PRJEB49093

[lskatz]

A colleague wrote:

Not sure if it is of interest, but PRJEB49093 has 151 M. tuberculosis isolates with matched Illumina and Nanopore data. Mixture of resistance profiles.
The raw nanopore data and all the metadata files can also be found in the Data Sharing section of our paper https://www.thelancet.com/journals/lanmic/article/PIIS2666-5247(22)00301-9/fulltext#seccestitle210

So I will check out that bioproject to see if any genomes are fast to assemble.

[lskatz]

After downloading all reads, these are the smallest file sizes after sorting and compressing with -9.

-rw-------. 1 gzu2 users 168M Mar 13 09:33 ERR9030520.fastq.gz
-rw-------. 1 gzu2 users 167M Mar 13 09:26 ERR9030361.fastq.gz
-rw-------. 1 gzu2 users 163M Mar 13 09:36 ERR9030334.fastq.gz
-rw-------. 1 gzu2 users 163M Mar 13 09:35 ERR9030284.fastq.gz
-rw-------. 1 gzu2 users 163M Mar 13 09:17 ERR9030249.fastq.gz
-rw-------. 1 gzu2 users 161M Mar 13 09:30 ERR9030283.fastq.gz
-rw-------. 1 gzu2 users 159M Mar 13 09:17 ERR9030474.fastq.gz
-rw-------. 1 gzu2 users 153M Mar 13 09:08 ERR9030398.fastq.gz
-rw-------. 1 gzu2 users 144M Mar 13 10:20 ERR9030424.fastq.gz
-rw-------. 1 gzu2 users 142M Mar 13 10:18 ERR9030315.fastq.gz
-rw-------. 1 gzu2 users 138M Mar 13 10:03 ERR9030319.fastq.gz
-rw-------. 1 gzu2 users 137M Mar 13 09:22 ERR9030505.fastq.gz
-rw-------. 1 gzu2 users 128M Mar 13 10:04 ERR9030329.fastq.gz
-rw-------. 1 gzu2 users 122M Mar 13 09:56 ERR9030258.fastq.gz
-rw-------. 1 gzu2 users 113M Mar 13 10:14 ERR9030503.fastq.gz
-rw-------. 1 gzu2 users 112M Mar 13 09:06 ERR9030316.fastq.gz

Assembling the smallest genome ERR9030316.fastq.gz takes this long

1236.86user 78.76system 5:36.83elapsed 390%CPU (0avgtext+0avgdata 8945592maxresident)k

And so I'm not sure if 5:36 is an acceptable time for a toy dataset.

[mbhall88]

Fair. You could always just subsample them? That'd assumable make them faster to assemble?

[lskatz]

I don't have subsampling in this repo but I will consider it for later. I don't necessarily want to add another dependency. But another advantage of this repo is that there is a hashsum, meaning that even if I did subsample, I would have to do it in a deterministic way.