Methods for peptide-protein binding prediction
OwnerStephen L. Hoffman, Sorin Istrail +4inventors
InventorsStephen L. Hoffman, Sorin Istrail, Liliana Florea, Bjarni V. Halldorsson, Oliver Kohlbacher, Russell S. Schwartz
Life of this patent
Estimating the term…
Abstract
The present teachings are related to methods for predicting the binding affinity of peptides for proteins. Individual methods for creating predictions and methods for combining predictions are also related. Some embodiments of the present teachings include methods for manufacturing a vaccine and a computer system for implementing the methods of the present teachings.
Claims84
A method for predicting the binding affinity of a peptide for a MHC protein, comprising providing a prediction of binding affinity by two or more methods, and then combining the predictions of those methods.
A method for assigning a categorical binding affinity for a protein to a candidate peptide based on information for a set of known peptides for a protein, comprising:
- obtaining sequence and binding affinity information for a set of known peptides;
- obtaining sequence data for a candidate peptide;
- generating an evaluation of the affinity of said candidate peptide for said protein using said sequence and binding affinity information for a set of known peptides and said sequence data for candidate peptides;
- ranking said candidate peptide and said known peptides on the basis of relative binding affinities for said protein;
- and
- assigning peptides to categories of binding affinity based on rank.
A method for evaluating the affinity of a candidate peptide for a protein, consisting of:
- obtaining sequence and binding affinity information for a set of known peptides;
- obtaining sequence data for a candidate peptide;
- generating a first evaluation of affinity of said candidate peptide for said protein by comparing the sequence of said candidate peptide with data developed from said information for a set of known peptides with a first prediction method;
- generating one or more additional evaluations of affinity of said candidate peptide for said protein using said information for a set of known peptides and one or more additional prediction methods different from said first method used to generate said first evaluation of affinity;
- and
- combining said first evaluation and said one or more additional evaluations into a combined evaluation.
An algorithmic quadratic programming (QP) method, comprising methods for predicting the relative affinities of peptides proteins.
A linear programming method of predicting peptide affinity for a protein.
A profile-based method of predicting peptide affinity for a protein, comprising:
- obtaining information for a set of known peptides;
- creating one or more motifs for peptides with affinity to said protein by analyzing said information for a set of known peptides;
- and
- evaluating the affinity of a candidate peptide based on said one or more motifs.
An iterative multiple alignment (aln) method of predicting the relative binding affinity of a peptide for a MHC protein, comprising:
- obtaining sequence and affinity information for a set of known epitopes for said MHC protein;
- deriving a motif from said information for a set of known epitopes via an iterative multiple alignment heuristic;
- generating a score for said peptide based on its similarity to said motif;
- and
- predicting the relative binding affinity of said peptide for said MHC protein based on said score.
A letter frequency (LetFq) method of predicting the relative binding affinity of a peptide for a MHC protein, comprising:
- obtaining sequence and affinity information for a set of known epitopes for said MHC protein;
- deriving a motif from said information for a set of known epitopes via a letter frequency heuristic;
- generating a score for said peptide based on its similarity to said motif;
- and
- predicting the relative binding affinity of said peptide for said MHC protein based on said score.
The method of 54, further comprising:
- deriving multiple motifs from a set of known epitopes via a letter frequency heuristic;
- generating scores for said peptide, each of said scores based on its similarity to one of said motifs;
- and
- predicting the relative binding affinity of said peptide for said MHC protein based on the selection of the score showing the highest binding affinity.
A method based on χ2 statistical significance (Ki2) tests predicting the relative binding affinity of a peptide for a MHC protein, comprising:
- obtaining sequence and affinity information for a set of known epitopes for said MHC protein;
- deriving a motif from said information for a set of known epitopes according to dependencies between peptide positions, revealed by (i2 significance tests;
- generating a score for said peptide based on its similarity to said motif;
- and
- predicting the relative binding affinity of said peptide for said MHC protein
A method of selecting peptides with a desired level of affinity for a protein, comprising:
- obtaining sequence data for a set of candidate peptides and data for peptides of known affinity for said protein known peptides;
- evaluating the affinity of said candidate peptides for said protein using two or more methods, at least one of which uses the data for the peptides of known affinity;
- and
- selecting peptides from said set of candidate peptides that are predicted by said two or more methods to have said desired level of affinity for said protein.
A method for providing treatment for a patient with, or at risk for, cancer or a viral infection, comprising:
- obtaining sequence data for candidate peptides;
- obtaining data for a set of known peptides;
- evaluating the affinity of said candidate peptides for a protein with two or more prediction methods;
- selecting candidate peptides with a desired level of affinity for the protein for inclusion in a treatment;
- and
- treating said patient by stimulating an immune response in said patient with said treatment.
A method for screening a set of candidate peptides, comprising:
- obtaining sequence data for candidate peptides;
- obtaining data for a set of known peptides;
- evaluating the affinity of said candidate peptides for a protein with two or more prediction methods;
- assigning said candidate peptides to two or more groups based on the evaluations of affinity;
- and
- screening said set of candidate peptides on the basis of group assignment.
A method for identifying epitopes, comprising:
- obtaining sequence data for candidate peptides;
- obtaining data for a set of known peptide epitopes that bind a MHC protein;
- evaluating the affinities of said candidate peptides for said MHC protein using two or more prediction methods and said data for a set of known peptide epitopes;
- assigning said candidate peptides to two or more groups based on the evaluations of affinity;
- and
- identifying a set of said candidate peptides by group assignment.
A computer system for predicting the affinity of candidate peptides for a protein, comprising:
- a database containing data for sets of known peptides;
- a database containing sequence information for said candidate peptides;
- a processor capable of evaluating peptides by multiple methods;
- and
- a database for storing evaluations.
Dependencies are inferred from the claim text.