Question 1

Marker gene survey or whole community sequencing?

Accepted Answer

It depends on the question. Amplicon surveys are cheaper per sample and suit large designs, but resolve taxonomy coarsely and infer function only indirectly. Whole community sequencing gives gene content directly and finer taxonomic resolution, at higher cost per sample and greater analytical burden. Many projects use amplicon breadth with deeper sequencing on a subset.

Question 2

Why does compositionality change the statistics?

Accepted Answer

Because the data carry only relative information. Total sequencing depth is fixed by the run, not by the sample, so proportions must sum to one and an increase in one taxon appears as a decrease elsewhere. Use log ratio approaches or tests designed for constrained data, and state the transformation applied before any difference testing.

Question 3

Which reference database should be used?

Accepted Answer

Choose one suited to the environment and record its exact release, since taxonomies are revised and results shift between versions. Report the fraction of reads left unassigned, which indicates how well the reference covers the community. Where possible, run a second database and note whether the main conclusions hold.

Question 4

How should low biomass samples be handled?

Accepted Answer

With controls at every stage. Include blank extractions and no template amplification controls in each batch, sequence them alongside the samples, and report what grew in them. Reagent contamination can supply most of the signal in a low biomass sample, so a taxon abundant in blanks should be treated with suspicion wherever it appears.

Question 5

Is rarefaction still acceptable?

Accepted Answer

It remains contested. Rarefying discards data and loses power, while ignoring uneven depth entirely biases diversity estimates. Many projects report results under more than one normalisation and show that conclusions do not hinge on the choice. Whatever route is taken, state it plainly and give the depth used.

Question 6

How far can functional inference be trusted?

Accepted Answer

Gene content indicates capability, not activity. Predicting function from marker gene surveys relies on reference genomes and works poorly for poorly characterised environments. Even from whole community sequencing, a gene present may never be expressed. Where activity is the claim, pair the profile with transcript or metabolite measurement.

Bioinformatics

Bioinformatics

Research Frontiers in Metagenomic Microbiome Profiling Methods