0
$\begingroup$

I am looking for homologous genes in various protozoan species using BLAST. The genome sequences of these species are deposited in NCBI's WGS database. NCBI’s webpage includes the global statistics for each WGS project and it indicates the assembly level (scaffold, contig, etc.), the assembly statistics (number of contigs, N50, L50, etc) and (if available) the genome coverage. Can any / some of these parameters be used to evaluate the genome assembly’s "quality", especially if there is more than one genome assembly for the same species? Can these data be used to determine if the genome is very fragmented or not?

I understand that quality is a relative term and that there are several options for assessing genome assemblies (QUAST, BUSCO, etc.) but I am just looking for a "simple" way to compare several genome assemblies (for e.g., this post How can I improve a long-read assembly with a repetitive genome? states that the reference genome is highly fragmented, but there is no explanation regarding how the OP determined this). I believe that the N50 value can be helpful, but I need some context to understand what is "good" or "bad". I am at a loss of where to start, so any information will be greatly appreciated!

-Leah

$\endgroup$
3
$\begingroup$

It is not so easy to determine the quality of an assembly - but sequence length, number of contigs, N50 and L50 can give you some guidance.

How much does the sequence length deviate from the expected length (or how much variation is among assemblies of the same species). Same for number of contigs/scaffolds. The sequence length, and number of contigs together with N50 and L50 give you an idea about the fragmentation. Very small N50 compared to expected genome size and chromosomes is a likely indication of high fragmentation, same for high L50 compared to expected number of chromosomes.

However, these number can be misleading. High N50 and low L50 can occur witho some very long contigs, so you may want to compare N75/L75 or better the Quast statistics

| improve this answer | |
$\endgroup$
1
$\begingroup$

Carambakaracho's answer is good. A high N50 should usually be a good sign, as well as agreement between the overall genome assembly length and the expected genome size. The number of 'N's in the assembly may also be important to check, as Ns are sometimes used to indicate single unknown bases, and sometimes to indicate unknown numbers of unknown bases, and Ns may be used to connect scaffolds,which may lead to artificially higher N50. If you have access to a set of genome sequence reads (Illumina) for the same organism, then another more detailed check you can make is to compare the kmer spectrum of the reads with the kmer spectrum of the assembly, using the kat tool.

Finally, if you want to compare several assemblies, you might align them against each other, to see how they compare qualitatively (e.g. using mauve).

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thank you very much for your answers, I really appreciate your help. @Jonathan Moore, just a couple of questions: (1) what would be considered a good / bad kmer spectrum of the reads? and (2) is it possible to align the whole xxx_genomic.fna files from NCBI using the mauve tool? $\endgroup$ – Leah Sep 19 '19 at 0:20
  • 1
    $\begingroup$ 1) a good assembly should have similar kmer spectrum for the reads as for the assembly, with the ratio in abundances between the two spectra reflecting the depth of coverage of the reads. If some kmers have a large difference in abundances in the reads compared to the assembly (double for instance), it may be that these parts of the assembly do not have haploid copies collapsed correctly, or some other issue with the assembly, and if there are many kmers in the reads which are present in large numbers but absent from the assembly, it may be evidence of missing parts of the genome. $\endgroup$ – Jonathan Moore Sep 20 '19 at 12:12
  • 1
    $\begingroup$ 2) I think mauve should be able to align the whole genome .fna files. It may be somewhat slow with larger genomes. $\endgroup$ – Jonathan Moore Sep 20 '19 at 12:14
  • $\begingroup$ Thank you very much @Jonathan Moore! I will give them a try. $\endgroup$ – Leah Sep 21 '19 at 3:08

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.