Apparatus and method for training a tunable data structure to predict internal ribosome entry site (IRES) activity
An apparatus for training a tunable data structure to predict internal ribosome entry site (IRES) activity includes at least a processor and a memory containing instructions configuring the at least a processor to assemble a training set including a plurality of nucleotide sequence data examples IRES sequences and a plurality of correlated observed IRES activity, partition the training set into at least a first section and a second section, train, using the first section at least an activity data structure to generate probable IRES activity using nucleotide sequence data, and iteratively retrain the at least an activity data structure using the second section, wherein each iteration of the iterative retraining includes generating a predicted IRES activity value using the at least an activity neural network and a nucleotide sequence data example, evaluating an error function, and tuning the activity data structure.
1. An apparatus for training a tunable data structure to predict internal ribosome entry site (IRES) activity, the apparatus comprising:
at least a processor; and
a memory communicatively connected to the at least a processor, the memory containing instructions configuring the at least a processor to:
assemble a training set, the training set comprising a plurality of nucleotide sequence data examples describing a plurality of IRES sequences and a plurality of correlated observed IRES activity, wherein:
assembling the training set further comprises converting each IRES sequence of the plurality of IRES sequences into a corresponding nucleotide sequence data example of the plurality of nucleotide sequence data examples using a representation generation data structure;
the representation generation data structure has an input having a first sequence size;
an IRES sequence of the plurality of IRES sequences has a second sequence size greater than the first sequence size; and
converting the IRES sequence to a nucleotide sequence data example of the plurality of nucleotide sequence data examples further comprises:
inputting a plurality of sections of the IRES sequence, wherein each section has a section sequence size equal to the first sequence size;
outputting a plurality of sectional nucleotide sequence data outputs;
aggregating the sectional nucleotide sequence data outputs; and
generating the nucleotide sequence data example using the aggregated sectional nucleotide sequence data outputs;
partition the training set into at least a first section and a second section;
train, using the first section, at least an activity data structure to generate probability of IRES activity using nucleotide sequence data; and
iteratively retrain the at least an activity data structure using the second section, wherein each iteration of the iterative retraining further comprises:
generating a predicted IRES activity value using the at least an activity data structure and a nucleotide sequence data example;
evaluating an error function using the predicted IRES activity value and an observed IRES activity correlated to the nucleotide sequence data example; and
tuning the at least an activity data structure using the evaluated error function.
2. The apparatus of claim 1 , wherein the representation generation data structure further comprises a neural network.
3. The apparatus of claim 2 , wherein the neural network further comprises a bidirectional encoder representations from transformer (BERT) neural network.
4. The apparatus of claim 1 , wherein the at least an activity data structure further comprises at least an activity neural network.
5. The apparatus of claim 4 , wherein the at least an activity neural network further comprises a graph neural network.
6. The apparatus of claim 4 , wherein the at least an activity neural network further comprises a convolutional neural network.
7. The apparatus of claim 1 , wherein the at least an activity data structure further comprises a plurality of activity data structures, and further comprising:
generating, for each activity data structure of the plurality of activity data structures, at least an efficacy metric; and
selecting an activity data structure of the plurality of activity data structures as a function of the at least an efficacy metric.
8. The apparatus of claim 7 , wherein each activity data structure of the plurality of activity data structures is trained with a distinct training data partition.
9. A method of training a tunable data structure to predict internal ribosome entry site (IRES) activity, the method comprising:
assembling, by a computing device, a training set, the training set comprising a plurality of nucleotide sequence data examples describing a plurality of IRES sequences and a plurality of correlated observed IRES activity, wherein:
assembling the training set further comprises converting each IRES sequence of the plurality of IRES sequences into a corresponding nucleotide sequence data example of the plurality of nucleotide sequence data examples using a representation generation data structure;
the representation generation data structure has an input having a first sequence size;
an IRES sequence of the plurality of IRES sequences has a second sequence size greater than the first sequence size; and
converting the IRES sequence to a nucleotide sequence data example of the plurality of nucleotide sequence data examples further comprises:
inputting a plurality of sections of the IRES sequence, wherein each section has a section sequence size equal to the first sequence size;
outputting a plurality of sectional nucleotide sequence data outputs;
aggregating the sectional nucleotide sequence data outputs; and
generating the nucleotide sequence data example using the aggregated sectional nucleotide sequence data outputs;
partitioning the training set into at least a first section and a second section;
training, using the first section, at least an activity data structure to generate probability of IRES activity using nucleotide sequence data; and
iteratively retraining the at least an activity data structure using the second section, wherein each iteration of the iterative retraining further comprises:
generating a predicted IRES activity value using the at least an activity neural network and a nucleotide sequence data example;
evaluating an error function using the predicted IRES activity value and an observed IRES activity correlated to the nucleotide sequence data example; and
tuning the at least an activity data structure using the evaluated error function.
10. The method of claim 9 , wherein the representation generation data structure further comprises a neural network.
11. The method of claim 10 , wherein the neural network further comprises a bidirectional encoder representations from transformer (BERT) neural network.
12. The method of claim 9 , wherein the at least an activity data structure further comprises at least an activity neural network.
13. The method of claim 12 , wherein the at least an activity neural network further comprises a graph neural network.
14. The method of claim 12 , wherein the at least an activity neural network further comprises a convolutional neural network.
15. The method of claim 9 , wherein the at least an activity data structure further comprises a plurality of activity data structures, and further comprising:
generating, for each activity data structure of the plurality of activity data structures, at least an efficacy metric; and
selecting an activity data structure of the plurality of activity data structures as a function of the at least an efficacy metric.
16. A non-transitory machine-readable medium containing instructions configured to cause at least a processor to:
assemble a training set, the training set comprising a plurality of nucleotide sequence data examples describing a plurality of IRES sequences and a plurality of correlated observed IRES activity, wherein:
assembling the training set further comprises converting each IRES sequence of the plurality of IRES sequences into a corresponding nucleotide sequence data example of the plurality of nucleotide sequence data examples using a representation generation data structure;
the representation generation data structure has an input having a first sequence size;
an IRES sequence of the plurality of IRES sequences has a second sequence size greater than the first sequence size; and
converting the IRES sequence to a nucleotide sequence data example of the plurality of nucleotide sequence data examples further comprises:
inputting a plurality of sections of the IRES sequence, wherein each section has a section sequence size equal to the first sequence size;
outputting a plurality of sectional nucleotide sequence data outputs;
aggregating the sectional nucleotide sequence data outputs; and
generating the nucleotide sequence data example using the aggregated sectional nucleotide sequence data outputs;
partition the training set into at least a first section and a second section;
train at least an activity data structure to generate probable IRES activity using nucleotide sequence data; and
iteratively retrain the at least an activity data structure using the second section, wherein each iteration of the iterative retraining further comprises:
generating a predicted IRES activity value using the at least an activity neural network and a nucleotide sequence data example;
evaluating an error function using the predicted IRES activity value and an observed IRES activity correlated to the nucleotide sequence data example; and
tuning the at least an activity data structure using the evaluated error function.