The high-throughput short-reads RNA-seq protocols often produce paired-end reads, with the middle portion of the fragments being unsequenced. We explore if the full-length fragments can be computationally reconstructed from the sequenced two ends in the absence of the reference genome - a problem here we refer to as de novo bridging. Solving this problem provides longer, more informative RNA-seq reads, and benefits downstream RNA-seq analysis such as transcript assembly, expression quantification, and splicing differential analysis. However, de novo bridging is a challenging and complicated task owing to alternative splicing, transcript noises, and sequencing errors. It remains unclear if the data provides sufficient information for accurate bridging, let alone efficient algorithms that determine the true bridges. Methods have been proposed to bridge paired-end reads in the presence of reference genome (called reference-based bridging), but the algorithms are far away from scaling for de novo bridging as the underlying compacted de Bruijn graph(cdBG) used in the latter task often contains millions of vertices and edges. We designed a new truncated Dijkstra's algorithm for this problem, and proposed a novel algorithm that reuses the shortest path tree to avoid running the truncated Dijkstra's algorithm from scratch for all vertices for further speeding up. These innovative techniques result in scalable algorithms that can bridge all paired-end reads in a cdBG with millions of vertices. Our experiments showed that paired-end RNA-seq reads can be accurately bridged to a large extent. The resulting tool is freely available at https://github.com/Shao-Group/rnabridge-denovo.

翻译：高通量短读长RNA-seq技术常产生双端测序读段，但片段中间部分未被测序。我们探究在无参考基因组的情况下，能否通过计算从测序的两个末端重建全长片段——这一问题在此称为"从头桥接"。解决此问题可提供更长、更富含信息的RNA-seq读段，并惠及下游RNA-seq分析，如转录本组装、表达量定量及剪接差异分析。然而，由于选择性剪接、转录噪声和测序错误的存在，从头桥接是一项具有挑战性的复杂任务。目前尚不清楚数据是否为准确桥接提供了充分信息，更遑论存在能确定真实桥接的高效算法。虽有方法提出在存在参考基因组时进行双端读段桥接（称为参考基因组桥接），但其算法远未达到可扩展至从头桥接的水平，因为后者依赖的紧凑型德布鲁因图（cdBG）常包含数百万个顶点和边。我们针对该问题设计了一种新的截断Dijkstra算法，并提出了一种创新算法——通过复用最短路径树避免对所有顶点从头运行截断Dijkstra算法以进一步加速。这些创新技术实现了可扩展的算法，能桥接含数百万顶点的cdBG中的所有双端读段。实验表明，双端RNA-seq读段可在很大程度上被准确桥接。该工具可免费获取于https://github.com/Shao-Group/rnabridge-denovo。