Question: window-length pros and cons

[JohnUrban]

Hi,

RaCon is great. I was wondering if there were pros and cons to the --window-length parameter, which is 500bp by default. I tested 5kb -- so I know it gets a lot slower with length, but I was hoping there would also be some arguments that favor using a longer window such as 5kb.

Also, between minimap2 and RaCon there is clearly an ultra-fast gap-filling solution. I am toying around w/ the 2 tools in a kind of clunky way to do so. Do you have any recommendations for that?

Best,

John

[rvaser]

Hi John,
longer window lengths should yield a bit higher accuracy, although there is no guarantee (I have a comparison somewhere, will update the comment next week). The speed downgrades more when the length of the longest sequence in a window multiplied with the maximal absolute value from the alignment scoring matrix exceeds the size of 16bit signed integer (~32k). 32bit integers are then used instead to avoid overflows in SIMD instructions. You can try window length 1500 with default scoring parameters.

If I understood correctly, you are trying to use minimap2 and racon to fill gaps of unfinished assemblies with TGS data? Or is it something else?

Best regards,
Robert

[JohnUrban]

Hi Robert,

Thanks for the tips.

As for gap-filling -- yes I was able to accidentally fill small gaps using minimap2 and RaCon. It then occurred to me that Minimap2 and RaCon could be combined to have the same functionality (I think) as the commonly-used PBJelly2 gap-filler. Heck -- PBJelly2 could probably be updated w/ these 2 programs b/c now it uses BLASR and PBDagCon.

Best,

John

[rvaser]

If miminap2 was able to map a whole read over a short gap (without break points), then racon will probably fill the gap. Otherwise, you have to fill the gap with any reads before polishing so that other reads can map to it.

Best regards,
Robert