Researchers from Cambridge University, in cooperation with the Chatgpt-4 advanced artificial intelligence model, have reached unconventional drug compositions that may be effective in eliminating cancer cells, using cheap and safe drugs that are usually used to treat high cholesterol or alcohol addiction.
In an unprecedented experience, artificial intelligence succeeded in proposing a mixture of drugs that showed the effectiveness of some traditional breast cancer treatments, which opens the door to a new pattern of cooperation between human and the machine in developing cancer treatments, through what is known as “re -employment of medicines.”
The researchers said that this result did not coincide with coincidence, but rather by using a huge Chatgpt-4 category that was trained to analyze huge quantities of scientific literature, in search of hidden patterns that may indicate unveiled therapeutic capabilities.
The researchers pointed out that the fundamental idea behind this work was simple in principle, but deep in its potential impact. Can a linguistic model such as Chatgpt-4 can contribute to the discovery of new drugs to treat cancer, by reading and understanding millions of pages of medical research that is difficult for any human beings to be complete?
The Cambridge team answered this question through a unique experience that combined artificial intelligence and experimental scientific research.
Doctor Catherine Getier is no longer analyzing biopsies under a microscope, but rather is based on a screen, before the role of artificial intelligence comes in helping to diagnose some types of cancer.
In the first stage of the experiment, scientists asked Chatgpt-4 to suggest new pharmaceutical groups that can be effective in targeting a specific line of breast cancer cells, a line that is widely used in medical research.
Artificial intelligence directing
But the instructions were very accurate, as scientists requested to avoid any medicine used in the treatment of cancer, and focus only on medications that could kill cancer cells without harm to healthy cells, while prioritizing the pre -approved medicines by the regulators, and be at hand in terms of price.
The model responded to these directives and suggested 12 drugs, and after testing these installations in the laboratory, it was found that three of them were more effective than standard medications currently used in the treatment of breast cancer.
It did not stop there, but the results of these tests were used to feed the model again, in what is known as the closed loop system, a system in which artificial intelligence learns from the results of the experiment, and proposes new options based on these data. Chatgpt-4 suggested four additional combinations, three of which showed promising results as well.
This repeated approach, where artificial intelligence is nurtured by ideas, then trial, then the results are returned to the model to suggest more, represents an unprecedented penetration in the relationship between humans and the machine in scientific research.
For the first time, as stated in the paper published in the Journal of the Royal Society, a large linguistic model is used in this interactive cooperative way in the field of cancer research.
One of the interesting aspects is that what is usually a “mistake” in the performance of artificial intelligence models, known as “hallucinations”, turned in this context into an advantage.
Chatgpt-4, like other linguistic models, is sometimes known for its production of inaccurate or even fabricated information.
But in the field of scientific research, especially in the field of drug exploration, “hallucinations” that propose unfamiliar ties may lead to ideas worthy of experimenting.
New methods of scientific research
This type of artificial intelligence provides a wide -ranging imaginary layer of scientific exploration, and opens for human scientists new ways that may not have occurred to them before, said the lead author of the study, Ross King, a researcher of the Department of Chemical Engineering and Biological Technologies at the University of Cambridge.
Artificial intelligence was not limited to submitting random suggestions, but rather relied on a profound analysis of existing scientific literature, and linking well -known biological information with similarities or non -visible patterns of human eye.
The researchers pointed out that this ability to move between the huge amount of knowledge, and the conclusion of indirect relationships that made the model closer to a “virtual world” that does not tire, explores the depths of knowledge at the pace of human superiority.
Among the pharmaceutical formulas that have emerged in laboratory experiments, a sympastatin mix, a drug that is usually used to lower cholesterol and deculfire, a drug used to treat alcohol addiction, was one of the most prominent options that showed effectiveness against breast cancer cells.
A study warned that artificial intelligence models may recommend various treatments for the same medical condition, based only on the social, economic and population features of the patient.
The researchers pointed out that in this proposal “the fundamental idea lies in the research”, which is the re -employment of already existing drugs in completely new uses, an idea that is scientifically known as the redirect of medicines, which is a quick and relatively low -cost approach than developing completely new drugs.
Although these results are promising, scientists stress that the road is still long, before these formulations are used to actually treat patients.
Every medicine or drug composition shows effectiveness in the laboratory, later needs to go through multiple stages of complex clinical trials, to ensure their safety and effectiveness in human bodies.
While the study author, Hector Zenil, a researcher at Kings in London, described the role of Chargpt-4 as a partner who does not know fatigue capable of exploring a huge satisfactory space of scientific possibilities, and suggesting ideas that would have taken a long time through human effort alone.
“What we do is not a replacement of scientists, but rather a founding of a new pattern of cooperation.”
Concepts, this study is an proof that artificial intelligence is no longer just a tool or metaphor for the human mind, but it can play a real role within the scientific research system, provided that it is under conscious human supervision.
Artificial intelligence in this case did not work in a vacuum, but rather was part of a repetitive link of analysis, experimentation, feedback, and modification, leading to continuous improvement of suggestions.
King talked about the fact that the artificial intelligence subject to supervision is able to propose hypotheses that exceed specializations, merge the previous results, and work through multiple repetitions. “This represents new limits for scientific research, as the term that was previously considered a metaphorical metaphor (the automatic world) has now become a reality backed by evidence.”
