3.7. Gene Markers for Diatoms

• Gap in the reference database [16,27,32,34,38,46,62,82,90,99], etc.
• The natural intraspecific and intragenomic variabilities of the barcoding marker (single taxon has multiple genotypes at the barcoding region, and members of that taxon might cluster into different Molecular Operational Taxonomic Units (MOTUs)) [35].
• Cryptic diversity—a single morphological species can represent different genetic groups (e.g., diatoms Sellaphora pupula, Pinnularia borealis, Hantzschia amphioxys and Nitzschia inconspicua and species of Stichococcus, Coccomyxa, Chlorokybus, Cryptomonas, etc.) [78,137,138,139,140,141,142].
• MOTU richness can be artificially inflated through technical errors at different steps of sample processing during amplification and sequencing [35].
• The MOTU delimitation approach influences the richness estimation and interpretation [35] (and references in it) (assessment of the bioinformatics pipelines provided in [14,15]).
• Complete absence of amplification on the whole due to a mismatch of the primer set used. For example, Salmaso et al. [27] did not find any species belonging to the Euglenales in the HTS results (with universal eukaryotic primers (TAReuk454FWD1 and TAReukREV3) for V4 18S), although they were present in LM. Hanžek et al. [66] reported that the taxa that contributed most to the biomass (Actinotaenium/Mesotaenium sp. and the species Cosmarium tenue, Pantocsekiella comensis, Sphaerocystis schroeteri and Synedropsis roundii) were not identified by eDNA metabarcoding (V9 18S region was amplified using the universal primer pair 1391F and EukB). Proeschöld and Darienko [140] noted that, although Stichococcus-like organisms are widely distributed in almost all habitats, they are not recorded in environmental studies based on HTS approaches, because the V4 or V9 regions of the SSU contain introns that obstruct amplification. Groendahl et al. [42] reported that Monorhaphidium sp., Selenastrum sp. and Trachelomonas sp. detected using the morphology-based approach were not identified by the metabarcoding approach, despite the fact that all three genera are included in the reference database.
• Uncertainties and lack of sensitivity of reference databases for the selected DNA markers [27].

• Diatom extracellular skeletons are counted in LM even if they come from dead cells. The valves of dead cells can be transported from locations other than the target assemblage. Metabarcoding will not detect these dead cells [35,78].
• The proportion of live diatoms found in environmental samples varies greatly, ranging from 2 to 98% [35].
• Small-celled species and pico-sized cells are often overlooked or underestimated by the morphological approach. For example, the valves of Fistulifera saprophila tend to dissolve during sample processing, which can explain why this species is often missed during morphological identification [78,80,82,99].
• LM misidentification (false LM positives) [27,78,99].
• Differences in the detection limits of the two methods: morphological and molecular approaches do not give the same insight into communities of algae and, therefore, do not have the same detection capacity for species [27,92,99].
• The different sample volumes settled for microscopy and metabarcoding [143].
• Underlying units used for microscopy (individual cells) and those used for metabarcoding (ASV sequences) are quite different, making direct comparisons imperfect [24,143].
• A short barcode gene fragment may have limited the taxonomic resolution [143]. For example, the resolution of the V4 18S region does not allow to unambiguously identify some species of Navicula [32]. For the V7–9 18S marker, a lack of intergenus taxonomic resolution was found (the MOTUs matched multiple genera, e.g., Alexandrium pseudogonyaulax and A. hiranoi, Chaetoceros neogracile and C. curvisetus and Thalassiosira eccentrica and T. antarctica) [144]. In some Chlamydomonas, the V9 region is very similar to that of prasinophytes clade VII A5 [122].