17806076. TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH THREE-DIMENSIONAL DISTANCE DATA simplified abstract (Microsoft Technology Licensing, LLC)
TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH THREE-DIMENSIONAL DISTANCE DATA
Organization Name
Microsoft Technology Licensing, LLC
Inventor(s)
TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH THREE-DIMENSIONAL DISTANCE DATA - A simplified explanation of the abstract
This abstract first appeared for US patent application 17806076 titled 'TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH THREE-DIMENSIONAL DISTANCE DATA
Simplified Explanation
The abstract describes a computing system that uses a transformer-based graph neural network to analyze molecular graphs and predict energy changes in a molecular system. Here are the key points:
- The system includes a processor that provides a training data set consisting of pre-transformation molecular graphs and corresponding energy parameter values.
- The pre-transformation molecular graphs are fully connected networks of normal nodes, connected by edges.
- The processor encodes structural information by representing the three-dimensional Euclidean distance along each edge as learnable embeddings.
- The training data set is inputted to a transformer-based graph neural network, which is trained to perform inference.
- During inference, the processor receives a pre-transformation molecular graph as input, and outputs the predicted post-transformation energy parameter value.
Potential applications of this technology:
- Drug discovery: The system can be used to analyze molecular structures and predict energy changes, aiding in the development of new drugs.
- Material design: It can be applied to analyze the structural properties of materials and predict their energy changes, assisting in the design of new materials with desired properties.
- Chemical reactions: The system can help understand and predict the energy changes that occur during chemical reactions, facilitating reaction optimization and synthesis planning.
Problems solved by this technology:
- Complex molecular systems: The system provides a way to analyze and predict energy changes in complex molecular systems, which is challenging using traditional methods.
- Efficient analysis: The use of a transformer-based graph neural network allows for efficient analysis of large amounts of molecular data, enabling faster and more accurate predictions.
- Structural encoding: The learnable embeddings of the three-dimensional Euclidean distances capture important structural information, improving the accuracy of predictions.
Benefits of this technology:
- Improved accuracy: By using a transformer-based graph neural network and structural encoding, the system can provide more accurate predictions of energy changes in molecular systems.
- Time and cost savings: The system enables faster analysis of molecular data, reducing the time and cost required for tasks such as drug discovery and material design.
- Enhanced understanding: The predictions provided by the system can help researchers gain a deeper understanding of molecular systems and their energy changes, leading to new insights and discoveries.
Original Abstract Submitted
A computing system is provided, including a processor configured to, during a training phase, provide a training data set including a pre-transformation molecular graph and post-transformation energy parameter value representing an energy change in a molecular system following an energy transformation. The pre-transformation molecular graph includes a plurality of normal nodes fully connected by edges. The processor is configured to encode structural information including a three-dimensional Euclidean distance along an edge connecting a pair of the normal nodes in each molecular graph as learnable embeddings. The processor is configured to input the training data set to a transformer-based graph neural network to train the network to perform an inference at inference time. The processor is further configured to receive inference-time input of the inference-time pre-transformation molecular graph at the trained transformer-based graph neural network, and output the inference-time post-transformation energy parameter value based on the inference-time pre-transformation molecular graph.