Fig. 5

Error types and number of haplotigs for nPhase and WhatsHap polyphase. nPhase and WhatsHap polyphase were both applied to our 20X test datasets of different ploidy and heterozygosity levels. nPhase was tested using its default parameters and WhatsHap polyphase was tested with all six possible values of its adjustable sensitivity parameter. This graph compares both tools using the following metrics: average number of haplotigs obtained for the genome, normalized by the ploidy, average accuracy, average error rate, and average missing rate. a Average accuracy, error, and missing rates for all tests using nPhase and WhatsHap polyphase on different sensitivity levels. The error rate for WhatsHap polyphase increases dramatically as the sensitivity level decreases, illustrating the tool’s tradeoff between accuracy and contiguity. b Average number of haplotigs per chromosome per haplotype for all tests using nPhase and WhatsHap polyphase on different sensitivity levels. The very high number of haplotigs per chromosome per haplotype for the highest sensitivity levels (5 and 4) shows that despite being highly accurate, they are not contiguous enough to be informative. Based on our results, nPhase outperforms WhatsHap polyphase in all of our tests. The tradeoff between accuracy and contiguity is extreme in WhatsHap polyphase, either the results are very accurate but so fragmented as to be uninformative, or they are about as contiguous as nPhase but less than 65% accurate