Salmon quants command keeps on getting killed

Question

[cengiz]

Hi all!

I'm a wetlab guy quite new to data analysis and would appreciate some help if possible!

Currently i'm trying to run Salmon quants on my desktop - using Ubuntu terminal for this and it seems that my computer is unable to do what i am ask it to do. Maybe i have a mistake with my command? Here is what i see;

(salmon) cengiz@cengiz:~/Documents$ salmon quant -p 1 -i salmon_index/default/ -l A -r raw_data/NGS_211_553_fastq/_1_Cengiz.fastq.gz -o salmon_output/
Version Server Response: Not Found
### salmon (selective-alignment-based) v1.5.2
### [ program ] => salmon 
### [ command ] => quant 
### [ threads ] => { 1 }
### [ index ] => { salmon_index/default/ }
### [ libType ] => { A }
### [ unmatedReads ] => { raw_data/NGS_211_553_fastq/_1_Cengiz.fastq.gz }
### [ output ] => { salmon_output/ }
Logs will be written to salmon_output/logs
[2021-11-08 20:03:00.308] [jointLog] [info] setting maxHashResizeThreads to 1
[2021-11-08 20:03:00.308] [jointLog] [info] Fragment incompatibility prior below threshold.  Incompatible fragments will be ignored.
[2021-11-08 20:03:00.308] [jointLog] [info] Usage of --validateMappings implies use of minScoreFraction. Since not explicitly specified, it is being set to 0.65
[2021-11-08 20:03:00.308] [jointLog] [info] Setting consensusSlack to selective-alignment default of 0.35.
[2021-11-08 20:03:00.308] [jointLog] [info] parsing read library format
[2021-11-08 20:03:00.455] [jointLog] [info] There is 1 library.
[2021-11-08 20:03:00.705] [jointLog] [info] Loading pufferfish index
[2021-11-08 20:03:00.716] [jointLog] [info] Loading dense pufferfish index.
-----------------------------------------
| Loading contig table | Time = 33.453 s
-----------------------------------------
size = 36981178
-----------------------------------------
| Loading contig offsets | Time = 1.0236 s
-----------------------------------------
-----------------------------------------
| Loading reference lengths | Time = 22.564 ms
-----------------------------------------
-----------------------------------------
| Loading mphf table | Time = 14.44 s
-----------------------------------------
size = 3784352032
Number of ones: 36981177
Number of ones per inventory item: 512
Inventory entries filled: 72229
-----------------------------------------
| Loading contig boundaries | Time = 10.044 s
-----------------------------------------
size = 3784352032
-----------------------------------------
| Loading sequence | Time = 6.9525 s
-----------------------------------------
size = 2674916722
-----------------------------------------
| Loading positions | Time = 96.926 s
-----------------------------------------
size = 3401221626
Killed

This has happened a few times - i was at first running with 3 threads and the initial steps were super fast before slowing/ stopping and crashing, then i tried 2, lastly 1. It does the exact same thing regardless. It gets as far as loading positions then stops with the line 'Killed'

IS it simply my computer unable to? I did notice it generally slowing down as i downloaded more and more files. I have 16GB of memory with an Intel Core i7-6700K CPU @4Ghz x8 so i hope powerful enough for this command.

Apart from add memory what are my options? Is there a problem with my script, or an alternative workflow one could use?

Many thanks, Cengiz

[GenoMax]

Just to confirm, are you aligning against transcriptome? salmon is not meant to map data to genome.

[cengiz]

Hi GenoMax,

Thanks for engaging. Yes, I am using an already indexed transcriptome for Salmon downloaded from the following web-page - http://refgenomes.databio.org/v3/assets/splash/2230c535660fb4774114bfa966a62f823fdb6d21acf138d4/salmon_sa_index?tag=default - so this is the index file I am using, it is pretty large (weighing in at 16.5GB)

There is an alternative, I hadn't tried - http://refgenomes.databio.org/v3/assets/splash/9a02d64909100f146272f8e16563178e9e93c218b1126ff9/salmon_index?tag=default - this one weighs in at just 617.9MB so maybe more manageable for my computer

Answer 1

[rpolicastro]

The larger one includes the genome in the decoy file, increasing the accuracy of quantification (but increasing memory requirements). In my experience using the genome as a decoy requires roughly 20G of memory, so your job is probably getting killed because of that.

[GenoMax]

I am going to second this. If you don't have access to any other hardware switch to the plain transcriptome reference (minus the decoy). That should need ~5-6 G of RAM.

[cengiz]

Thanks for the help and feedback! Indeed I was able to go ahead and run salmon quant using the alternative transcriptome file. I am now getting into the DESeq2 analysis and seems to have been just fine !