Methods and systems for predicting membrane protein expression based on sequence-level information

1. A method for membrane protein expression, comprising:

selecting an initial set of sequences with known protein expression level for a membrane protein;

transforming the initial set of sequences to sequence features including nucleotide features, protein features, or both;

dividing the initial set of sequences into a training set of the sequence features and a test set of the sequence features;

creating a supervised machine learning model based on the training set of the sequence features and related expression levels, the model providing statistical analysis of the sequence features of the training set in relation to the related expression levels;

validating the supervised machine learning model by confirming relation between the sequence features of the test set and the related expression levels;

providing a distinct set of membrane protein sequences, each membrane protein sequence of the distinct set having an untested or unknown protein expression level;

performing a statistical modeling on each membrane protein sequence of the distinct set with the supervised machine learning model to provide an expression level prediction for each membrane protein sequence of the distinct set;

ranking membrane protein sequences of the distinct set having different expression levels, wherein the ranking is based on the expression level predictions, the ranking identifying candidate membrane protein sequences above a 50 th percentile of an entirety of the distinct set of membrane protein sequences,

selecting an experimentation subset having membrane protein sequences selected from the candidate membrane protein sequences having positive expression and identified to be above the 50 th percentile;

performing experimental testing on the experimentation subset such that expression levels are determined for each member of the subset, creating a positive result set of sequences that have positive expression of the membrane protein as determined by the experimental testing; and

expressing into a host membrane other proteins having the membrane protein sequences of the positive result set of sequences.

2. The method of claim 1 , wherein the initial set of sequences are alpha-helical membrane proteins.

3. The method of claim 2 , wherein the alpha-helical membrane proteins have a single transmembrane domain.

4. The method of claim 2 , wherein the alpha-helical membrane proteins have multiple transmembrane domains.

5. The method of claim 1 , wherein the features are transformed using one or more of the following methods: scaling, centering, PCA, autoencoding, and discretization.

6. The method of claim 1 , wherein at least one feature is described by one or more of the following categories: overall codon usage, codon pair score, tRNA adaptation index, 5′ codon usage, GC content, 5′ RNA Structure, overall RNA structure, Shine-Dalgarno-like sites, overall disorder, loop disorder, TM size/count, hydrophobicity, loop charge, loop length, and length/pl.

7. The method of claim 1 , wherein the supervised machine learning model uses a supervised-learning statistical framework.

8. The method of claim 6 , wherein the statistical framework is a support vector machine.

9. The method of claim 6 , wherein the statistical framework is a neural network.

10. The method of claim 6 , wherein the statistical framework is a decision tree.

11. The method of claim 1 , wherein the distinct set of membrane protein sequences are alpha-helical membrane protein sequences.

12. The method of claim 11 , wherein the alpha-helical membrane proteins have a single transmembrane domain.

13. The method of claim 11 , wherein the alpha-helical membrane proteins have multiple transmembrane domains.

14. A non-transient computer readable media comprising instructions to carry out the method of claim 1 .

15. The method of claim 1 , wherein at least one sequence feature of the sequence features is described by one or more of the following categories: overall codon usage and codon pair score.

16. A computer-implemented method for membrane protein expression, comprising:

collecting an initial set of sequences with known protein expression level for a membrane protein;

applying a transformation on the initial set of sequences to provide sequence features including nucleotide features, protein features, or both, the initial set of sequences including a training set of the sequence features and a test set of the sequence features;

running a supervised machine learning model trained on the training set of the sequence features and their known expression levels on the test set of sequence features, the model providing statistical analysis of the sequence features of the training set in relation to the related expression levels;

providing a distinct set of membrane protein sequences, each membrane protein sequence of the distinct set having an untested or unknown protein expression level;

running the supervised machine learning model to create a statistical modeling on each membrane protein sequence of the distinct set, providing an expression level prediction for each membrane protein sequence of the distinct set in a ranking by expression level predictions;

creating a set of candidate membrane protein sequences above a 50 th percentile in the ranking, the set of candidate membrane protein sequences containing an experimentation subset;

performing experimental testing on the experimentation subset such that expression levels are measured for each member of the subset, creating a positive result set of sequences that have positive expression of the membrane protein as determined by the experimental testing; and

expressing into a host membrane other proteins having the membrane protein sequences matching the positive result set of sequences.