Primer bias in microbial diversity studies

The most intriguing result is that all these innovative, high-throughput molecular methods have their serious inherent biases. For example, amplicon-based methods depend strongly on taxonomic resolution of the barcode (DNA fragment used for identification), primer-template mismatches, the presence of introns and the overall length of the barcode; but conversely, PCR-free methods are affected strongest by the availability of taxonomic reference information, which differs enormously for fungal classes and phyla.

Motivated by criticism of a Science article on global fungal diversity, where the abundance of certain fungi were suspected of being underestimated, the authors went back to the lab to test if their initial approach may have involved a primer bias, i.e. distorted taxon distribution as a result of imperfect primer matches with some templates in their application of metabarcoding. 

What a great example of serious science. The authors took the criticism, went back to the drawing board, did experiments and published their findings in two journals (as response in Science, and the full study in MycoKeys) even though those confirmed doubts about some results in their earlier paper. There is no need to retract any paper but the colleagues re-checked some of their results and had the guts to write about it. Their new findings are immensely important for the community at large which makes this even more commendable. Some might wonder why I am so excited by this. Well, in times where the amount of retracted papers in big journals increases and the reasons for this are mostly rather embarrassing for the scientific community (e.g. falsification, non-reproducible, methodologically flawed), every example that shows how it is supposed to be done is more than welcome. Also, quite often responses and comments in scientific journals are more representative of personal conflicts between scientists than constructive dialogue meant to advance our knowledge.

And the results? 

The colleagues demonstrated substantial methodological biases in soil fungal diversity. High-throughput Illumina sequencing of DNA metabarcodes and the whole soil metagenome revealed strong methodological biases in taxonomic insights into soil fungal diversity. All methods had their inherent biases and shortcomings, but reached roughly similar ecological conclusions indicating the greatest role of floristic variables on soil fungal communities in the mountainous Papua New Guinea. 

This study demonstrates that PCR-free metagenomics and amplicon-based approaches perform in a comparable fashion in recovering major fungal classes in spite of certain statistical differences. Within the amplicon data set, barcode-primer pair combinations differed strongly in recovering relative abundance of fungal classes and OTU richness (see also Tedersoo et al. 2015b). Nonetheless, these were all in agreement about trends in OTU richness and disentangling the key drivers of community composition. We found no evidence for reduced statistical performance in barcodes with relatively conserved sequences, but the use of conserved barcodes seriously hampers biological relevance of the data due to the inability to approximate species level (or explicitly any taxonomic level) and to assign functional categories such as trophic status (except arbuscular mycorrhizal mutualism). Considering the taxonomic resolution and primer bias, we recommend targeting the ITS2 barcode when using the current HTS [High throughput] technologies that permit <700 bases of high-quality reads.