8
$\begingroup$
  1. Is there a standard definition for "assembly polishing" in the field?

  2. Is there a standard definition for what polishing algorithms do?

My understanding of "polishing" is strongly influenced by Pilon:

Whereby users input a reference FASTA along with the BAM of aligned reads against this reference FASTA. The polishing algorithm then does alignment to discover local misassemblies and other alignment inconsistencies. These are then "corrected", i.e. "polished".

$\endgroup$
1
  • 2
    $\begingroup$ Anything that makes something more correct (or shinier) would be polishing it in English, there's no special meaning here. $\endgroup$
    – Devon Ryan
    May 15 '18 at 20:29
5
$\begingroup$

In long-read assembly, "polish" refers to the step to improve the base accuracy of contig sequences. I believe the terminology was originated from the HGAP paper:

The final consensus–calling algorithm Quiver, which takes into account all of the underlying data and the raw quality values inherent to SMRT sequencing, then polishes the assembly for final consensus accuracies in excess of 99.999% (QV of >50)...

Pilon was published more than a year later and in fact, neither its paper nor README mentions word "polish".

By the way, "error correction" refers to the step to fix sequencing errors in raw reads, not in contigs. In Canu/HGAP/Falcon, that step is done by pbdagcon or falcon_sense.

$\endgroup$
1
  • $\begingroup$ Cool I didn't know that! $\endgroup$
    – Kamil S Jaron
    May 16 '18 at 6:21
4
$\begingroup$

Disclaimer, this answer is based on feeling I got from talks and papers, but I do not have any hard reference supporting it.

I believe that genome polishing is a technique that was introduced for correction of individual bases of long noisy read assemblies - i.e. fixing SNPs and short indels. Guessing a correct base from long reads during assembly seems to be bit more complicated task in comparison to short read assemblies, therefore it is isolated into separated algorithms like Quiver/Arrow, Nanopolish (for correction by long reads) or the tool you mentioned - Pilon (correction by short reads).

Miss-assemblies could be corrected even prior long reads for instance by matching the assembly to optical mapping, or by tools like REAPR or TGnet. However, I have never heard term "genome polishing" used for this type of genome correction.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.