Selecting N50 and L50: implication for genome

[bio_informatics]

Hi Members,

I'm stuck with N50 and L50.
Have paired end genome. E. coli isolates. Assembled with SPAdes, and quality using Quast.

My understanding is based on information spread across different resources, and blogs. For I'm using Quast, I read its manual.

http://quast.bioinf.spbau.ru/manual.html#sec2.1

Based on different sources, blogs, and posts of N50, and L50 I'm looking for a base line or a benchmark to assess genome quality.

My question:
If I'm to decide about quality of my genome based on L50 and N50, how would I be doing it?

1) Minimum L50, Or,
2) Maximum N50, Or,
3) Minimum N50, Or,
4) Maximum L50

And, most importantly: Why?

Based on reply by user: Jeremy Leipzig at

https://www.biostars.org/p/723/

What he suggests I fail to understand.

If I were to say, then minimum number of Contigs, that is, minimum L50.
Or Maximum N50 (the length ).

Of course N50, and L50 cannot be considered as gold standard to decide of genome quality.

Any help would be greatly appreciated.

[westerman]

What Jeremy Leipzig at Biostars says:

you want L50 to be large and N50 to be small, so you wouldn't need many contigs to cover most of the assembly

Could be confusing. Unfortunately there is some disagreement in the field as to what the 'L' and 'N' mean. I think that most people take 'N' to be the 'length' and 'L' to be the number but there are some people (e.g., Brian of BBMap fame) who use 'N' to be the number of contigs and 'L' to be the weighted median length. Brian's use of 'L'ength and 'N'umber certainly makes sense from an English usage standpoint but does go contrary to the long standing use of 'N' for length.

So, going back to the quote, it could be that Jeremy was using L and N in the Length and Number sense.


Going back to your original question and using N as Length, given the choice of a single statistic to optimize I would go with a maximum N50 over a minimum L50. But this single statistic masks so many other considerations (e.g., did my assembly match expected genome size? percentage of unknown bases? and so on) that it is basically worthless by itself.

[bio_informatics]

What Jeremy Leipzig at Biostars says:



Could be confusing. Unfortunately there is some disagreement in the field as to what the 'L' and 'N' mean. I think that most people take 'N' to be the 'length' and 'L' to be the number but there are some people (e.g., Brian of BBMap fame) who use 'N' to be the number of contigs and 'L' to be the weighted median length. Brian's use of 'L'ength and 'N'umber certainly makes sense from an English usage standpoint but does go contrary to the long standing use of 'N' for length.

So, going back to the quote, it could be that Jeremy was using L and N in the Length and Number sense.


Going back to your original question and using N as Length, given the choice of a single statistic to optimize I would go with a maximum N50 over a minimum L50. But this single statistic masks so many other considerations (e.g., did my assembly match expected genome size? percentage of unknown bases? and so on) that it is basically worthless by itself.

Hi westerman,
Thanks much for your reply.
I see, it makes sense what Jeremy suggested, for the prevalent confusion of 'L' and 'N'.

I agree merely N, L 50s would be worthless if we overlook other parameters if genome length, number of Ns, etc.
I've already filtered poor assembled genomes based on genome size. Wanted to zoom in more if I could for another task

Thank you for your reply.