| Gene sequencing,as a new type of genetic testing work,plays an important role in predicting diseases and explaining the rationality of individual behavior.It is one of the important ways for humans to explore the mysteries of life,greatly promoting the development of biological and medical fields.Gene sequencing technology was introduced in the 1970s and has undergone three generations of improvement and development.The first generation sequencing technology,Sanger sequencing,accuracy rate of which is high,is used as the gold standard for genetic testing.However,its throughput is too low and cost is high.With the development of science and technology,the second generation sequencing technology based on reversible termination and simultaneous synthesis has developed.However,due to the use of PCR amplification,the nucleotide sequences obtained by sequencing are too short,which is not conducive to subsequent genome splicing.Today,the third generation sequencing technology represented by Oxford Nanopore Sequencing Technology has been developed.Compared to the previous two generations of sequencing technology,the third generation of sequencing technology has many advantages,including the need not to use PCR amplification which could lead to a deviation of GC in downstreaming analysis and obtaining longer nucleotide sequences that cover repetitive regions of the genome.Due to the advantages of long length,fast speed of sequencing,and portability,Nanopore Sequencing plays an important role in biological research.Sequencing data is widely used in downstream analysis such as Modification Detection,especially Methylation Detection,Genome Splicing,whole field transcript detection,and detection of single nucleotide polymorphism.In addition,Nanopore Sequencing is also widely used to detect large structural variations in the biomedical context,describe the whole field transcriptome and complex transcriptional events,characterize epigenetics,identify genomic variations of interest in cancer research,and be used for infectious and genetic disease research and epidemic monitoring,which is of great significance.The key to Nanopore Sequencing is to identify the corresponding bases of the original electrical signals which is called basecalling,and the quality of the data seriously affects subsequent downstream analysis.However,today’s base recognition tools are mostly focused on Oxford Nanopore 5-meror 6-mer data,without processing tools for domestic Nanopore 4-mer data.Therefore,in response to this situation,we have conducted a.complete process of basecaller and creatively introduced denoising technology based on wavelet decomposition to preprocess the original electrical signal sequence.Firstly,we use DeepSimulatorl.5 to generate the simulated 4-mer Nanopore signals.In order to obtain the standard dataset,we need to label the original electrical signals.The main issue in this regard is mapping end-to-end sequences.We compared two different mapping methods and ultimately choose the alignment method cwDTW based on electrical signal space,which is also the first time it has been applied to data labeling of electrical signals.After obtaining the base level mapping between the electrical signals and the corresponding nucleotide sequences,we construct the benchmark.In addition,we divids the benchmark into the training set and several test sets with different average lengths,respectively,for training the model and verifying its effectiveness.Afterwards,the training model in Bonito is introduced,which combines a one-dimensional time channel separable convolutional neural network structure with a connectionist time classifier to calculate losses for training and decoding the predicted output sequence.Afterwards,we creatively introduced a denoising method based on wavelet transform to preprocess the original electrical signals,and the results show that it can improve the stability of the model to a certain extent. |
