Mathematical processes for determination of peptidase cleavage
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11069427-B2
Publication Date
2021-07-20
Expiration Date
Abstract
This invention relates to the identification of peptidase cleavage sites in proteins and in particular to identification protease cleavage by the endopeptidases. The present invention utilizes a bioinformatic methodology for prediction of peptidase cleavage sites based on principal component analysis and based on training sets obtained by experimental protein cleavage. This invention is not limited to training sets derived from CSL approaches, nor to any other experimental determination of cleavage site.
Core Innovation
The invention is a computer implemented process for identifying peptidase cleavage sites in a polypeptide. The process obtains an amino acid sequence for a target polypeptide and applies a structured prediction workflow based on experimentally derived amino acid physicochemical properties. Cleavage predictions are generated for cleavage-site dimers and then consolidated to identify amino acid subsets in the target polypeptide that are predicted to be cleaved.
A central part of the process derives an ensemble of peptidase cleavage prediction equations for each possible cleavage site dimer. Experimentally derived data of physicochemical properties are processed to produce a correlation matrix, and Principal Component Analysis is used to derive multiple uncorrelated dimensionless, weighted and ranked proxy descriptors describing at least 80% of the variance in the physicochemical properties of individual amino acids. Probabilistic modeling is used to derive a cleavage prediction equation for each specific cleavage site dimer.
The process further derives independently derived ensembles by repeatedly comparing cleaved and uncleaved peptide sets for the same cleavage site dimer and generating a plurality of independently derived prediction equations. For every possible cleavage site dimer up to a total of 400 such ensembles, the derived ensembles are stored on a non-transitory computer readable medium. The target polypeptide sequence is then processed by applying the proxy descriptors to derive vectors describing peptides of defined length, applying the ensembles to predict peptidase cleavage sites, and identifying cleavage probabilities at any cleavage site dimer by consensus.
Beyond cleavage-site identification, the invention links predicted cleavage to immunological selection and immunogen preparation. Amino acid subsets predicted to be cleaved are determined, and the resulting subsets are evaluated for MHC binding properties and to function as a T cell epitope that generates an immune response. An immunogen is prepared from the selected amino acid subsets predicted to have MHC binding and T-cell epitope function, and the immunogen is administered to a subject under conditions such that an immune response is generated.
Claims Coverage
The independent claim is directed to a computer implemented process that predicts peptidase cleavage sites and uses the predicted cleavage outcomes to select subsets with MHC binding/T-cell epitope function, prepare an immunogen, and administer it. The claim includes multiple inventive features, and dependent claims further refine modeling, peptide context, physicochemical feature selection, and peptidase specificity.
Ensemble derivation of cleavage-site prediction equations from PCA proxy descriptors
Deriving an ensemble of peptidase cleavage prediction equations for each possible cleavage site dimer by assembling experimentally derived data of amino acid physicochemical properties, producing a correlation matrix, deriving Principal Component Analysis proxy descriptors describing at least 80% of the variance, and using said proxy descriptors to describe peptide sets comprising experimentally determined cleaved and uncleaved cleavage site dimers.
Probabilistic modeling for each cleavage-site dimer using cleaved and uncleaved peptide sets
Comparing, via probabilistic modeling, the amino acid descriptors of peptide sets comprising experimentally determined cleaved and uncleaved cleavage site dimers to derive a cleavage prediction equation for said specific cleavage site dimer based on that peptide set.
Independent ensembles up to 400 cleavage-site dimer ensembles with consensus cleavage probabilities
Deriving an ensemble of independently derived prediction equations for every possible cleavage site dimer up to a total of 400 ensembles by repeatedly deriving independently derived prediction equations for said specific cleavage site dimer, storing said ensembles on a non-transitory computer readable medium, applying said ensembles to peptides of defined length from the target polypeptide, and identifying the probability of cleavage at any given cleavage site dimer in said target polypeptide by consensus.
MHC binding and T-cell epitope screening of predicted cleaved amino acid subsets
Defining amino acids subsets in said target polypeptide predicted to be cleaved by a peptidase and determining if the resulting amino acid subsets are predicted to have MHC binding properties and to function as a T cell epitope that generates an immune response.
Immunogen preparation from screened subsets and administration to generate immune response
Preparing an immunogen from amino acid subsets predicted to have MHC binding properties and to function as a T cell epitope that generates an immune response, and administering an effective amount of the immunogen to a subject under conditions such that an immune response is generated.
Probabilistic model type selection for cleavage prediction
Performing said probabilistic modeling using a probabilistic neural network or a support vector machine.
Accuracy threshold for probabilistic neural network cleavage prediction
Predicting a peptidase cleavage site with greater than about 70%, 80%, 90%, or 95% accuracy using a probabilistic neural network.
Overlapping peptide subsets generated from the N-terminus
Generating selected subsets of amino-acid sequences as overlapping 8-mers beginning at an N-terminus position and continuing so that overlapping 8-mers start at n and at n+1 until n+1 ends at the C-terminus for the selected length of the peptides.
Specific physicochemical property set for PCA input
Selecting physicochemical properties of individual amino acids from a specified group of polarity, hydrophobicity/hydrophilicity, and chromatography/thermodynamics-derived retention and hydration parameters, including combinations thereof.
Peptidase specificity to cathepsin
Carrying out the process using a peptidase that is specifically a cathepsin.
Overall, the claim set covers deriving PCA-based proxy descriptors from experimentally derived amino acid physicochemical properties, using probabilistic modeling to generate and repeatedly derive independently derived ensembles of cleavage prediction equations for cleavage-site dimers, applying consensus cleavage probabilities to predict cleaved amino acid subsets, and then screening those subsets for MHC binding and T-cell epitope function to prepare and administer an immunogen. Dependent claims refine the probabilistic model type and performance thresholds, constrain peptide subset construction, restrict the physicochemical property inputs, and restrict the peptidase to cathepsin.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?