DeepNano-blitz: a fast base caller for MinION nanopore sequencers

Abstract

Motivation Oxford Nanopore MinION is a portable DNA sequencer that is marketed as a device that can be deployed anywhere. Current base callers, however, require a powerful GPU to analyze data produced by MinION in real time, which hampers field applications. Results We have developed a fast base caller DeepNano-blitz that can analyze stream from up to two MinION runs in real time using a common laptop CPU (i7-7700HQ), with no GPU requirements. The base caller settings allow trading accuracy for speed and the results can be used for real time run monitoring (i.e. sample composition, barcode balance, species identification, etc.) or prefiltering of results for more detailed analysis (i.e. filtering out human DNA from human–pathogen runs). Availability and implementation DeepNano-blitz has been developed and tested on Linux and Intel processors and is available under MIT license at https://github.com/fmfi-compbio/deepnano-blitz. Supplementary information Supplementary data are available at Bioinformatics online.

Pair consensus decoding improves accuracy of neural network basecallers for nanopore sequencing

“…Our software can readily be adapted to work with the output of other neural network basecallers. Application to the recent DeepNano-blitz [11], showed a similar gain in accuracy from consensus decoding. We also applied our algorithm to the ONT basecaller Bonito [12], a research basecaller inspired by recent successes of purely convolutional neural networks in speech recognition, and compared results with Guppy, an earlier ONT basecaller which can make use of 1D 2 .…”

Processing Oxford Nanopore Long Reads Using Amazon Web Services

“…Наиболее часто используемым и эффективным программным пакетом для процедуры перевода сигналов нанопорового секвенатора в нуклеотидные последовательности является guppy_basecaller. Часть конкурирующих програмных продуктов либо не поддерживается производителем («Albacore Oxford Nanopore Basecaller», Великобритания), либо обладает более низкими эксплуатационными характеристиками [9]. Точность секвенирования при использовании расширенных нейросетевых моделей в guppy_basecaller достигает 99.73%.…”

Nanopore base calling on the edge

“…Guppy, a base caller provided by ONT, is based on recurrent neural networks (RNN) and provides two different architectures: a fast base caller, which can base call with 85–92% median read accuracy in real time, using recent GPU cards and a high-accuracy base caller (90–96% median read accuracy), which is too slow to be used in real time without specialized setup. DeepNano-blitz trades off a bit of accuracy in order to provide real-time base calling on a common CPU using a specifically engineered RNN, thus obviating the need for GPUs ( Boza et al , 2020 ). Other RNN-based base callers, including Chiron ( Teng et al , 2018 ), are too slow for real-time base calling.…”

“…Our new base caller DeepNano-coral provides real-time base calling that is significantly more energy efficient than existing approaches running on GPU (e.g. Guppy base caller) or CPU ( Boza et al , 2020 ). These advantages, coupled with easy availability and a low price of Coral devices, make this a practical solution for the problem of real-time base calling.…”

“…Our experiments show that DeepNano-coral achieves the accuracy comparable to other real-time base callers, Guppy fast and DeepNano-blitz. Such accuracy is sufficient for real-time monitoring tasks, such as monitoring barcode composition in pooled libraries or species composition in environmental or clinical samples ( Boza et al , 2020 ). DeepNano-coral achieves this goal much more energy efficiently, using only 0.6–0.7 Wh of energy to base call a test sample of 40.8 Mb of nanopore sequences at a speed of 1.54 million signal samples per second (on the same setup, the closest competitor, Guppy fast, uses 1.4 Wh of energy, processes 4.37 million signal samples per second, with up to 2 percentage points lower accuracy depending on the dataset).…”