Scientists have used artificial intelligence to “predict” the formulation of newly designed drugs, with the stated goal of helping improve their regulation. Artificial intelligence has generated formulas for nearly 9 million potential new drugs.
Researchers at the University of British Columbia (UBC) used a deep neural network in this work to teach it to construct the chemical structure of potential new drugs.According to their LearnThe computer intelligence released this week performed better on this task than scientists expected.
The research team used a database of known design drugs (synthetic psychoactive substances) to train the AI to train its structure.The designer drug market is constantly changing as their manufacturers constantly adjust their formulations to circumvent restrictions and produce new ones “legal” The researchers said that although it would take several months for law enforcement agencies to crack their structure.
“The vast majority of these designed drugs have never been tested in humans and are completely unregulated. They are a major public health problem in emergency departments around the world,” Said Dr. Michael Skinnider, one of the researchers and a UBC medical student.
After training, artificial intelligence can generate approximately 8.9 million potential design drugs. Subsequently, the researchers analyzed the data sheets of about 196 new drugs that appeared in real life after the model was trained, and found that more than 90% of them had been predicted by the computer.
“In fact, we can predict which special drugs may appear on the market before they actually appear. This is a bit like the 2002 sci-fi movie “Minority Report”, in which anticipation of upcoming criminal activities helps to significantly reduce criminal activities. .future world” Said Dr. David Wishart, senior author and professor of computer science at the University of Alberta.
The research team pointed out that identifying completely unknown substances is still a problem for artificial intelligence, but they hope that it may help accomplish this task, because the computer can also predict which formulas for design drugs are more likely to be created and hit the market.The model “72% of the time, the correct chemical structure of an unidentified design drug was ranked in the top 10 drug candidates,” This is an easy-to-obtain measurement method when performing spectral analysis, increasing the accuracy to around 86%.
“We were shocked that the model performed so well, because it is generally considered an unsolvable problem to elucidate the entire chemical structure only through accurate mass measurements.” Skinnyder said.
Do you think your friends will be interested? Share this story!