How To Manually Check Your Putative Chimeric Sequences

Question

[samlambrechts299]

Hi everybody,

Various sources tell me it is not recommended to blindly discard all sequences flagged as chimeras by tools such as Chimera Slayer and UCHIME, and that you should manually check if they really represent PCR artifacts. Now I have been searching the literature as to how to do this but I can't seem to find an answer. Are there any tips as to how to further analyze putative chimeric sequences?

To clarify: We are talking about some 100 OTU representative sequences (16S bacterial pyrosequencing reads, approximately 250bp each) which remain unclassified at any level after alignment with SILVA, Greengenes, LTP, and RDP, and I want to dig deep as to whether these are PCR artifacts or possible new taxa (and not just rely on the UCHIME software).

Kind regards,

Sam

Answer 1

[Istvan Albert]

One possible avenue would be to align these sequences via BLAST against known sequences. The ends of a chimeric read would presumably align well to different known taxa whereas a new species should align reasonably well over the entire read to a related species.

Answer 2

[Josh Herr]

+1 on Istvan's response. Just adding to his comment about BLASTing to known sequences: you should be able to determine the point where the chimera starts by inspecting your BLAST alignment from your sequences to your database. I would use a known reference database, but especially BLAST to your own raw dataset, as chimeras will arise from "hybrid sequences" within your own dataset, which may reflect novel 16S sequences not represented in SILVA, RDP, or greengenes and would reduce the search time of BLASTing to large reference databases.

To save time you can use the segregated putative chimeras from UCHIME or chimera slayer and/or parse out any BLAST hits to your database which appear to be chimeras (as Istvan said: search for uniform evenness vs. strand differences) and then inspect them manually. The chimera question is not an easy one to answer as it's estimated that a portion of reference sequence databases consist of chimeras also.

Answer 3

[cdsouthan]

I have had some past experience with chimeric mRNAs and the key manual check was inspecting the BLAST graphical outputs. What you may see is very clear breakpoints as vertical discontinuities. The matches either side of the chimeric break can tell you what's going on.

Answer 4

[samlambrechts299]

Thank you very much! The first part was something that had crossed my mind, but I doubted and left the idea because I couldn't find a reference in the literature or something on the internet. Your answers are very clear and helpfull! Hallelujah for this website!

Again, thank you very much,

a student from Belgium

[Josh Herr]

Please put all comments by editing your question above. Thanks!

Answer 5

[samlambrechts299]

Hi,

As I am manually checking my putative chimeras, I have ran into a few cases where it's a bit confusing. I added one example below, where there are hits that align pretty well to the entire length of the read (uniform evenness), as well as strand differences that look like evidence for a chimera. Are these natural chimeras? Or how should I interpret these?

Thank you,

Sam

[Josh Herr]

Sam, this is not an answer to your question. Please put this information in the question section; this will allow us to be able to see what evidence you have in regards to your problem.

Right now, you're just showing BLAST alignment scores for your sequences. These alignment scores could be caused by lots of factors (species diversity, haplotypes, etc.) so you'll have to show us more evidence of your chimeric sequences if we are to help you.