Molecular interaction predictors

Abstract

Adaptive threading models for predicting an interaction between two or more molecules such as proteins are provided. The adaptive threading models have one or more learnable parameters that can be learned from all or some of the available data. The available data can include data relating to known interactions between the two or more molecules, the composition of the molecules and the geometry of the molecular complex.

1 . A system stored on computer-readable media, comprising: 
 two or more parameters machine learned at least in part utilizing data relating to a set of known interactions between proteins and ligands, two of the two or more parameters relating to pairwise contact potentials and weights; and    a prediction component configured to predict information relating to an unknown interaction between a protein and a ligand by employing information about the protein's sequence, the ligand's sequence, a geometry of a protein-ligand complex and an adjusted threading model comprising the two or more parameters.    
   
   
       2 . The system of  claim 1 , further comprising an inference component configured to machine infer the geometry of the protein-ligand complex.  
   
   
       3 . (canceled)  
   
   
       4 . The system of  claim 1 , wherein the adjusted threading model has a soft step function.  
   
   
       5 . The system of  claim 4 , wherein the soft step function is a sigmoid function.  
   
   
       6 . The system of  claim 4 , wherein the soft step function has at least one machine learned parameter.  
   
   
       7 . The system of  claim 6 , wherein the at least one machine learned parameter comprises a threshold distance and a smoothness of the soft step function.  
   
   
       8 . The system of  claim 1 , wherein the information relating to the unknown interaction is a binding energy.  
   
   
       9 . The system of  claim 1 , wherein the unknown interaction between the protein and the ligand is between an MHC molecule and a peptide of about 8-11 amino acids in length.  
   
   
       10 . The system of  claim 9 , wherein the MHC molecule is a synthetic molecule.  
   
   
       11 - 20 . (canceled)  
   
   
       21 . Computer-executable instructions stored on computer-readable media, the computer-executable instructions encoding a method, comprising: 
 providing two or more parameters machine learned at least in part utilizing data relating to a set of known interactions between proteins and ligands, two of the two or more parameters comprising pairwise contact potentials and weights; and    predicting information relating to an unknown interaction between a protein and a ligand using the two or more parameters and information about the protein's sequence, the ligand's sequence and a geometry of a protein-ligand complex.    
   
   
       22 . The computer-executable instructions of  claim 21 , the method further comprising inferring the geometry of the protein-ligand complex.  
   
   
       23 . The computer-executable instructions of  claim 21 , wherein the information relating to the unknown interaction between the protein and the ligand is a binding energy.  
   
   
       24 . The computer-executable instructions of  claim 21 , wherein the unknown interaction between the protein and the ligand is between an MHC molecule and a peptide of about 8-11 amino acids in length.  
   
   
       25 . The computer-executable instructions of  claim 24 , wherein the MHC molecule is a synthetic molecule.  
   
   
       26 . A method for generating an adaptive threading system, the method stored on computer-readable media, the method comprising: 
 machine learning two or more parameters at least in part utilizing data relating to a set of known interactions between proteins and ligands, two of the two or more parameters comprising pairwise contact potentials and weights.    
   
   
       27 . The method of  claim 26 , wherein machine learning comprises employing a Bayesian network.  
   
   
       28 . The method of  claim 26 , wherein the set of known interactions between proteins and ligands comprises a set of binding energies.  
   
   
       29 . The method of  claim 26 , wherein the proteins are MHC molecules and the ligands are peptides of about 8-11 amino acids in length.  
   
   
       30 . The method of  claim 29 , wherein at least one of the MHC molecules is a synthetic molecule.