I am interpreting microbiome data output from illumina miseq. The genus read counts are significantly higher than genus read clusters, and I am trying to understand the difference between them. A sequencing company performed the bioinformatics and I'm finding it hard to make sense of this and can't get any good explanations from them so far. Here's an example:
Genus Hafnia
Genus.Counts.Reads 328486
Genus.Counts.Clusters 18547
%Composition Clusters 26.4
Any ideas would be much appreciated.
Thank you.
I don't know which software they used, but I assume that it's a clustering software. Maybe is the number of clusters classified as a specific genus (Genus.Counts.Clusters) and the Genus.Counts.Reads is the number of reads that belongs to those clusters.
Counts.Reads will be the raw number of reads, which have likely passed various quality filters.
Counts.Clusters will be the number of unique sequences within the total number of reads.
Even if I am not an expert in microbiome seq data analysis, looking at your example I think it means that there are 328K reads that are specific for the genus Hafnia. These reads are able to cluster into 18K clusters (small genome parts). I do not understand the %Composition.
Counting the reads that belong to different genus would allows you to get an idea of the relative frequency of these genus in the sample.
Hope this help
Dear Philip,
It is hard to say without knowing what the company actually did and what was in your library (PCR amplicons or metagenomes?). My guess is that they grouped all reads by sequence similarity in order to simplify the dataset (or create MOTUs if your run contained PCR amplicons). In that case, Genus.Counts.Clusters would correspond to the total number of groups of sequences (that are 100% or e.g. 97 % similar), that correspond to the genus Hafnia (the way they did their taxonomic assignments remains to be clarified…). Hence, your results would make sense as each of these groups of sequences may be represented by 1 or more reads.
HTH
Illumina MiSeq produces paired end reads. This means 2 reads are read off of each cluster. Since your read count is not quite 2x the cluster number, the company may be reporting the higher quality read counts only.
Thank you all for your help. They were 16S PCR amplicons Lucie, and I think that your explanation makes total sense. It is sometimes just confusing how "clusters" is used at times, as it seems it can have different meanings
Genus.Counts.Reads: The number of reads assigned to this genus
Genus.Counts.Clusters: The number of OTUs (Species, Clusters) belonging to this genus (I guess you used 0.03 cutoff, then the OTU will be species)
Hi Philip,
First of all, without any more detail on the kind of experiment and on the bioinformatic analysis which was performed any answers will more or less be speculative.
Anyhow, from the term microbiome i infer that you're speaking of PCR amplicons which were sequenced. Then the bioinformatic analysis usually involves the generation of OUTs, which is usually carried out by clustering. Thereby, your sequencing reads are clustered at a specific distance such that reads within that distance to each other end up in the same cluster (very very coarse). After that, the cluster can be taxonomical annotated, again usually by picking a cluster representative and taxonomical classifying it. Then, the Genus.Counts.Clusters refers to the number of clusters that were - in this manner - assigned to this genus. Genus.Counts.Reads then is just the number of all reads those cluster are containing, i.e. the summed cluster sizes. Ideally, during clustering you would like to end up with one single cluster for each category representing all reads of exact that category. However, due to PCR and NGS biases, errors and the like, a high degree of variation is added to the sequencing reads which, in addition to other aspects (copy number variations) can led to such inflation. i would guess that the %Composition Clusters is referring to the relative amount these clusters to the total number of clusters.
One last note. The number of total reads and also the number of clusters seem very high for me for an common amplicon study. Are those from one single sample?
I hope this helps,
Sebastian
One additional note, as i can already read it here again and this it's one central aspect which may contribute a lot to this confusion (and always did):
An OTU is NOT a Genus/Species or anything like that!
If OTUs were generated by clustering, this means (in this context), again:
A Cluster is NOT a Genus/Species or anything like that!
It's an (operationla taxonomic) unit, a union of reads sharing some specific constraints, which can be assigned to belong to a Genus/Species, but that does not magically transform them into being a taxa. First, because of the more or less arbitrary clustering distance (or the cutoff for hierarchical clustering) which just can not reliable separate between the desired taxa and, second, because of the already mentioned artificial or natural variation in the data. So, if you have e.g. 100 OTUs (say clustered at 97%; putting aside any critique on this standard for now), you don not necessarily have to have 100 species.
As I look at the information you provided, I can say
Genus.Counts.Reads 328486 : These must be the toal number of reads sequeced in that genus
Genus.Counts.Clusters 18547 : These must be the number of clusters obtained after grouping 328486 reads basing on some similarity factor
%Composition Clusters 26.4: This could be the read composition present in that cluster with respect to the total number of sequenced reads
- Badges
- Science method