sample contamination?

[sjkimble]

I seem to have some contamination in my demultiplexed sequences, but I'm not sure what to make it: is it phix contamination? library prep contamination? something else? details:

gDNA from 96 eukaryotic individuals were ddRADseq prepped per Parchman et al. 2012, sequenced on one late of illumina nextseq500 with 83bp PE reads. Raw files were demultiplexed with process_radtags in STACKS:

process_radtags -P -p ./raw_reads_fastq_format/ -b ND_barcodes5.txt -o ./demultiplexed_jan_2019/ -r -i fastq -y gzfastq --inline_null --renz_1 ecoRI --renz_2 mseI -s 10 -w 0.15 --disable_rad_check -D

blasting resulting sequences hits mostly prokaryotic genomes, including phix. further, some of the files contain a low diversity of sequences. any insight would be appreciated!

[SNPsaurus]

What kind of organisms were you genotyping. Some samples we do, like small insects, can have significant bacterial contamination due to the microbiome of the organism. Other kinds of samples, like museum samples or field collected samples that weren't stored in a way to prevent bacterial growth, also have high bacterial loads.

If we are working on a project with a little known species, the blast will hit any contaminating bacteria but not the species of interest, but that just reflects the absence of a closely related genome to the host. So if you aren't getting many blast hits with a high proportion of the ones you do get being bacteria then it might be OK...the ones you want just aren't blasting. If you get a blast hit to every read and it is nearly all bacteria then you have a problem!

If you have the ability to throw a few samples onto the next run, you might try just doing WGS on a few samples (with or without contamination) and seeing how the DNA looks independent of the ddRAD prep. Maybe the cut site frequency is much higher in your bacterial genomes, for instance.

[sjkimble]

SNPsaurus:

Thanks for your reply. Yes, these were frog toes so perhaps we've sequenced the cutaneous microbiome! I will do more BLASTING to get a better proportion of the sequences are frog and yes, there are very few amphibians out there in sequenceland.

Cheers,
Steve

[sjkimble]

This leads me to a more general question, which is: is BLASTing the best way to detect contamination in any sequence data? It is computationally expensive!

[GenoMax]

It depends. If you want a quick scan you could use something like fastq screen. You will need to select the database carefully.

If you know the genome you are interested in then you could use bbsplit.sh from BBMap to recover the reads you are interested in and leave the contamination behind in a different file/pool.