What is more, the Rep-accuracy and genome coverage were increasing a little with the increasing of read length, which indicated the completeness of assembled repeats and assembly redundant were increasing simultaneously.
For all datasets, GeneStitch significantly improves the completeness of assembled genes as compared with initial assembly's genes (with higher gene coverage), and the number of complete genes, especially for the datasets with lower sequencing depths (6× or 13×).
Consequently, the completeness and accuracy of assembling repeats are what SWA is mainly concerned about the completeness of assembled repeats and nonrepeats rather than the continuity of whole genome assembly.
For this purpose three transcriptome assemblers, Trinity, Velvet/OASES and SOAPdenovo-Trans, were used to maximise diversity and completeness of assembled transcripts.
Rep-accuracy and N50 were a little down but still good; when depth dropped to 2 or 1, TRC and NNC were increasing, while N50, Max, and Rep-accuracy were decreasing; meanwhile the completeness of assembled repeats is getting worse.
For comparisons in real NGS datasets, the metrics such as TRC, NNC, and Rep-size are used to evaluate the completeness of assembled repeats and nonrepeats; the metrics such as Rep-accuracy, C-accuracy, and CN-accuracy are used to evaluate the accuracy of assembled repeats and nonrepeats, while the N50, N90, and maximum contig are used to evaluate the continuity of whole genome assembly.
Results indicated that SWA outperformed other leading assemblers in the completeness and correctness of assembling repeats, but the continuity was not better than some of the others.
Finally, the extensive comparisons with other eight leading assemblers show that SWA outperformed others in terms of completeness and correctness of assembling repeats and nonrepeats.
But for the repeats longer than read length, most of them performed poorly in completeness and accuracy of assembling repeats from NGS data.
