How may mis-scaffolding happen about arcs HOT 6 CLOSED

Hello Qihua,

The most important parameters in controlling assembly stringency are LINKS parameters -l and -a.
Increasing -l (try 10?) and decreasing -a (0.3 or 0.1) will help mitigate errors. We tend to keep -r (0.05) and -e (30000) at default, unless you have a fragmented assembly draft. The other ARCS parameter you could tune is -c, which controls the minimum number of aligned read pairs per sequence edge (you could try 5 to begin with, increasing will make barcode mapping more specific, but you will have to probably decrease LINKS -l accordingly).

Best,
Rene

from arcs.

Hi Rene,

Thanks for the suggestions on tuning LINKS parameters. I will try and see the alignments of new scaffolds with other available data.

Besides, recently I realize more than 1 (typically 1 to 10 according to 10x) HMW DNAs can be attached to the same GEM cell. Unlike 10x LongRanger which relies on a high quality reference and thus can cluster 10x reads based on the alignment locations, BWA alignments of 10x reads towards the draft assembly may not help to demultiplex 10x reads by the alignment locations along the reference. I am interested in how ARCS deals with demultiplexing reads with the same barcode back to the same HMW DNA.

The pre-print ARCS paper does not explain this in details, is there other available materials explaining how ARCS is designed?

Thanks
Qihua

from arcs.

Hi Qihua,
Yes, that is correct - barcodes are reused.
ARCS uses the number of barcodes supporting two sequences to suggest joins for LINKS. We consider every possible pair of sequences and look at evidence (reads aligning to the sequence edges only) supporting a relationship between two sequences. When barcodes are re-used, the sequences those reads aligned to must be supported by multiple (LINKS -l) barcodes to be linked. Barcode re-used would have for effect to create tips in the graph, but they would not be supported by a majority of distinct barcodes if the edge is valid.
I hope it makes sense.
Rene

from arcs.

Hi Rene,

I read the LINKS paper and I am still not clear about the algorithm part on picking dominant pairing(s) from all possible parings by -a. How is "maximum link ratio between two best contig pairs" calculated? Are we choosing among contig pairs that are supported by same multiple (LINKS -l) links?

Thanks
Qihua

from arcs.

Hi Qihua,

Please refer to slides 4 and 8 of this RECOMB presentation:

https://warrenlr.github.io/papers/2017_arcs_recombseq.pdf

Cheers,
Rene

from arcs.

Closing this old issue -- feel free to re-open if you still have questions.

from arcs.