Metagenomes sequenced to profile prokaryotic communities also carry eukaryotic reads. The size and source of that signal limit how such profiles can be read: a eukaryotic fraction that varies by habitat lowers prokaryotic depth unevenly across samples. The quantitative evidence here comes from one project, Metadata Correlates of Eukaryotic Contamination in NMDC Prokaryote-Targeted Metagenomes. It used read-based taxonomic classifications from gottcha2 stored in the nmdc-results tables of nmdc 1.
Prevalence of the Eukaryotic Signal
The project examined 2,759 NMDC ReadbasedAnalysis runs from 9 studies. Of these, 77% carry detectable eukaryotic reads, with a median eukaryotic fraction of 2.7% and a mean of 13.3%. In 20% of runs the eukaryotic fraction exceeds 20%. Among runs with detectable eukaryotic reads, plastid (plant or algal chloroplast) DNA makes up a median 100% of the eukaryotic signal. Because the mean sits well above the median, the distribution of per-run fractions has a long upper tail. These are per-run summaries, not read-count-weighted totals, so they do not show what share of all eukaryotic reads comes from which runs. Only GOTTCHA2 gave a usable eukaryotic fraction. The report states that its absolute values depend on the reference database and should be read as relative or ordinal, not as calibrated absolute contamination 1.
Habitat-Specific Sources
The composition of the eukaryotic signal differs sharply by habitat 1:
- Freshwater (aquatic): 99.5% eukaryote detection, a plastid share of 1.00, dominated by algal plastid from phytoplankton chloroplasts 1.
- Soil (terrestrial): 55.7% eukaryote detection, a plastid share of 0.43, a mix of plant plastid and soil fungi or protists 1.
- Plant roots: 100% eukaryote detection, a plastid share of 0.03, dominated by non-plastid root-associated fungi or protists 1.
Interpretation: Co-Sampled Environmental DNA, Not Contamination
The report interprets the eukaryotic “contamination” in these metagenomes mainly as co-sampled photosynthetic environmental DNA rather than laboratory or animal-host contamination. In this reading, the signal comes from algal chloroplasts in freshwater, plant chloroplasts and soil fungi in terrestrial samples, and root-associated fungi or protists in plant samples. This is an interpretation of the source composition, not a direct test of contamination pathways. The habitat-level source split has not yet been confirmed within batch-controlled cohorts (see Open Directions), so it may be partly entangled with study-level differences of the kind discussed in study-batch-confounding-of-environmental-associations 1.
Literature Context
The project places its findings against published work. The items below are literature findings reported by the source, not project measurements 1:
- Eisenhofer, Alberdi & Woodcroft (2026, mSystems, PMID 41854267) report biome-specific prokaryotic fractions across 136,284 metagenomes. This is consistent with the habitat dependence observed here 1.
- Sobolev et al. (2025, IJMS, PMID 41373768) show that sample matrix, together with extraction kit, drives eukaryotic admixture. This supports a habitat or matrix explanation, but it also raises extraction protocol as a co-varying factor that the habitat comparison does not separate out 1.
- The plastid/chloroplast dominance is reported to match Chevokina et al. (2025, Front Plant Sci, PMID 41560914) and the plant-host-depletion literature (Wang et al. 2026, Plant Biotechnol J, PMID 41078118) 1.
- Anthony et al. (2024, Environ Microbiome, PMID 39095861) likewise attribute poor soil-metagenome resolution to plant/eukaryotic DNA 1.
Open Directions
- Extend the within-study design beyond soil: test, in the aquatic and plant-associated studies, whether the vegetation/geography effect seen within soil generalizes, and whether the split between algal plastid and root fungal sources holds within batch-controlled cohorts. This would close the gap between cross-habitat contrasts and study or batch effects 1.
- Separate matrix from extraction protocol: NMDC does not populate DNA-extraction kit, host-depletion or library-prep fields, so extraction kit currently remains an unmeasured residual. The first step is to get per-study wet-lab protocols from the original study records or publications, or to use a collection that records them. Eukaryotic fraction can then be modelled on habitat and kit jointly, to test the matrix-plus-kit driver cited from Sobolev et al. against the habitat-only reading 1.
- Classifier dependence: re-estimate plastid share on a subset of runs with a second read classifier and reference database. This would check whether the median 100% plastid share among detectable runs reflects the biology or the GOTTCHA2 reference database 1.