estimated size VS de novo total contig size

[biocyberman]

Hi,
Could you comment about where this method can be more or less accurate than the size obtained by summing sizes of contigs from de novo assembly of the same dataset? Let's say I already have the best possible de novo assembly output.

[wrf]

Kmer depth is commonly used to confirm the assembly size, and so both are often reported in genome papers. The advantage of the size estimate is if you do not have an assembly or do not know how big the genome should be, you can guess without assembling the data. The estimated size should roughly be the sum of the contigs, though it can overestimate sometimes (maybe 5 to 10 percent).

[josephryan]

Some genome assembly software estimates genome size using a similar k-mer approach and then makes sure that the resulting assembly matches the size estimate. However, some assemblers will compact repeats and the resulting assembly will be very much an underestimate of the real genome size. A very defined peak using estimate_genome_size.pl should be a good approximation of size. Consequently, if the size estimated from these scripts is substantially larger than the size of an assembly, it may indicate excessive compaction of repeats in the assembly.