摘要：SummarySingle-molecule sequencing technologies produce much longer reads compared with next-generation sequencing, greatly improving the contiguity ofde novoassembly of genomes. However, the relatively high error rates in long reads make it challenging to obtain high-quality assemblies. A computationally intensive consensus step is needed to resolve the discrepancies in the reads. Efficient consensus tools have emerged in the recent past, based on partial-order alignment. In this study, we discovered that the spatial relationship of alignment pileup is crucial to high-quality consensus and developed a deep learning-based consensus tool, CONNET, which outperforms the fastest tools in terms of both accuracy and speed. We tested CONNET using a 90× dataset ofE. coliand a 37× human dataset. In addition to achieving high-quality consensus results, CONNET is capable of delivering phased diploid genome consensus. Diploid consensus on the above-mentioned human assembly further reduced 12% of the consensus errors made in the haploid results.Graphical AbstractDisplay OmittedHighlights•Deep learning methods outperform existing approaches in assembly consensus•Spatial relationships in alignment pileup are crucial to high-quality consensus•Diploid consensus can further reduce errors made in haploid consensus•CONNET can be used for both consensus and polishingGenomics; Bioinformatics; Sequence Analysis
