AI Is Making A Drug That Is Being Tested On Humans
From an AI-created plan that took just 46 days, a new pharmaceutical drug is now in clinical trial stages.
AI is being used by just about every major industry, including the pharmaceutical drug industry. According to Singularity Hub, a biotech company called Insilico Medicine is starting Phase 2 clinical trials on a drug discovered and designed by AI. The AI designed a molecule that targets a protein in fibrosis, and the process of designing it took only 46 days.
This AI drug was just a proof of concept four years ago, but the company was able to train the model on existing drugs for the protein, allowing them to expedite the process to get to the human clinical trials stage.
While there are many dark sides to AI, this could be one of the more positive uses of the technology. While discovering and designing these pharmaceuticals is typically a time and resource-intensive process, AI could help to make the drug discovery process much faster and cheaper overall.
As Singularity Hub points out, an average of one molecule out of every one million screened ever makes it to the clinical trials and approval phases of production. This process can usually take anywhere from 12 to 15 years and $1 billion in estimated costs. The 46 days of discovery for the AI and the four-year turnaround to clinical trials for a drug are highly efficient and cost-effective in comparison.
Plus, AI technology is only getting better and more accessible, and as it takes on more training data it can possibly discover and design effective drugs even faster. Insilico’s idiopathic pulmonary fibrosis drug has already cut down on time by a third and cost by a tenth.
It’s easy to imagine that the time and costs could be cut down even further, hopefully encouraging pharma companies to price their drugs lower.
Insilico used two different forms of AI to design the drug — a generative adversarial network (GAN) and reinforcement learning.
A generative adversarial network utilizes two neural networks that work together to generate new objects such as text, images, or molecule chemical structures. One network will generate an output, and the other network determines if the output is true or false.
The reinforcement learning AI is a machine learning algorithm that learns through trial and error from its own feedback. With these AI algorithms at work, Inscilio was able to develop the drug INS018_055. INS018_055 works by slowing the thickening and scarring of tissues in the lungs.
For its Phase 2 human trials, Insilico has brought on 60 patients from China and the US with idiopathic pulmonary fibrosis who will take the drug for 12 weeks.
With Phase 2 clinical trials, Insilico will be testing the safety and efficacy of the AI-designed drug. If the Phase 2 trials go well and show promise, the drug moves on to Phase 3 trials which looks at side effects and the efficacy of the drug across a larger sample size of subjects.
It’ll be interesting to see how well this drug does during these trial phases, as its success could certainly mean some big things for the future of AI in the pharmaceutical industry.
