We have lab established culture setup of marine synechococcus PCC 7002 sp. and heterotrophic bacteria. Recently we did metagenomic analysis of it (sequencing was done by an outside reputed company). According to the metagenomic analysis our sample contains other cynaobacterial algae corrosponding to synechococcus and Lyngbia genera (taking upto less than 1 to 2% of the reads for different samples from same culture).
We tried to identify this other algae morphologically using microscopy, flowcytometry and plate culture method but at no point we observed any other algal cells than Synechococcus PCC 7002. 16S amplicon sequencing also did not show any presence of other algae.
What might be the reason for such a different result from metagenomic ( apart from the assumption that contamination were introduced while sequencing as the company is reputed and we have sequenced different samples processed from same culture system)?
It may be an issue with the classification of reads, rather than contamination. If you get a single classification for every read (and that's quite normal to only get one per read) you may get an incorrect classification if there are multiple hits which are given the same score. For example if one of the reads scores 100% identity for both PCC 7002 and Lyngbia then the classification would depend on the order they occur in the list of hits as only the top one will be reported.
What you can do is isolate the reads that seem to be from the wrong organism and run them through Blast searches to see exactly what is happening.
Shailesh Nair were whole or partial genomes reconstructed from the metagenomic dataset for the particular species, or was it just bunch of unassembled reads that were assigned as that?
It may be an issue with the classification of reads, rather than contamination. If you get a single classification for every read (and that's quite normal to only get one per read) you may get an incorrect classification if there are multiple hits which are given the same score. For example if one of the reads scores 100% identity for both PCC 7002 and Lyngbia then the classification would depend on the order they occur in the list of hits as only the top one will be reported.
What you can do is isolate the reads that seem to be from the wrong organism and run them through Blast searches to see exactly what is happening.
Deni Ribicic The results are from two separate analyses. The first one is from the sequencing company>>>> which they didn't is from partial constructed genes using Diamond. The second analysis I did by just running an unassembled dataset. Now I am in the process of constructing the genomes but still won't get all the genomes of present bacteria.
Vincent Mulholland Sorry I didn't quite understand what you mean by " single classification for every read " (I AM AMETUAR IN THIS FIELD). Do you mean single read for the said classification than NO almost 1% of the all classified reads were taken up by the said classification
Shailesh Nair When you classify a read you search it against a database. If you were doing a search on just one sequence you might get a list of similar organisms starting with the most similar going down to less similar sequences. Because of the number of sequences you are dealing with, what comes back from database searches is not a list of sequences for each read, it is only the top hit for each sequence. So if the sequence is identical to several sequences on a database it would just return one of those sequences, not all of them. Say the sequence is 100% similar to both species A and species B, you could classify the sequence as either A or B depending on what the order the list was in. So if you have a sequence which is 100% identical to PCC 7002 and Lyngbia it would depend on which was the top hit when the read was classified. It might be a PCC 7002 sequence, but because it is not the number 1 on the classification search it could mis-classified as Lyngbia. Classifications at the species level are often unreliable because there are too many similar sequences from closely related organisms. 1% of the reads mis-classified does not seem too bad.
Vincent Mulholland yes, that's correct. I will isolate these specific reads and then run blast individually on them. Thanks for the great help.
Vincent Mulholland has summed up pretty well why you could get such an output.
I would just add that when doing shotgun metagenomics you are dealing with really short fragments of genes (even when assembled into contings) and mismatching against databases will therefore occur. The safest bet IMO is to try to reassemble genomes from the dataset and then see what do you have.
Deni Ribicic yes.Thanks for the heads up. currently, I am doing that..reassembling the raw files..thanks
I would also encourage a reverse analysis. Say take Synechococcus PCC 7002 genome (or the closest available relative) and map your sequencing reads to that genome. Check the coverage and the proportion of recruited reads. Do the same but now with a genome, or a set of genomes, of potential contaminants. Does the recruitment rate change? If Synechococcus PCC 7002 is still recovering the majority of the reads, I would confidently ignore the Diamond results, and focus on the Synechococcus recruited reads. Hope this helps.
- Badges
- Science topic
- Similar topics
- Biological Science
- Microbiology