whether these compounds are made by AI and may feel uneasy about taking drugs developed by AI,” Sharma said.
Given these risks, biotech companies may need to assess how patients perceive AI medicines when participating in clinical trials.
Meanwhile, the FDA has yet to decide how to regulate AI in drug development and is expected to release guidance on the topic. At the same time, as the adoption of AI in industry continues to grow, the FDA over 100 Drug and biological product applications submitted in 2021 with an AI or machine learning component.
“Given the speed at which AI and software are developing, in some cases we may be late in learning about the need for regulation, especially when it comes to medical devices that may use AI,” Sharma said.
But once developed, there may be less need for regulation, Sharma said, because the clinical trials and drug approval process will still remain the gold standard for quality, despite AI being involved.
