Making a in silico mock community

[horribulas]

Hello All!

I am looking to some microbiome work of fecal samples (16S V4 MiSeq). We are interested in detection a number of specific groups of bacteria in the samples. Specifically E. Coli and different Klebsiella species.

I would like to know in silico whether we will be able to differentiate between species after we run the sequencing.

What is the best method for testing whether our V4 sequencing will be able to differentiate to species level? Should I just be doing a bunch of blasting or would making a mock in silico community work best?

Thanks for any help

Mark

[Brian Bushnell]

If you are looking for specific microbial species (or really doing anything with NGS reads), BLAST is not really the best tool. You can calculate abundance by mapping or kmer-matching your sequences to just the set of references of interest.

And certainly, you should first validate that your approach works on synthetic data before proceeding. That will allow you to quantify true positive, false positive, and false negative rates, which may lead you to adjust your methodology prior to doing any real work.

For kmer-matching, I recommend Seal (part of the BBMap package), which is fast and easy to use for quantifying sequence expressions levels. For example:

seal.sh in=reads.fq ref=ecoli.fa,klebsiella1.fa,klebsiella2.fa stats=stats.txt refstats=refstats.txt

If you generate synthetic reads with errors, and annotate the reads with the name of the organism they came from, you can then objectively calculate the accuracy. Alternatively, if you just want to measure levels of various bacteria, you can generate synthetic reads and mix them in a specific ratio, then evaluate your approach by calculating how closely the output of expression levels matches the ratio of reads you mixed together. That's easier as it does not require parsing the read names.