Nanopore sequencing base calling
View Patent ↗Disclosed herein are systems and methods for nanopore sequencing basecalling. In one embodiment, the method can include: receiving raw nanopore sequencing data comprising a plurality of continuous data acquisition (DAC) values corresponding to a biomolecule; normalizing the raw nanopore sequencing data to generate normalized nanopore sequencing data comprising a plurality of normalized DAC values; generating, using a first neural network (NN) and a normalized DAC value of the plurality of normalized DAC values, a vector of transformed probability values, segmenting the plurality of normalized DAC values into a plurality of discrete events; generating, using a second neural network and the event vector, an element determination of the biomolecule.
1. A system for nanopore sequencing basecalling, comprising:
a membrane defining a pore wherethrough a biomolecule is threaded to generate raw nanopore sequencing data;
a computer-readable memory storing executable instructions; and
one or more hardware-based processors programmed by the executable instructions to perform steps comprising:
receiving the raw nanopore sequencing data comprising a plurality of continuous data acquisition (DAC) values corresponding to the biomolecule;
normalizing the raw nanopore sequencing data to generate normalized nanopore sequencing data comprising a plurality of normalized DAC values in a streaming manner;
generating a vector of transformed probability values using a first neural network (NN) and a normalized DAC value of the plurality of normalized DAC values, wherein the first NN comprises a first recurrent neural network (RNN), wherein the first RNN comprises a first plurality of long short term memory (LSTM) layers, wherein a LSTM layer of the first plurality of LSTM layers comprises a first unidirectional LSTM layer, wherein the first unidirectional LSTM is followed by one or more fully connected layers; and
determining a base in the biomolecule using a second NN and the generated vector of transformed probability vectors, wherein the second NN comprises a second RNN, wherein the second RNN comprises a second plurality of LSTM layers, and wherein a LSTM layer of the second plurality of LSTM layers comprises a second unidirectional LSTM layer, wherein the second unidirectional LSTM is followed by one or more fully connected layers.
2. The system of claim 1 ,
wherein a position of the vector corresponds to an oligonucleotide, and
wherein a transformed probability value, of the transformed probability values of the vector, at the position of the vector comprises a probability of the oligonucleotide being represented by the DAC value.
3. The system of claim 2 , wherein the oligonucleotide comprises a 4-base oligonucleotide.
4. The system of claim 1 , wherein normalizing the raw nanopore sequencing data comprises using median normalization of the plurality of DAC values to generate the normalized nanopore sequencing data.
5. The system of claim 1 ,
wherein a DAC value of the plurality of DAC values corresponds to a read start DAC value,
wherein the raw nanopore sequencing data comprises a second plurality of DAC values upstream of the read start DAC value, and
wherein normalizing the raw nanopore sequencing data comprises using one or more second DAC values of the second plurality of DAC values upstream of the read start DAC value to generate the normalized nanopore sequencing data.
6. The system of claim 1 , wherein the first plurality of LSTM layers comprises three LSTM layers.
7. The system of claim 6 , wherein the three LSTM layers each comprise 75, 100, and 50 nodes, respectively.
8. The system of claim 1 , wherein the first plurality of LSTM layers is followed by one or more fully connected layers.
9. The system of claim 1 , wherein the method further comprises:
generating a second vector of transformed probability values using the first NN, a second normalized DAC value of the plurality of normalized DAC values, and a memory value.
10. The system of claim 1 , wherein the vector of transformed probability values comprises a vector of logits of transformed probability values.
11. The system of claim 1 , wherein generating the vector of transformed probability values comprises: generating, using the first NN and the normalized DAC value of the plurality of normalized DAC values, the vector of transformed probability values and a memory value.
12. The system of claim 1 , wherein generating the vector of transformed probability values comprises: generating, using the first NN and the normalized DAC value of the plurality of normalized DAC values, the vector of transformed probability values at a delay.
13. The system of claim 12 , wherein the delay comprises a duration corresponding to a number of DAC values of the plurality of DAC values, and wherein the number of DAC values of the plurality of DAC values comprises 33 DAC values of the plurality of DAC values.
14. The system of claim 1 , wherein the method further comprises:
segmenting the plurality of normalized DAC values into a plurality of discrete events.
15. The system of claim 1 , wherein the base corresponds to two base insertions or single deletion.
16. The system of claim 1 , wherein the base comprises a non-standard base.
17. The system of claim 16 , wherein the method further comprises:
determining the non-standard base using a classifier.
18. The system of claim 17 , wherein the classifier comprises a random forest.
19. The system of claim 17 , wherein determining the non-standard base using the classifier comprises using DAC values of the plurality of DAC values within an event.
20. The system of claim 1 , further comprising training the first NN and the second NN, wherein the training the first NN comprises training the first LSTM layer of the first plurality of LSTM layers, followed by training the second LSTM layer of the first plurality of LSTM layers, and the training the second NN comprises training the first LSTM layer of the second plurality of LSTM layers, followed by training the second LSTM layer of the second plurality of LSTM layers, wherein the training the first NN and the second NN further comprises obtaining a training data set using a HMM-based basecaller, wherein obtaining the training data set comprises mapping a long read and aligning a raw electric current signal through a called event to a genome.
21. The system of claim 20 , wherein the training data set is based on a cDNA.
22. The system of claim 20 , wherein the training data set is based on a cDNA from Escherichia coli.
23. A method of nanopore sequence basecalling, comprising:
providing a membrane defining a pore wherethrough a biomolecule is threaded to generate raw nanopore sequencing data;
receiving the raw nanopore sequencing data comprising a plurality of continuous data acquisition (DAC) values corresponding to the biomolecule;
normalizing the raw nanopore sequencing data to generate normalized nanopore sequencing data comprising a plurality of normalized DAC values in a streaming manner;
generating a vector of transformed probability values using a first neural network (NN), and a normalized DAC value of the plurality of normalized DAC values, wherein the first NN comprises a first recurrent neural network (RNN), wherein the first RNN comprises a first plurality of long short term memory (LSTM) layers, wherein a LSTM layer of the first plurality of LSTM layers comprises a first unidirectional LSTM layer, wherein the first unidirectional LSTM is followed by one or more fully connected layers; and
determining a base in the biomolecule using a second NN and the generated vector of transformed probability vectors, wherein the second NN comprises a second RNN, wherein the second RNN comprises a second plurality of long short term memory (LSTM) layers, and wherein a LSTM layer of the second plurality of LSTM layers comprises a second unidirectional LSTM layer, wherein the second unidirectional LSTM is followed by one or more fully connected layers.
24. The method of claim 23 , further comprising: segmenting the plurality of normalized DAC values into a plurality of discrete events,
wherein a position of the vector corresponds to an oligonucleotide, wherein a transformed probability value, of the transformed probability values of the vector, at the position of the vector comprises a probability of the oligonucleotide being represented by the DAC value,
wherein a discrete event is associated with an event vector of averaged transformed probability values, wherein a position of the event vector corresponds to the oligonucleotide, and wherein an averaged transformed probability of the oligonucleotide comprises an averaged probability of the oligonucleotide being represented by DAC values of the plurality of DAC values within the event.