Is 10x coverage suitable for resequecing 1Gb genome using 454?

[fusu]

Choosing between sequencing 30x coverage using Solexa and 10x coverage using 454, which is better for reference based assembly?

[westerman]

What type of organism? Reference base to a mostly non-repetitive organism such as E. coli then Solexa with its short reads would be great at 30x. To wheat which is highly repetitive (and, yes, is not fully sequenced) then 454 would probably be better.

Frankly at 10x coverage *any* system is likely to be good. Being able to afford 10x coverage, especially on a large genome, is in my mind the key point.

[swbarnes2]

For heterozygous SNPs, you want more coverage, maybe more like 50x average to be sure.

Paired ends on Solexa/Solid alleviate many of the downsides to the shorter read lengths.

[fusu]

Thx, westerman. The cost is very important, so I want to combine a low coverage 454 sequencing with high coverage Solexa sequencing. When we sequenced large genome using Sanger method, only 6x coverage can be afforded, but a deeper coverage would make our assembly result better. Since 10 fold is good, how about 6 fold in 454?

[lh3]

If cost matters, you may prefer Illumina. For resequencing of a diploid genome, 10X is not enough to get heterozygotes called correctly. 30X Illumina reads would be enough. Of course the decision also depends on how much repeats you want to recover.

[westerman]

lh3 and swbarnes both have good points -- coverage does depend on what your project goals are. Why don't you give the nature of your project -- SNP discover, de-novo assembly, or other -- and the approximate size of your target organism and we might be able to better discuss details.

[fusu]

Thx, westerman. We are planning to sequence a new genome about 900Mb, which had about 95% similarity to a sequenced genome according to cDNA comparison. So, our goal is to do de novo assembly and gene annotation. Thinking about cost, we want to use 454 to sequence a low coverage (under 10x), then adding Solexa or Solid data. So, do you think a 6x coverage 454 data is enough?

[mchaisso]

One thing to consider his how the assembly will proceed. If you are doing true de novo assembly, you will not be able to combine 454 and illumina reads with newbler, and with 6x coverage many regions will not assemble. If you are performing "reference guided de novo assembly", a 95% similarity may drop a bunch of alignments in areas of elevated divergence (fewer if using GAII longer reads), so you may not see a gain when using Illumina reads. Of course I would like to promote euler for assembling mixtures of reads, but unless you have a fat memory machine (256G), ABySS may be the only possibility.

-mark