# Can pair end reads with high heterozygosity be used to polish PacBio assembly?

I have used canu (correct) and smartdenovo to assembly PacBio reads. Next I am going to polish my assembly using illumina pair end reads by Pilon. However, I found there is high heterozygosity in my pair end reads when I used jellyfish to do k-mer analysis. There are two peaks in the k-mer frequencies distribution plot. The size of my genome is about 190M, and it is a diploid organism.
What I am doubting is that can I use this pair end data to polish my assembly?

If not, I have to do sequencing again.

• In general polishing using short reads is depreciated because of mapping problems. If you have sufficient coverage, I would just stick to arrow. Mar 7, 2018 at 15:10

The answer, yes, you can use Pilon + short reads to correct the PacBio assembly even with a highly heterozygous sample. To do so you should use the --diploid option in Pilon to let it know that there will be heterozygous sites. The way that it corrects heterozygous SNPs and heterozygous indels is different, and it seems to perform best with SNPs and sometimes does not correctly identify heterozygous indels (source).
Pilon itself doesn't handle phasing, but just corrects with whatever reads are in the bam file that it is provided with. The end result of providing a bam with reads from both haplotypes is that the corrected sequence will be a mosaic of the dominant allele at the heterozygous site as was observed in the short read data. This may or may not select for some alleles based on sequencing coverage bias of Illumina sequencers.