Official implementation of our Nano Letters paper Expanding the Molecular Alphabet of DNA-Based Data Storage Systems with Nanopore Readouts.
Programming language: Python 3.7. Tested on operating systems: Windows 10, CentOS 7.7.1908.
Our method can be divided into two components: extracting signals of interest from raw fast5 files, and perform neural network classification on time series.
The ONT (Oxford Nanopore Technologies) raw fast5 files for all 77 tetramers are stored in [link]. We also store the final results after signal extraction step as npy files in [link].
The overall pipeline of this component is shown in the following figure. Please refer to Figure S6 for more details.
To use the code, you need to first download the fast5 files into some directory and set the following data path accordingly.
python signal_extraction.py --data_path="raw_fast5" --out_dir="processed_files/extracted_signals"
python build_dataset.py --data_path="processed_files/extracted_signals" --pool_path="raw_fast5"
--out_path="processed_files/dataset" --std_thres_ub=10.0
With the built dataset, we are ready to perform classification with deep learning to achieve high classification accuracy.
python main_allclass_convplus.py --data_path="processed_files/dataset/"
--lr=1e-3 --test_ratio=0.4 --batch_size=256
--max_epoch=200 --early_stop=25
--chkpt="./chkpt/convplus/allclass_1e-3"
--max_sample=3500 --num_trails=1
Confusion Matrix
python main_usecase_resnet1d.py --data_path="processed_files/dataset/"
--lr=1e-3 --test_ratio=0.4 --batch_size=64
--max_epoch=400 --early_stop=50
--chkpt="./chkpt/resnet1d/U2_1e-3"
--max_sample=5000 --num_trails=1
--usecase_short="U2" --model_param="[64,4,1,2]"
Confusion Matrix
