E.coli contamination?

mbk0asis · 11-08-2013, 05:12 AM

Hi, all!

We recently ran Chip-seq and RNA-seq and got decent amount of data.
When we the raw sequences by running in NCBI blast, all the sequences from Chip-seq were aligned to E.coli genomes. However, only a few sequences were mapped on E.coli genomes.
FYI, our samples were from mouse cells.
Has anyone experienced this?

Richard Finney · 11-08-2013, 05:22 AM

I've seen bacteria and virus sequences in data from humans. Some is probably really in the sample, some is contamination from the air.
It's certainly not unheard of.

GenoMax · 11-08-2013, 06:02 AM

Quote:

Originally Posted by mbk0asis

When we the raw sequences by running in NCBI blast, all the sequences from Chip-seq were aligned to E.coli genomes. However, only a few sequences were mapped on E.coli genomes.
FYI, our samples were from mouse cells.
Has anyone experienced this?

Can you clarify if *all* sequences mapped to E. coli or only some did (what fraction of total)?

If it is the first case then you either did not get data back from your own sample or something went terribly wrong in your sample library prep.

mbk0asis · 11-10-2013, 04:49 PM

Sorry about late reply, GenoMax.
I tested 10 reads on blastn, and all of them were aligned to E.coli sequences.
I also tested 10 reads from another sequencing data we ran last year, and 4 out of 10 sequences were from E.coli. As Richard said, it's somewhat common to see those sequences.
Thanks!

GenoMax · 11-10-2013, 05:56 PM

Have you tried to analyze the full data set? 10 sequences out of a few million hitting the E. coli genome by blast (are the hits full length identities) may not be a worrisome thing if majority of the sequences are indeed mouse. Until you are convinced otherwise, I would advise moving forward with the alignments to the mouse genome/transcriptome (and if you want to be dead certain Ecoli genome independently). See what fraction of reads align in the two comparisons.

mbk0asis · 11-10-2013, 07:03 PM

Thank you for the advice.
I already tried mapping with bowtie on the mouse genome, and less than 1% of raw reads (~3000 reads) were mapped on the mouse. I think I will try bowtie on Ecoli genome to see how many reads would be aligned.
I think it's most likely the Ecoli contamination.
If so, in which step during library prep, do you think that could happen?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Second Strand Step - E coli Ligase?	lucialucia11678	RNA Sequencing	0	03-06-2013 03:33 AM
Mapping E.coli PE reads	aggp11	Bioinformatics	3	05-14-2012 10:48 AM
E coli de novo sequencing	SHB	General	5	10-28-2011 04:47 AM
Illumina for resequencing E. coli	jkersh	Illumina/Solexa	16	11-11-2010 01:05 PM
Sequencing of E. coli strains	pDNA	454 Pyrosequencing	2	05-12-2010 01:45 PM

11-08-2013, 05:12 AM	#1
mbk0asis Member Location: Korea Join Date: Dec 2011 Posts: 41	E.coli contamination? Hi, all! We recently ran Chip-seq and RNA-seq and got decent amount of data. When we the raw sequences by running in NCBI blast, all the sequences from Chip-seq were aligned to E.coli genomes. However, only a few sequences were mapped on E.coli genomes. FYI, our samples were from mouse cells. Has anyone experienced this?

11-08-2013, 05:22 AM	#2
Richard Finney Senior Member Location: bethesda Join Date: Feb 2009 Posts: 700	I've seen bacteria and virus sequences in data from humans. Some is probably really in the sample, some is contamination from the air. It's certainly not unheard of.

11-10-2013, 04:49 PM	#4
mbk0asis Member Location: Korea Join Date: Dec 2011 Posts: 41	Sorry about late reply, GenoMax. I tested 10 reads on blastn, and all of them were aligned to E.coli sequences. I also tested 10 reads from another sequencing data we ran last year, and 4 out of 10 sequences were from E.coli. As Richard said, it's somewhat common to see those sequences. Thanks!

11-10-2013, 05:56 PM	#5
GenoMax Senior Member Location: East Coast USA Join Date: Feb 2008 Posts: 7,087	Have you tried to analyze the full data set? 10 sequences out of a few million hitting the E. coli genome by blast (are the hits full length identities) may not be a worrisome thing if majority of the sequences are indeed mouse. Until you are convinced otherwise, I would advise moving forward with the alignments to the mouse genome/transcriptome (and if you want to be dead certain Ecoli genome independently). See what fraction of reads align in the two comparisons.

11-10-2013, 07:03 PM	#6
mbk0asis Member Location: Korea Join Date: Dec 2011 Posts: 41	Thank you for the advice. I already tried mapping with bowtie on the mouse genome, and less than 1% of raw reads (~3000 reads) were mapped on the mouse. I think I will try bowtie on Ecoli genome to see how many reads would be aligned. I think it's most likely the Ecoli contamination. If so, in which step during library prep, do you think that could happen?