Aladdin Healthcare Technologies validates its proprietary Deep Learning Algorithms that significantly improve Drug Discovery Performance
DGAP-News: Aladdin Healthcare Technologies SE
/ Key word(s): Miscellaneous
Aladdin Healthcare Technologies validates its proprietary Deep Learning Algorithms that significantly improve Drug Discovery Performance - Aladdin develops a new deep learning method for drug discovery with superior performance validated on six chemical property and toxicity datasets - Average prediction performance improved by 5-10% compared to other deep learning methods and by 20% compared to traditional computational methods - 1.7 times more accurate than current state of the art methods BERLIN/LONDON, July 30, 2020 - Aladdin Healthcare Technologies SE ("Aladdin", ISIN: DE000A12ULL2), a leading developer of Artificial Intelligence (AI) based healthcare diagnostics and drug discovery applications, had successfully validated its deep learning algorithms (Message Passing Neural Network) for drug discovery. Aladdin's technology has been validated last week by the 'International Joint Conferences on Artificial Intelligence (IJCAI)', a world leading organization for scientific purpose that spreads information on Artificial Intelligence (AI) by means of conferences at which educational materials like books and proceedings are presented. Aladdin's team of experts developed a new deep learning-based graph model for molecular representation. A good drug candidate should not only be efficient but also have appropriate ADMET properties at a therapeutic dose. ADMET stands for absorption, distribution, metabolism, excretion, and toxicity of the drug candidate. Extensive experiments demonstrated that Aladdin's technology has achieved superior prediction performances (ADMET properties prediction) on six chemical property and toxicity datasets, improving other deep learning methods by 5-10% and traditional computational methods by over 20%. Furthermore, Aladdin's method is 1.7 times more accurate than other state-of-the-art methods when the training data set has no similar scaffold to the test set. The successful results of Aladdin's deep learning algorithms have been accepted by the 29. International Joint Conference on AI among 4,717 valid submissions. The overall acceptance rate was only 12.6%. Constructing proper representations of molecules lies at the core of numerous tasks such as molecular property prediction, virtual screening and drug design. Deep learning methods by using Graph neural networks, especially Message Passing Neural Networks (MPNN) and its variants, have recently made remarkable achievements in drug molecular modeling. Albeit powerful, the one-sided focus on atom (node) or bond (edge) information of existing MPNN methods leads to insufficient representations of the attributed molecular graphs. Aladdin has now developed a Communicative Message Passing Neural Network (CMPNN) to improve the molecular embedding by strengthening the message interactions between nodes and edges through a communicative kernel. Wade Menpes-Smith, CEO of Aladdin Healthcare Technologies SE, comments: "This is a further validation of our world-class AI technology capability and another key milestone that is a catalyst for further breakthroughs in AI drug discovery." Further information about Aladdin's method and results can be found on https://www.ijcai.org/Proceedings/2020/392.
30.07.2020 Dissemination of a Corporate News, transmitted by DGAP - a service of EQS Group AG. |
Language: | English |
Company: | Aladdin Healthcare Technologies SE |
Unter den Linden 10 | |
10117 Berlin | |
Germany | |
Phone: | 030 700140449 |
E-mail: | info@aladdinid.com |
Internet: | www.aladdinid.com |
ISIN: | DE000A12ULL2 |
WKN: | A12ULL |
Listed: | Regulated Market in Dusseldorf |
EQS News ID: | 1105817 |
End of News | DGAP News Service |
|
1105817 30.07.2020