How to identify the contamination of the genome from the RNA seq data?

[tianyub836]

Hi, guys!

i am working on the analysis of the RNA seq data of my group but i deeply doubt that my data has been contamintated by the tags generated from the genome because the ratio of alignment to genome is extremely high (more than 85%) however the ratio of transcripts is very low (lower than 20%).

i am trying to find an approach to prove that and any suggestion is welcomed!

thank you!

[cedance]

Ratio usually means two parameters. What do you mean by "ratio of alignment to genome" and "ratio of transcripts is low"?

By tags, do you mean barcodes? If so, you'll have to remove them as a part of your preprocessing step. But I think you already know that.

[tianyub836]

Ratio usually means two parameters. What do you mean by "ratio of alignment to genome" and "ratio of transcripts is low"?

By tags, do you mean barcodes? If so, you'll have to remove them as a part of your preprocessing step. But I think you already know that.

i am sorry for not clearly expressing myself. actually i refered to the percentages of tags matched to the reference genome and transcripts RNA.
and for the tags, i meant the reads generated from the Illumina RNAseq platform.

i hope that might express myself clearly

thanks

[gringer]

Have you filtered out ribosomal RNA first? I've recently analysed some yeast RNA reads, and about 70-80% of the reads matched to the yeast ribosome fragments (which is consistent with the values you have here, assuming the 20% is non-filtered).

If you're doing RNA expression analysis, most of the reads will likely match to the ribosomal RNA, rather than transcripts (this is likely to be the case even when using a workflow that attempts to remove rRNA). The rRNA will match the genome, but not transcripts.

[cedance]

From my understanding, there are two ways of library preparation. From the mRNA, you could either obtain fragmentation of mRNA-cDNA complex by random priming or you could first create mRNA-cDNA complex and then obtain fragments.

Correct me if I am wrong, but in both cases, we rely on the poly-A signal for primer selection = poly-T (for double-strand formation). Therefore, I don't agree that they are ribosomal RNA. How do you obtain the fragments and how does priming occur in that case???

[swbarnes2]

From my understanding, there are two ways of library preparation. From the mRNA, you could either obtain fragmentation of mRNA-cDNA complex by random priming or you could first create mRNA-cDNA complex and then obtain fragments.

Correct me if I am wrong, but in both cases, we rely on the poly-A signal for primer selection = poly-T (for double-strand formation). Therefore, I don't agree that they are ribosomal RNA. How do you obtain the fragments and how does priming occur in that case???

Benchwork techniques don't always do the job 100% as well as you'd want them to. There is a lot more rRNA in a cell than there is mRNA, so unless your selection method is virtually perfectly effective, there will be rRNA. So don't disimiss the possibility just because the protocol theoretically screens out rRNA.

[cedance]

swbarnes2, Thanks for your reply. I don't disagree the presence of rRNA. My knowledge of lab protocols are entirely theoretical. I read from Illumina's RNA-Seq (before fragmentation + cDNA construction ...) that

This process purifies the poly-A containing mRNA molecules using
poly-T oligo-attached magnetic beads.

You say that here, there might still be sufficient amount of rRNA after purification, I suppose?

[tianyub836]

thanks for your replies, gringer, cedance and swbarnes2.

i do have read the Illumina technical manual and i understand the procedures of mRNA purification by adopting magnetic beads. And i am not sure if there was a kind of process called 'washing' to wash away other nucleotides such as genome or rRNA during the purification.

what i found in my work was that the reads , which i prefer to say tags, that failed to match to the transcripts matched to the introns in the genome.

i am not able to figure it out any way. That might be the differences between theory and application.

by the way, i am working on the DEGs of plants and i am not sure about the impact of rRNAs.

[cedance]

Hi tianyub836,
I can relate to the fact that reads map to introns of your genome. They just mean that there are quite a bit of "intron retention events". And in plants they are the most predominant form of Alternative splicing events when compared to humans, for example, where exon skipping is predominant. However, what I am not able to answer is the difference in the reads you get when you map to the genome and transcriptome. This means that almost all your coding genes have a (or many) intron retention events (assuming everything else is right and the transcriptome you use is complete).

[tianyub836]

Hi tianyub836,
I can relate to the fact that reads map to introns of your genome. They just mean that there are quite a bit of "intron retention events". And in plants they are the most predominant form of Alternative splicing events when compared to humans, for example, where exon skipping is predominant. However, what I am not able to answer is the difference in the reads you get when you map to the genome and transcriptome. This means that almost all your coding genes have a (or many) intron retention events (assuming everything else is right and the transcriptome you use is complete).

thank you for your tips, cedance.
now maybe i have to check it out on the alternative splicings of my transcripts.
it is might be a hard jobs.

[cedance]

Actually, think of it, it makes more sense for you to map a lot of reads to the genome (across introns, that too) and not to your transcriptome because of excessive pre-mRNA (rather than rRNA assuming not too much contamination in lib preparation). I suggest you use the --butterfly-search option in tophat while trying to map against transcriptome and see how it does. I haven't used the option and don't know how good it is, but its worth a try.

[Marsa]

I am wondering, when I use short aligner (Bowtie) to align reads (75nt) against rRNA for determination of contamination how many mismatches should I allow?
Thank you

[cedance]

BWA is designed for small mismatches anyways. Its fastest with no mismatches (I wouldn't go more than 5). In my case, deciding mismatches depends on the species for which I have the transcriptome data and the reference genome I have to map to (due to non-availability) and their divergence. How do you obtain the rRNA reference btw?

[Marsa]

@cedance
I work with human data. That's why I doubt about mismatches. In my case should I make alignment more strict or allow a couple of mismatches? When I aligned with 0 mismatches less than 1% of all reads mapped to the rRNA and chrM, when I allowed 2 mismatches 5% of reads where mapped. I don't think that this is significant contamination. However I would like to know what how many mismatches would you allow in this cace. rRNA sequences I downloaded from ncbi. I was able to find all rRNAs except 16S (I am not sure if this one can introduce high contamination).

[cedance]

Marsa,
what's the fold/coverage you obtained? n=2 seems fine to me and as you say, you don't have to be alarmed as it doesn't seem significant.