Methods and systems for engineering biomolecules

Abstract

Disclosed are methods for building a sequence activity model with reference to structural data, which model can be used to guide directed evolution of proteins having beneficial properties. Some embodiments use genetic algorithms and structural data to filter out uninformative data. Some embodiments use a support vector machine to train the sequence activity model. The filtering and training methods can generate a sequence activity model having higher predictive power than conventional modeling methods. Systems and computer program products implementing the methods are also provided.

What is claimed is: 
     
         1 . A method, implemented using a computer system comprising one or more processors and system memory, of conducting directed evolution, the method comprising:
 (a) receiving, by the computer system, an unfiltered data set comprising the following information for each of a plurality of variant biomolecules: (i) the variant biomolecule's activity on a ligand in a binding site of the variant biomolecule, (ii) a sequence of the variant biomolecule, wherein the sequence is a nucleic acid sequence or a protein sequence, and (iii) one or more geometric parameters characterizing the geometry of the variant biomolecule regarding the ligand in the binding site of the variant biomolecule;   (b) filtering, by the one or more processors, the unfiltered data set to generate a filtered data subset, comprising:
 obtaining a plurality of data subsets from the unfiltered data set, each data subset is obtained by removing from the unfiltered data set at least one of the geometric parameters of the plurality of variant biomolecules and/or removing certain variant biomolecules having geometric parameter values outside defined ranges, 
 training a plurality of sequence activity models, wherein each of the plurality of sequence activity models is trained using a data subset of the plurality of data subsets, and wherein the sequence of the variant biomolecule and one or more of the geometric parameters are independent variables of the sequence activity models and the variant biomolecule's activity on the ligand is a dependent variable of the sequence activity models, 
 testing the plurality of sequence activity models' predictive abilities to predict variant biomolecules' activities on the ligand, and 
 identifying, as the filtered data subset, a data subset that is used to train a sequence activity model having a higher predictive ability than a sequence activity model trained with the unfiltered data set; 
   (c) applying, by the one or more processors, a sequence activity model trained using the filtered data subset to a plurality of new variant biomolecules, and identifying, using the one or more processors, one or more new variant biomolecules that are predicted by the sequence activity model trained using the filtered data subset to have a desired level of the activity on the ligand, wherein each variant biomolecule of the plurality of new variant biomolecules is different from the plurality of variant biomolecules; and   (d) synthesizing or expressing the one or more new variant biomolecules identified in (c), wherein the one or more new variant biomolecules' activities on the ligand are improved over the plurality of variant biomolecules as measured by an assay.   
     
     
         2 . The method of  claim 1 , wherein the filtering the unfiltered data set is performed with a genetic algorithm that varies thresholds for removing information associated with the geometric parameters for one or more variant biomolecules. 
     
     
         3 . The method of  claim 1 , wherein applying the sequence activity model trained using the filtered data subset comprises recombining or mutating one or more known variant biomolecules to obtain potential new variant biomolecules, the one or more known variant biomolecules are predicted by the sequence activity model to have activities meeting a criterion. 
     
     
         4 . The method of  claim 3 , further comprising evaluating the potential new variant biomolecules using the sequence activity model trained using the filtered data subset as a fitness function to identify the one or more new variant biomolecules. 
     
     
         5 . The method of  claim 1 , further comprising
 producing a structural model for each of the new variant biomolecules; and   using the structural models to generate geometric parameters characterizing the geometry of the ligand in the binding sites of the new variant biomolecules.   
     
     
         6 . The method of  claim 1 , further comprising receiving structural models of variant biomolecules and determining the one or more geometric parameters using the structural models. 
     
     
         7 . The method of  claim 6 , wherein the structural models are homology models, and optionally wherein the homology models are prepared using physical structural measurement details of biomolecules. 
     
     
         8 . The method of  claim 7 , wherein the physical structural measurement details of biomolecules comprise three-dimensional positions of atoms obtained by NMR or x-ray crystallography. 
     
     
         9 . The method of  claim 6 , further comprising using a docker to determine the one or more geometric parameters. 
     
     
         10 . The method of  claim 1 , wherein the information for each of the plurality of variant biomolecules further comprises (iv) an interaction energy characterizing the interaction of the ligand in the binding site. 
     
     
         11 . The method of  claim 10 , further comprising using a docker to determine the interaction energy. 
     
     
         12 . The method of  claim 1 , wherein the sequence activity model trained using the filtered data subset is obtained by a support vector machine, a multiple linear regression, a principal component regression, a partial least square regression, or a neural network. 
     
     
         13 . The method of  claim 1 , wherein the plurality of variant biomolecules comprises a plurality of enzymes. 
     
     
         14 . The method of  claim 13 , wherein the activity of the variant biomolecule on a ligand is the activity of an enzyme on a substrate. 
     
     
         15 . The method of  claim 1 , further comprising assaying the new variant biomolecules for activity. 
     
     
         16 . The method of  claim 1 , wherein the filtering the unfiltered data set comprises removing from the unfiltered data set certain variant biomolecules having geometric parameter values outside defined ranges. 
     
     
         17 . The method of  claim 1 , wherein the filtering the unfiltered data set is performed with a genetic algorithm. 
     
     
         18 . The method of  claim 1 , wherein each of the one or more new variant biomolecules has a sequence that differs from the sequences of the variant biomolecules providing information for the data set of (a).