Using artificial intelligence (AI) and one of the fastest supercomputers in the world, Chinese scientists are developing otherwise unknown chemicals that may be used clinically in the future.

The Tianhe-2 supercomputer in south China’s Guangdong Province, ranked among the world’s 10 fastest computers in the TOP 500 released this month, has been used as a platform for drug discovery. Now AI-based algorithms make the machine even smarter.

Scientists from Sun Yat-sen University and Beijing-based AI startup Galixir, along with those from Georgia Institute of Technology and Massachusetts Institute of Technology, presented a practical deep learning toolkit for predict biosynthetic pathways of natural products (NPs) or NP-like compounds in Tianhe-2.

Natural products are the main source of clinical drug discovery. More than 60% of small molecule drugs approved by the FDA in the United States are NPs or their derivatives.

Over 300,000 NPs have been registered to date, but due to complex production know-how, only a tenth have been developed as a substrate or product, with computer-assisted screening urgently needed.

In a recent study published in Nature Communications, researchers presented a tool called BioNavi-NP to optimally propose NP biosynthetic pathways from simple building blocks, which do not require any already known biochemical rules.

First, a one-step bio-retrosynthesis prediction model is trained to generate candidate precursors for a target NP. According to the study, the fully data-driven model achieves 1.7 times more accurate prediction accuracy than the previous rule-based model.

Next, an automatic retro-biosynthetic route planning system efficiently samples plausible biosynthetic pathways.

The study reveals that the toolkit can successfully identify biosynthetic pathways for 90.2% of the 368 compounds tested.

Additionally, researchers combined an existing enzyme prediction tool to provide a user-friendly, publicly available web server capable of predicting biosynthetic pathways. It can also assess the biological feasibility of these pathways based on estimated species and enzyme preference.

By entering all the relevant NP molecules into the online toolbox, one can get several predicted ways to synthesize them in minutes.

Rapid results are only made possible by Tianhe-2’s strong parallel computing capability and custom GPU resources, which can reduce training and testing time from over two weeks to one day.

China’s Tianhe-2 supercomputer has been widely used to promote health and medical research.