Shift-invariant predictions
AssigneeMicrosoft Corporation
InventorsNebojsa Jojic, David Heckerman, Noah Zaitlen, Manuel Gomez
Life of this patent
Estimating the term…
Abstract
Shift invariant predictors are described herein. By way of example, a system for predicting binding information relating to a binding of a protein and a ligand can include a trained binding model and a prediction component. The trained binding model can include a hidden variable representing an unknown alignment of the ligand at a binding site of the protein. The prediction component can be configured to predict the binding information by employing information about the protein's sequence, the ligand's sequence and the trained binding model.
Claims20
A system for predicting binding information relating to a binding of a protein and a ligand, the system stored on computer-readable media, the system comprising:
- a trained binding model comprising a hidden variable representing an unknown alignment of the ligand with the protein at a binding site of the protein;
- and
- a prediction component configured to predict the binding information by employing information about the protein's sequence, the ligand's sequence and the trained binding model.
The system of claim 6, wherein the MHC class II molecule is a synthetic molecule.
A method of making a treatment by utilizing predicted binding information about a protein and a ligand, comprising:
- computing a probability distribution relating to an alignment of the ligand when bound to the protein;
- employing the probability distribution to predict binding information about the protein and the ligand;
- using the predicted binding information to identify a treatment;
- and
- making the treatment.
Computer-executable instructions encoding a method of generating a binding predictor, the computer-executable instructions stored on computer-readable media, the method comprising:
- estimating one or more parameters of the binding predictor from training data;
- estimating a probability distribution that relates to a geometric configuration of a ligand bound to a protein from the training data;
- and
- iteratively re-estimating the one or more parameters and the probability distribution until convergence.
Dependencies are inferred from the claim text.