UK markets closed
  • FTSE 100

    7,500.89
    +34.98 (+0.47%)
     
  • FTSE 250

    20,338.96
    +93.53 (+0.46%)
     
  • AIM

    933.20
    +4.84 (+0.52%)
     
  • GBP/EUR

    1.1820
    +0.0002 (+0.02%)
     
  • GBP/USD

    1.2139
    -0.0064 (-0.52%)
     
  • BTC-GBP

    20,113.82
    +77.29 (+0.39%)
     
  • CMC Crypto 200

    574.64
    +3.36 (+0.59%)
     
  • S&P 500

    4,280.15
    +72.88 (+1.73%)
     
  • DOW

    33,761.05
    +424.38 (+1.27%)
     
  • CRUDE OIL

    91.88
    -2.46 (-2.61%)
     
  • GOLD FUTURES

    1,818.90
    +11.70 (+0.65%)
     
  • NIKKEI 225

    28,546.98
    +727.65 (+2.62%)
     
  • HANG SENG

    20,175.62
    +93.19 (+0.46%)
     
  • DAX

    13,795.85
    +101.34 (+0.74%)
     
  • CAC 40

    6,553.86
    +9.19 (+0.14%)
     

AI model discovers drugs 1,000-times faster than current methods

·2-min read
A team from MIT built an AI model called EquiBind that is 1,200 times faster than existing models at finding potential drug molecules (Getty Images/iStockphoto)
A team from MIT built an AI model called EquiBind that is 1,200 times faster than existing models at finding potential drug molecules (Getty Images/iStockphoto)

Researchers have developed an artificial intelligence model that can find potential drug molecules more than 1,000 times faster than current state-of-the-art methods.

The team from Massachusetts Institute of Technology (MIT) say the AI model, called EquiBind, will significantly reduce the chances and costs of drug trial failures.

The number of molecules that have potential drug-like traits is gargantuan, estimated to be around 1060. By comparison, the Milky Way galaxy has around 108 stars.

The EquiBind model is able to successfully bind these drug-like molecules to proteins at a rate that is 1,200 times faster than one of the quickest existing computational molecular docking models.

EquiBind achieves this through built-in geometric reasoning that allows it to predict which proteins will fit to a molecule without any prior knowledge of its target pocket.

“We were amazed that while other methods got it completely wrong or only got one correct, EquiBind was able to put it into the correct pocket, so we were very happy to see the results for this,” said Hannes Stärk, a first-year graduate student at the MIT Department of Electrical Engineering and Computer Science and lead author of the paper describing the research.

The findings have already attracted the attention of industry figures, with hopes that it can be used to find treatments for lung cancer, leukemia and gastrointestinal tumours.

“EquiBind provides a unique solution to the docking problem that incorporates both pose prediction and binding site identification,” said Pat Walters, the chief data officer for drug discovery firm Relay Therapeutics.

“This approach, which leverages information from thousands of publicly available crystal structures, has the potential to impact the field in new ways.”

The paper, titled ‘EquiBind: Geometric Deep Learning for Drug Binding Structure Prediction’, will be presented at the International Conference on Machine Learning (ICML).

Our goal is to create a safe and engaging place for users to connect over interests and passions. In order to improve our community experience, we are temporarily suspending article commenting