Why two stool sequencing methods can disagree about the same sample

Two assays that answer different questions

The cheaper approach amplifies one conserved bacterial gene and uses variation within it to sort sequences into groups of related organisms. The more expensive approach sequences fragments of all the DNA present, which allows finer identification and some inference about genetic capability. The first cannot generally resolve species or strain, and it says nothing directly about which functions the community can perform.

The second produces far more data, costs considerably more to run, and requires substantially heavier computation to interpret. A report built on one method is therefore not directly comparable with a report built on the other, even for the same person on the same day.

Where the primer choice decides the answer

Marker-gene sequencing amplifies only a region of the gene, and different laboratories choose different regions for defensible reasons. The primers used to amplify a region match some organisms better than others, so certain groups are systematically over-represented or missed. Comparing results across laboratories that chose different regions is closer to comparing two different assays than to comparing two measurements.

This is why apparently contradictory findings about a particular bacterial group sometimes reflect methodology rather than biology. Well-conducted work states the amplified region and the primer sequences explicitly, and consumer-facing reports frequently omit both of those details entirely.

Extraction and the tough-cell problem

DNA has to be released from cells before anything can be sequenced, and bacterial cell walls differ enormously in how readily they break. Gentle extraction under-represents organisms with tough walls, while aggressive mechanical disruption shears DNA and creates its own artefacts. Because no protocol is neutral, the extraction step alone can shift the apparent proportions of major groups substantially.

Sample handling contributes too, since time and temperature before freezing allow some organisms to continue growing in the container. Standardisation efforts exist for exactly this reason, and adoption across commercial testing services remains uneven enough that protocols differ from one provider to the next.

Compositional data and the ratio trap

Sequencing returns proportions rather than absolute counts, because the total number of reads is set by the machine and not by the sample. That means a genuine increase in one organism necessarily lowers the reported proportion of every other organism present. Reading such a decrease as a real loss is a well-known statistical error, and it appears constantly in popular microbiome interpretation.

Methods exist to handle compositional data properly, and they change conclusions relative to naive comparison of proportions. Obtaining absolute quantities requires an additional measurement step, using cell counting or a spiked internal standard, that most routine pipelines simply do not include.

Reading a microbiome report sensibly

The organisms named in a report are matched against a reference database, and database coverage is uneven across the tree of life. Organisms absent from the reference are either misassigned or discarded, so a report describes what the database knows as much as what the sample contains. Interpretive layers that rate a result as good or bad rest on associations from research populations that may not resemble the person tested.

There is at present no agreed definition of a healthy gut community, which makes any scored verdict an editorial choice rather than a finding. Digestive symptoms that persist deserve clinical assessment, since a sequencing report cannot diagnose the conditions that actually explain them.