Gene-specific prediction
A gene-specific prediction tool for classifying and interpreting gene tests is described. The prediction tool includes a classifier trained and tested using databases of gene variants and their known phenotypes. The classifier uses differences between features of amino acids in obtaining attributes used to perform classification and generate predictions, including for benign and pathologic outcomes, for uncertain gene variants.
1. A method of predicting a result of genetic mutation, comprising:
determining differences between (i) values of a first plurality of features, of amino acids encoded by a variant of a wild-type polynucleotide sequence, and (ii) values of a second plurality of features, of amino acids encoded by the wild-type polynucleotide sequence;
wherein the features of the first plurality are the same as the features of the second plurality; and
by a processor and based on the differences, determining a predicted phenotype severity of the variant.
2. The method of claim 1 , wherein the prediction includes an indication of a disease.
3. The method of claim 1 , wherein at least one of the first plurality of features is selected from the group consisting of alpha NH chemical shifts, normalized frequency of C terminal helix, normalized frequency of chain reversal IR, normalized positional frequency at helix termini N2, partition coefficient Garel, relative preference value at C2, relative preference value at N1, weights for beta sheet at the window position of 0, amino acid distribution, average relative fractional occurrence in A0(i), average relative probability of inner beta sheet, composition, effective partition energy, free energy in alpha helical region, frequency of the third residue in turn, helix formation parameters (delta delta G), hydrophobicity, membrane buried preference parameters, normalized frequency of beta structure, normalized frequency of con, normalized positional frequency at helix termini Cc, STERIMOL maximum width of the side chain, and Zimm Bragg parameter sigma.
4. The method of claim 1 , wherein at least one of the first plurality of features is selected from the group consisting of alpha NH chemical shifts, normalized frequency of C terminal helix, normalized frequency of chain reversal R, normalized positional frequency at helix termini N2, partition coefficient Garel, relative preference value at C2, relative preference value at N1, weights for beta sheet at the window position of 0, amino acid distribution, average relative fractional occurrence in A0(i), average relative probability of inner beta sheet, composition, effective partition energy, free energy in alpha helical region, frequency of the third residue in turn, helix formation parameters (delta delta G), hydrophobicity, membrane buried preference parameters, normalized frequency of beta structure, normalized frequency of coil, normalized positional frequency at helix termini Cc, STERIMOL maximum width of the side chain, and Zimm Bragg parameter sigma.
5. A method of predicting a result of genetic mutation, comprising:
determining differences between (i) values of a first plurality of features, of amino acids encoded by a variant of a wild-type polynucleotide sequence, and (ii) values of a second plurality of features, of amino acids encoded by the wild-type polynucleotide sequence;
wherein the features of the first plurality are the same as the features of the second plurality;
by a processor, training a machine learning classification algorithm using the differences; and
determining, by the trained algorithm, a prediction of phenotype severity of the gene variant.
6. The method of claim 5 , wherein the machine learning classification algorithm includes at least one of Zero Rules (ZeroR), naive Bayesian, Simple Logistic Regression (Simple Logistic), Support Vector Machine (SMO), k-nearest neighbor (IBk), or Random Forest Regression (Random Forest).
7. The method of claim 5 , wherein the plurality of features comprises at least one of physical, chemical, conformational, physiochemical, biochemical, or energetic properties of amino acids.
8. The method of claim 5 , further including selecting the plurality of features by performing correlation-based feature subset selection and best-first greedy hill-climbing search and identifying a subset of properties that differentiate benign mutations from pathogenic mutations.
9. A non-transitory computer-readable medium encoded with machine-executable instructions for:
determining differences between (i) values of a first plurality of features, of amino acids encoded by a variant of a wild-type polynucleotide sequence, and (ii) values of a second plurality of features, of amino acids encoded by the wild-type polynucleotide sequence;
wherein the features of the first plurality are the same as the features of the second plurality; and
by a processor and based on the differences, determining a predicted phenotype severity of the variant.
10. The method of claim 9 , wherein the prediction includes an indication of a disease.
11. The method of claim 9 , wherein at least one of the first plurality of features is selected from the group consisting of alpha NH chemical shifts, normalized frequency of C terminal helix, normalized frequency of chain reversal R, normalized positional frequency at helix termini N2, partition coefficient Garel, relative preference value at C2, relative preference value at N1, weights for beta sheet at the window position of 0, amino acid distribution, average relative fractional occurrence in A0(i), average relative probability of inner beta sheet, composition, effective partition energy, free energy in alpha helical region, frequency of the third residue in turn, helix formation parameters (delta delta G), hydrophobicity, membrane buried preference parameters, normalized frequency of beta structure, normalized frequency of coil, normal zed positional frequency at helix termini Cc, STERIMOL maximum width of the side chain, or Zimm Bragg parameter sigma.
12. The method of claim 9 , wherein at least one of the first plurality of features is selected from the group consisting of alpha NH chemical shifts, normalized frequency of C terminal helix, normalized frequency of chain reversal R, normalized positional frequency at helix termini N2, partition coefficient Garel, relative preference value at C2, relative preference value at N1, weights for beta sheet at the window position of 0, amino acid distribution, average relative fractional occurrence in A0(i), average relative probability of inner beta sheet, composition, effective partition energy, free energy in alpha helical region, frequency of the third residue in turn, helix formation parameters (delta delta G), hydrophobicity, membrane buried preference parameters, normalized frequency of beta structure, normalized frequency of coil, normalized positional frequency at helix termini Cc, STERIMOL maximum width of the side chain, or Zimrn Bragg parameter sigma.
13. A non-transitory computer-readable medium encoded with machine-executable instructions for:
determining differences between (I) values of a first plurality of features, of amino acids encoded by a variant of a wild-type polynucleotide sequence, and (ii) values of a second plurality of features, of amino acids encoded by the wild-type polynucleotide sequence;
wherein the features of the first plurality are the same as the features of the second plurality;
training a machine learning classification algorithm using the differences; and
determining, by the trained algorithm, a prediction of phenotype severity of the gene variant.
14. The method of claim 13 , wherein the machine learning classification algorithm includes at least one of Zero Rules (ZeroR), naive Bayesian, Simple Logistic Regression (Simple Logistic), Support Vector Machine (SMO), k-nearest neighbor (lBk), or Random Forest Regression (Random Forest).
15. The method of claim 13 , wherein the plurality of features comprises at least one of physical, chemical, conformational, physiochemical, biochemical, or energetic properties of amino acids.
16. The method of claim 13 , further including selecting the plurality of features by performing correlation-based feature subset selection and best-first greedy hill-climbing search and identifying a subset of properties that differentiate benign mutations from pathogenic mutations.