Song He casually opened a new program and, based on the Irema Laboratory's data, began reproducing the drug prediction process.
He didn't dare accept someone else's results wholesale. It was safer to verify them himself!
First was the classic DDI prediction—drug-drug interactions—which explored the combined effects produced when patients took multiple drugs within a short period.
Back at the Jinghai University laboratory, Lu Chengen had explained the five principles of traditional Chinese medicine prediction: mutual restraint, mutual inhibition, mutual attraction, mutual assistance, and mutual aversion. These were also the principles behind drug-drug interactions, except that modern DDI prediction had delved into the molecular level.
When Song He had read the relevant papers earlier, he had seen DDI predictions based on graph convolutional neural networks, as well as predictions based on balance theory.
The former was too one-sided, considering only the elements of the adjacency matrix. The latter, meanwhile, was too absolute. In short, it claimed that "a friend's friend is a friend" and "an enemy's friend is an enemy." But in actual drug development, relationships were far more intricate than balance theory suggested.
On the computer screen, the Irema Laboratory had actually abandoned both methods and was clearly using a new approach for DDI prediction!
After studying it for a long time, Song He basically understood the idea... deriving results from a massive database!
The Irema Laboratory had attempted to expand its enormous collection of compounds, sort out their familial relationships, and draw a spectacular tree diagram, with each branch connected to a similar compound.
Then, based on the medicinal properties of compounds that had already been identified, they boldly guessed the effects of unknown compounds.
It was like knowing that the husband was a biologist and the wife was a sprinter, then guessing that their future grandson might... sprint