Keeping in mind the enormous cost and time investment required to develop a new drug and make it available on the market, an American startup scaled up its AI to make a drug and see what happened. The whole thing was just a test, but the machines not only proved to have enormous potential to shorten the current process; And cut costs dramatically – but they were able to produce a new molecule with medicinal properties.
Artificial intelligence in the service of medicine
According to Gregory Barber of Wired, according to some estimates. The process between developing and marketing new drugs can be incredibly slow and require the investment of billions of dollars, depending on the drug. But, as many experts bet, AI could step in and dramatically reduce research time and hence the cost of creating new drugs.
In the study, an artificial intestine, called intestinal organoid. Allowed researchers to cultivate human biopsy specimens while maintaining the basic physiology that exists within a human being. They also used cutting-edge genomic techniques to track the abundance of different molecules in the colon tissue of more than 150 patients.
MicroRNAs work as negative dials; The more abundant this molecule is, the more a target gene will be deleted.
Patients with subtype 1, unlike subtype 2, often do not respond well to medications and develop stenoses – extreme narrowing of the digestive tract, requiring surgery after it develops. Markers such as miR-31 may be helpful in the future for doctors to predict whether a patient should undergo preventive surgery before the condition gets worse.
Using miR-31, doctors could potentially separate individuals with Crohn’s disease into subtypes to more accurately determine whether a particular drug works for one subtype and not the other.
“Our study suggests that miR-31 is not only a predictor of clinical outcomes, but may be functionally relevant in conducting the disease,” says Praveen Sethupathy, senior co-author of the study.
Future work will explore exactly what miR-31 does and what role it can play in the integrity of the intestinal epithelium. “Our long-term goal is to better understand which molecular level is why the disease is so different in each patient, and to use that knowledge to develop more effective therapies,” says Terrence Furey, another study researcher.
This computational method is already widely used by the pharmaceutical industry to identify viable drug candidates, and this system only works with known molecules. Already ReLeaSE is able to create and evaluate new molecules focused on medicine. Olexandr Isayev, co-creator of the project, explains;
A scientist doing virtual sorting is like a customer placing orders in a restaurant. What can be ordered is usually limited by the menu. We want to give scientists a grocery store and a personal chef who can create any dish they want
The intention of the researchers is not to eliminate human labor. But to offer a new system that helps advance the pharmaceutical industry. In one example, ReLeaSE was able to find ways to find compounds that inhibit the main enzyme responsible for leukemia.
The algorithm’s ability to design new, and therefore patentable. Chemical compounds with specific biological activities and optimal safety profiles must be highly attractive to an industry that is constantly looking for new approaches to reducing the time required to bring in a new drug candidate. clinical trials.
In fact, there are countless companies developing molecules and working on the creation of new drugs. But it is essential to observe the cellular response and test the substances designed in the laboratory to prove the efficacy and safety of a possible treatment. Insilico Medicine began by training its AI to think like a medical chemistry expert would think and put the results to the test to see how molecules would behave in the real world.
Still, it is important to clarify that Insilico Medicine’s AI is not yet that “smart”; As well as that of other companies working on similar projects; Since, as the work has been presented now, the molecule would have to pass by tweaking and altering to advance to the clinical trials phase. Which would take time, which is exactly what we are looking to save. In addition, the team focused on an area (tissue healing, remember?). That already has a lot of information to feed AI, and the challenge would be different if it was a rare disease, for example.