Artificial Intelligence: Ping An’s New Deep Learning Framework Uses Unlabelled Data For Training

June 22, 2021 | Artificial Intelligence, News

AI-driven pharmaceutical research in the discovery of new drugs gets a boost with Ping An’s breakthrough.

Ping An Insurance (Group) Company of China, Ltd. (SSE: 601318) announced Monday that research by Ping An Healthcare Technology Research Institute and Tsinghua University was successful in the development of a deep learning framework for drug discovery that learns molecular representations from large volumes of unlabelled molecules. (CISION PR NEWSWIRE)

“An effective self-supervised framework for learning expressive molecular global representations to drug discovery”

The research by Ping An and Tsinghua uses graph neural networks (GNN), a tool for modeling molecular data. The researchers created MPG, a novel deep learning framework that learns molecular representations from large volumes of unlabelled molecules.

They devised MOIGNet, a powerful GNN that modeled molecular graphs. Together with Pairwise Half-Graph Discrimination, a self-supervised pre-training strategy, the researchers succeeded in getting MOIGNet to capture meaningful patterns of molecules that could generate an interpretable representation, after pre-training on 11 million unlabelled molecules.

The significance of unlabelled data for training datasets

Traditionally, drug discovery has been hampered by the extremely long time frames from invention to market – as high as 10 to 15 years. High failure rates and prohibitive costs accompany the large number of experiments that may ultimately result in success.

Though AI has shown promise in some areas of pharmaceutical research, molecular modeling has suffered from the limited amount of labeled data available for training datasets.

“The experimental results showed that MPG achieved the state-of-the-art performance on multiple drug discovery tasks,” Ping An said in a statement on the new research. “It is an important first step towards graph-level self-supervised learning on large-scale molecule datasets.”

Ping An said this technology was being used by Ping An Shionogi, a joint venture founded by it with Shionogi & Co., Ltd. a Japanese research-driven pharmaceutical company, for research and development of new drugs and drugs repurposing.

Related Story:  A Molecule For Alzheimer’s Created By Exscientia’s AI Enters Clinical Trials

Free Industry News

Subscribe to our free newsletter for updates and news about alternatives investments.

  • This field is for validation purposes and should be left unchanged.


Latest Alternative Investment News
FinTech: Lloyds To Acquire Stake In Loyalty App Bink
January 4, 2022     FinTech, News

Lloyds Banking Group (LON: LLOY), Britain’s biggest mortgage provider, will acquire a minority stake in loyalty app Bink, according to a report by Sky News for an undisclosed amount that…
Digital Assets: Coinbase CEO Armstrong Said To Have Splurged $133M On Home In LA
January 4, 2022     Digital Assets, News, Real Estate

An iconic property in Bel Air, Los Angeles, designed by internationally acclaimed English architect John Pawson changed hands last month for $133 million and the buyer was Coinbase (NASDAQ: COIN)…
Alternative Investments/ESG: VegTech Invest Launches Plant-Based Innovation & Climate ETF

VegTech Invest advisory has launched the VegTech Plant-based Innovation & Climate ETF (Ticker: EATV), its first financial product. The ETF offers exposure to publicly traded companies actively innovating with plants…超·世界.png
Venture Capital: Chinese AI Startup Parametrix Raises $100M, Turns Unicorn
January 4, 2022     Artificial Intelligence, News, Venture Capital

Chinese AI company has raised $100 million in a Series B round led by Sequoia China and joined by existing investors 5Y Capital and Gaorong Capital. Though the valuation…