Real Time Nanopore Sequencing at Scale - Read Until and Run Until on PromethION scale devices.

核酸的快速测序改变了我们对生物世界的理解。新技术和新方法为开发解决生物学问题的最佳方法提供了新的挑战。只有牛津纳米孔技术（ONT）仪器能够实时测序，序列数据可以在数据生成过程中进行分析。纳米孔测序的工作原理是测量DNA通过小纳米孔时的电流。电流依赖于存在的DNA序列，因此产生可解码的电流轨迹，从而揭示DNA序列。我们已经以多种方式利用了这项技术，特别是建立了“read until”。在这里，分子的起始部分在被测序的同时被分析。如果分子与感兴趣的区域匹配，测序就可以继续。如果不是，特定通道上的电压可以反转，读取被拒绝。原则上，这可以从库中对特定分子进行选择性测序。我们是第一个证明这种方法是有效的，使用动态时间扭曲来映射纳米孔弯曲到一个参考。随后，通过安大略省和诺丁汉大学之间的iCASE学生合作，我们共同开发了一种方法，可以在测序时直接从卷曲轨迹中获取碱基序列数据。这种方法消除了在信号中搜索的限制，并且可以很容易地被具有相同硬件访问权限的其他人使用。目前，该方法仅限于特定亚型的ONT测序仪（GridION和MinION）。这些平台具有每个flowcell生成10- 20gb数据的能力。PromethION测序仪可以同时从多达48个位置生成超过100 Gb的序列数据。这使得人类基因组测序在60小时内的单个流式细胞长读取。原则上，这种规模的选择性测序可以实现许多有用的应用，例如低覆盖率的全基因组加上更高覆盖率的特定区域，如已知的致癌SNPs或其他感兴趣的结构变异。然而，“read until”会对液流池的性能产生负面影响（特别是对PromethION），因此实验设计需要根据精确的实验目标进行优化。我们正在与ONT合作，以确定造成这种情况的潜在原因，并开发在更高吞吐量的平台上实现“Read Until”的方法。在本研究中，我们将开发两种实时纳米孔测序方法。1) Run Until，一旦达到特定的实验目标，测序就可以自动停止；2)将这种迭代类型的方法与“read Until”相结合，这样一旦达到特定的目标，就可以拒绝不再需要的读取。方法将包括基因组组装，全基因组区域的靶向覆盖，以及宏基因组的过滤。我们还将寻求建立可以同时或依次利用一个或多个测序位置的方法。至关重要的是，这些方法必须是动态的，并且在定制库准备之上提供灵活性，这可能会丰富特定目标。

Rapid sequencing of nucleic acids transformed our understanding of the biological world. New technologies and methods provide new challenges developing optimal approaches to address biological questions. Only Oxford Nanopore Technologies (ONT) instruments enable real time sequencing where sequence data can be analysed during data generation. Nanopore sequencing works by measuring current flow as DNA passes through a small nanopore. Current flow depends on the sequence of DNA present and so generates a current trace which can be decoded revealing the DNA sequence.We have exploited this technology in a number of ways, notably establishing "read until". Here the start of a molecule is analysed whilst being sequenced. If the molecule matches a region of interest, sequencing can continue. If not the voltage over the specific channel can be reversed and the read rejected. In principle, this enables selective sequencing of specific molecules from a library. We were the first to demonstrate that this method works using dynamic time warping to map the nanopore squiggle to a reference. Subsequently, working through an iCASE studentship between ONT and Nottingham, we have jointly developed an approach to basecall sequence data directly from the squiggle trace whilst sequencing. This method removes limitations of searching within signals and can be readily used by others with access to the same hardware. At present, the method is limited to a specific subtype of ONT sequencers (the GridION and MinION). These platforms have the capacity to generate 10-20 Gb of data per flowcell. The PromethION sequencer can generate in excess of 100 Gb of sequence data per flowcell from up to 48 positions simultaneously. This enables sequencing of a human genome within 60 hours on a single flowcell with long reads. In principle, selective sequencing at this scale could enable many useful applications such as a low coverage whole genome coupled with higher coverage of specific regions such as known cancer-causing SNPs or other structural variants of interest.However, "read until" has a negative impact on the flowcell performance (particularly on the PromethION) and so experimental design requires optimisation depending on the precise experimental goal. We are working with ONT to determine the underlying reasons for this and develop methods to implement "Read Until" on higher throughput platforms. In this studentship we will develop methods to implement two real time approaches around nanopore sequencing. 1) Run Until, whereby sequencing can be automatically halted once a specific experimental goal has been achieved and, 2), coupling this iterative type of approach with 'read until' such that once a specific goal has been achieved, reads can be rejected which are no longer required. Methods will include genome assembly, targeted coverage of regions of whole genomes, and filtering of metagenomes. We will also look to establish methods that can exploit one or more sequencing positions simultaneously or sequentially. Crucially, these methods must be dynamic and offer flexibility over and above custom library preparation which might enrich for specific targets.