Current techniques to determine the composition of the microbiome are inadequate, say Belén Serrano-Antón et al. in PLOS ONE. As a result, much research in this area is unreliable.
The most commonly used methods to study the microbiome are whole genome sequencing (WGS) and amplicon techniques, such as polymerase chain reaction (PCR). In both methods, researchers compare DNA sequences with the sequences available in databases. And therein lies the problem, according to the Spanish researchers: especially when these databases are incomplete, it is impossible to correctly characterize all bacteria in the microbiome.
To test the reliability of WGS and 16S rRNA-PCR, Serrano-Antón et al. created virtual microbiomes that mimic bacterial populations found in humans, insects, soil, water, etc. The researchers analyzed these microbiomes using both techniques.
Depending on the method used, the researchers characterized different bacterial species when they analyzed the same microbiome. The overlap between 16S and WGS was less than 50 percent at the species level, dropping to less than 10 percent when the database was incomplete. Remarkably enough, a double identification – by both 16S and WGS – did not guarantee the actual occurrence of a species in the virtual microbiome. The number of false positives increased when the researchers used an incomplete database.
To test whether the predictions of the virtual microbiomes also apply to physical microbiomes, the researchers analyzed the microbiome of wax moth larvae (Galleria mellonella) with WGS and 16S. Here too, they found significant discrepancies between the two methods in the characterization of the bacteria.