Methods and systems for identification of a protein binding site
AssigneeLaboratory Corporation of America Holdings
InventorsMark Evans, Mojgan Haddad
Life of this patent
Estimating the term…
Abstract
Processes for identification of binding sites on protein molecules such as epitopes are provided. The disclosed methods in some embodiments use a combination of protein sequence data, structural information, experimental data on binding affinity, and computational modeling in order to identify binding sites on protein molecules. Systems and computer readable media for implementing the disclosed methods are provided. Also provided are compositions comprising a binding site or antibody that interacts with the binding site as an active ingredient and methods of using such compositions.
Claims13
A system for identifying a ligand binding site of a protein, comprising:
- a computer readable medium;
- and
- a processor in communication with the computer readable medium, the program configured to:
- receive sequence information for a plurality of sequence variants of an immunogenic protein;
- receive biological activity information for each of the plurality of the sequence variants for the ligand;
- receive data regarding a three-dimensional model of the immunogenic protein's structure to define patches of the immunogenic protein structure containing a plurality of residues within close proximity of each other, wherein the patches comprise surface residues, non-surface residues that are in close proximity, or a combination thereof;
- correlate the sequence information within each of the defined patches with the biological activity data for each sequence variant, wherein the information on the sequence variability within each defined patch and the biological activity data for each sequence variant is represented in a non-binary or binary form, compiled as a vector file;
- apply two or more computational methods to the sequence information and biological activity data within each of the defined patches, wherein the two or more computational methods comprise i) at least one method involving multivariate analysis, at least one method involving univariate analysis, and at least one machine learning algorithm, or ii) two or more of multiple regression analysis, logistic regression analysis, support vector machine, Fischer's Exact test, Hidden Markov Models, Neural Networks, Random Forest, Decision Trees, or Bayesian Networks;
- compare the results obtained by applying the two or more computational methods;
- and
- identify at least one residue of the binding site based on agreement between the two or more computational methods.
Dependencies are inferred from the claim text.