Neural network for processing aptamer data
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for obtaining data defining a sequence for an aptamer, the aptamer comprising a string of nucleobases; encoding the data defining the sequence for the aptamer as a neural network input; and processing the neural network input using a neural network to generate an output that characterizes how strongly the aptamer binds to a particular target molecule, wherein the neural network has been configured through training to receive the data defining the sequence and to process the data to generate predicted outputs that characterize how strongly the aptamer binds to the particular target molecule.
1. A computer-implemented method for training a neural network to model results of an iterative SELEX process, wherein the iterative SELEX process begins with an initial pool of aptamers and iteratively updates the pool of aptamers at each of a plurality of different rounds, the method comprising:
obtaining results of the iterative SELEX process applied to the initial pool of aptamers;
obtaining data defining a plurality of sequences for a plurality of aptamers in the initial pool of aptamers, each aptamer from the plurality of aptamers comprising a string of nucleobases;
generating, based on the results of the iterative SELEX process, a respective label for each sequence from the plurality of sequences that comprises characteristics of how strongly the respective aptamer binds to at least one target molecule, wherein the respective label for each sequence from the plurality of sequences includes a plurality of sub-labels and each sub-label corresponds to a different round in the iterative SELEX process and characterizes a result of the corresponding round for the respective aptamer;
generating training data for a neural network from the data defining the plurality of sequences and the respective labels for each of the sequences; and
training the neural network on the training data, the neural network configured to generate, for a particular aptamer, a model output that comprises a plurality neural network outputs, each neural network output corresponding to a different round in the iterative SELEX process, and, wherein:
each neural network output is a prediction of a result of the corresponding SELEX round, if the particular aptamer were in the initial pool of aptamers as of the beginning of the corresponding SELEX round, and
the neural network output corresponding to a first SELEX round in the iterative SELEX process comprises a count, a binary value or a confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.
2. The method of claim 1 , wherein the neural network output corresponding to the first SELEX round in the iterative SELEX process comprises the binary value or the confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.
3. The method of claim 1 , wherein the count is a numerical value that represents a predicted count of the particular aptamer in the pool of aptamers for the first SELEX round.
4. The method of claim 1 , wherein the model output comprises a vector of fluorescence levels, where a brightness of each fluorescence level indicates how strongly the particular aptamer binds to the at least one target molecule.
5. The method of claim 1 , wherein the neural network is a convolutional neural network.
6. The method of claim 1 , wherein the data defining the plurality of sequences comprises, for each aptamer from the plurality of aptamers in the initial pool, data specifying secondary structure of the respective aptamer.
7. The method of claim 6 , wherein the data specifying the secondary structure comprises data specifying a pattern of hydrogen bonds in the respective aptamer or an energy dot plot matrix of the respective aptamer.
8. The method of claim 1 , wherein the neural network is configured process, for the particular aptamer, an encoding of data defining a sequence for the particular aptamer, the encoding comprising a plurality of one-hot vectors, each one-hot vector corresponding to a respective nucleobase in the particular aptamer.
9. The method of claim 2 , wherein the first SELEX round is before a second SELEX round in the iterative SELEX process.
10. A system comprising one or more computers and one or more storage devices storing instructions that when executed by one or more computers cause the one or more computers to perform operations for training a neural network to model results of an iterative SELEX process, wherein the iterative SELEX process begins with an initial pool of aptamers and iteratively updates the pool of aptamers at each of a plurality of different rounds, the operations comprising:
obtaining results of the iterative SELEX process applied to the initial pool of aptamers;
obtaining data defining a plurality of sequences for a plurality of aptamers in the initial pool of aptamers, each aptamer from the plurality of aptamers comprising a string of nucleobases;
generating, based on the results of the iterative SELEX process, a respective label for each sequence from the plurality of sequences that comprises characteristics of how strongly the respective aptamer binds to at least one target molecule, wherein the respective label for each sequence from the plurality of sequences includes a plurality of sub-labels and each sub-label corresponds to a different round in the iterative SELEX process and characterizes a result of the corresponding round for the respective aptamer;
generating training data for a neural network from the data defining the plurality of sequences and the respective labels for each of the sequences; and
training the neural network on the training data, the neural network configured to generate, for a particular aptamer, a model output that comprises a plurality neural network outputs, each neural network output corresponding to a different round in the iterative SELEX process, and, wherein:
each neural network output is a prediction of a result of the corresponding SELEX round, if the particular aptamer were in the initial pool of aptamers as of the beginning of the corresponding SELEX round, and
the neural network output corresponding to a first SELEX round in the iterative SELEX process comprises a count, a binary value or a confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.
11. The system of claim 10 , wherein the neural network output corresponding to the first SELEX round in the iterative SELEX process comprises the binary value or the confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.
12. The system of claim 10 , wherein the count is a numerical value that represents a predicted count of the particular aptamer in the pool of aptamers for the first SELEX round.
13. The system of claim 10 , wherein the model output comprises a vector of fluorescence levels, where a brightness of each fluorescence level indicates how strongly the particular aptamer binds to the at least one target molecule.
14. The system of claim 10 , wherein the neural network is a convolutional neural network.
15. The system of claim 10 , wherein the data defining the plurality of sequences comprises, for each aptamer from the plurality of aptamers in the initial pool, data specifying secondary structure of the respective aptamer.
16. The system of claim 15 , wherein the data specifying the secondary structure comprises data specifying a pattern of hydrogen bonds in the respective aptamer or an energy dot plot matrix of the respective aptamer.
17. The system of claim 10 , wherein the neural network is configured process, for the particular aptamer, an encoding of data defining a sequence for the particular aptamer, the encoding comprising a plurality of one-hot vectors, each one-hot vector corresponding to a respective nucleobase in the particular aptamer.
18. The system of claim 10 , wherein the first SELEX round is before a second SELEX round in the iterative SELEX process.
19. A system comprising:
one or more computers; and
one or more storage devices storing instructions, which, when executed by the one or more computers, cause the one or more computers to perform operations for modeling, for a particular aptamer and at least one target molecule, the results of an iterative SELEX process that begins with an initial pool of aptamers and iteratively updates the initial pool of aptamers at each of a plurality of rounds, the operations comprising:
obtaining data defining a sequence for the particular aptamer, the aptamer comprising a string of nucleobases;
encoding the data defining the sequence for the particular aptamer as a neural network input; and
processing the neural network input using a trained neural network to generate an output that comprises a plurality of neural network outputs, wherein:
each neural network output corresponds to a different one of the plurality of rounds in the iterative SELEX process, and
each neural network output is a prediction of a result of the corresponding SELEX round, if the particular aptamer were in the initial pool of aptamers as of the beginning of the corresponding SELEX round, wherein the neural network output corresponding to a first SELEX round in the iterative SELEX process comprises a characterization of how strongly the particular aptamer would bind to a particular target molecule in the first SELEX round if the particular aptamer were in the initial pool of aptamers as of the beginning of the first SELEX round.
20. The system of claim 19 , wherein the neural network output corresponding to the first SELEX round in the iterative SELEX process comprises a binary value or a confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.
21. The system of claim 20 , wherein the neural network output corresponding to a second SELEX round in the iterative SELEX process comprises a numerical value that represents predicted count of the particular aptamer in the pool of aptamers for the second SELEX round, and
the second SELEX round is after the first SELEX round in the iterative SELEX process.
22. The system of claim 19 , wherein the output comprises a vector of fluorescence levels, where a brightness of each fluorescence level indicates how strongly the particular aptamer binds to the at least one target molecule.
23. The system of claim 19 , wherein the neural network is a convolutional neural network.
24. The system of claim 19 , wherein the data defining the sequence comprises data specifying secondary structure of the particular aptamer.
25. The system of claim 24 , wherein the data specifying the secondary structure comprises data specifying a pattern of hydrogen bonds in the particular aptamer or an energy dot plot matrix of the particular aptamer.
26. The system of claim 19 , wherein encoding the data defining the sequence comprises converting the data defining the sequence to a plurality of one-hot vectors, each one-hot vector corresponding to a respective nucleobase in the particular aptamer.
27. One or more non-transitory computer-readable media storing instructions, which, when executed by one or more computers, cause the one or more computers to perform operations for modeling, for a particular aptamer and at least one target molecule, the results of an iterative SELEX process that begins with an initial pool of aptamers and iteratively updates the initial pool of aptamers at each of a plurality of rounds, the operations comprising:
obtaining data defining a sequence for the particular aptamer, the aptamer comprising a string of nucleobases;
encoding the data defining the sequence for the particular aptamer as a neural network input; and
processing the neural network input using a trained neural network to generate an output that comprises a plurality of neural network outputs, wherein:
each neural network output corresponds to a different one of the plurality of rounds in the iterative SELEX process, and
each neural network output is a prediction of a result of the corresponding SELEX round, if the particular aptamer were in the initial pool of aptamers as of the beginning of the corresponding SELEX round, wherein the neural network output corresponding to a first SELEX round in the iterative SELEX process comprises a characterization of how strongly the particular aptamer would bind to a particular target molecule in the first SELEX round if the particular aptamer were in the initial pool of aptamers as of the beginning of the first SELEX round.
28. The non-transitory computer-readable media of claim 27 , wherein the neural network output corresponding to the first SELEX round in the iterative SELEX process comprises a binary value or a confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.
29. The non-transitory computer-readable media of claim 28 , wherein the neural network output corresponding to a second SELEX round in the iterative SELEX process comprises a numerical value that represents predicted count of the particular aptamer in the pool of aptamers for the second SELEX round, and
the second SELEX round is after the first SELEX round in the iterative SELEX process.
30. One or more non-transitory computer-readable storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations for training a neural network to model results of an iterative SELEX process, wherein the iterative SELEX process begins with an initial pool of aptamers and iteratively updates the pool of aptamers at each of a plurality of different rounds, the operations comprising:
obtaining results of the iterative SELEX process applied to the initial pool of aptamers;
obtaining data defining a plurality of sequences for a plurality of aptamers in the initial pool of aptamers, each aptamer from the plurality of aptamers comprising a string of nucleobases;
generating, based on the results of the iterative SELEX process, a respective label for each sequence from the plurality of sequences that comprises characteristics of how strongly the respective aptamer binds to at least one target molecule, wherein the respective label for each sequence from the plurality of sequences includes a plurality of sub-labels and each sub-label corresponds to a different round in the iterative SELEX process and characterizes a result of the corresponding round for the respective aptamer;
generating training data for a neural network from the data defining the plurality of sequences and the respective labels for each of the sequences; and
training the neural network on the training data, the neural network configured to generate, for a particular aptamer, a model output that comprises a plurality neural network outputs, each neural network output corresponding to a different round in the iterative SELEX process, and, wherein:
each neural network output is a prediction of a result of the corresponding SELEX round, if the particular aptamer were in the initial pool of aptamers as of the beginning of the corresponding SELEX round, and
the neural network output corresponding to a first SELEX round in the iterative SELEX process comprises a count, a binary value or a confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.
31. The one or more non-transitory computer-readable storage media of claim 30 , wherein the neural network output corresponding to the first SELEX round in the iterative SELEX process comprises the binary value or the confidence value indicating whether the particular aptamer would exist in the pool of aptamers at completion of the first SELEX round.