System and method for prediction of protein-ligand interactions and their bioactivity
Inventors
Bastas, Orestis • Bucher, Alwin • Pabrinkis, Aurimas • Demtchenko, Mikhail • YANG, Zeyu • Jamieson, Cooper Stergis • Joĉys, {circumflex over (Z)}ygimantas • Tal, Roy • Knuff, Charles Dazler
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11176462-B1
Publication Date
2021-11-16
Expiration Date
Abstract
A system and method for computationally tractable prediction of protein-ligand interactions and their bioactivity. According to an embodiment, the system and method comprise two machine learning processing streams and concatenating their outputs. One of the machine learning streams is trained using information about ligands and their bioactivity interactions with proteins. The other machine learning stream is trained using information about proteins and their bioactivity interactions with ligands. After the machine learning algorithms for each stream have been trained, they can be used to predict the bioactivity of a given protein-ligand pair by inputting a specified ligand into the ligand processing stream and a specified protein into the protein processing stream. The machine learning algorithms of each stream predict possible protein-ligand bioactivity interactions based on the training data.
Core Innovation
The invention provides a system and a method for prediction of protein-ligand interactions and their bioactivity using two trained machine learning components. Chemical notation for a target molecule and chemical notation for a target protein segment are received and processed separately, with the target molecule processed through a trained graph-based neural network to obtain a first vector result and the target protein segment processed through a trained sequence-based neural network to obtain a second vector result.
The first vector result and the second vector result are concatenated to obtain a concatenated vector result. A prediction is then made as to the bioactivity of the target molecule and the target protein segment using the concatenated vector result. In dependent implementations, ligand parsing is used to create a graph-based representation associated with molecule bioactivity data describing known or suspected interactions.
The dependent implementations also convert protein segments into vector representations and associate these with protein bioactivity data describing known or suspected ligand interactions and resulting bioactivity. The graph-based neural network is further constrained to a message passing neural network, and the sequence-based neural network is further constrained to a long short term memory neural network or a multi-head attention transformer.
Claims Coverage
The provided material includes two independent claims: a system claim and a method claim. Each independent claim is centered on using two trained neural networks, producing vector results, concatenating the vectors, and using the concatenated vector result for bioactivity prediction.
Dual-stream bioactivity prediction using graph and sequence vector analyses
A system that receives chemical notation for a target molecule and chemical notation for a target protein segment, processes the target molecule through a trained graph-based neural network to obtain a first vector result, processes the target protein segment through a trained sequence-based neural network to obtain a second vector result, concatenates the first and second vector results to obtain a concatenated vector result, and makes a prediction as to the bioactivity of the target molecule and target protein segment using the concatenated vector result.
Dual-stream bioactivity prediction using concatenated vector result
A method for prediction of protein-ligand interactions and their bioactivity that receives chemical notation for a target molecule and for a target protein segment, processes the target molecule through the trained graph-based neural network to obtain a first vector result, processes the target protein segment through the trained sequence-based neural network to obtain a second vector result, concatenates the first and second vector results to obtain a concatenated vector result, and makes a prediction as to the bioactivity of the target molecule and target protein segment using the concatenated vector result.
Across both independent claims, the inventive core is the combination of a trained graph-based neural network analysis of a target molecule, a trained sequence-based neural network analysis of a target protein segment, concatenation of the resulting vectors, and making a bioactivity prediction using the concatenated vector result.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?