Towards Finished Genome Assemblies using SMRT® Sequencing

Plant and animal genomes are replete with repetitive sequences that pose challenges for short-read-based assemblies. By using reads greater than 5 kb, SMRT^® Sequencing can span many of these repeats, which allows the joining of contigs that would otherwise be split at a repeat, thereby improving assembly continuity. Two recent algorithm advances increase the utility of SMRT Sequencing data. First, the hierarchical genome assembly (HGA) process reduces library requirements to a single 10 kb or longer SMRTbell™ library by simultaneously utilizing the information from both longer and shorter reads. The approach enables a more efficient de novo assembly by performing a pre-assembly step that generates highly accurate, consensus-based reads that can be used as input for existing genome assemblers. Second, the Quiver consensus caller is a hidden Markov model-based method that can be used to polish assemblies, resulting in per-base accuracy of >QV50. By modeling random insertions and deletions, which are more common in single molecule sequencing, the algorithm can estimate near-perfect consensus sequences. We demonstrate these methods in a number of examples.