From defaults to databases: parameter and database choice dramatically impact the performance of metagenomic taxonomic classification tools
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
In metagenomic analyses of microbiomes, one of the first steps is usually the taxonomic classification of reads by comparison to a database of previously taxonomically classified genomes. While different studies comparing metagenomic taxonomic classification methods have determined that different tools are ‘best’, there are two tools that have been used the most to-date: Kraken ( k -mer-based classification against a user-constructed database) and MetaPhlAn (classification by alignment to clade-specific marker genes), the latest versions of which are Kraken2 and MetaPhlAn 3, respectively. We found large discrepancies in both the proportion of reads that were classified as well as the number of species that were identified when we used both Kraken2 and MetaPhlAn 3 to classify reads within metagenomes from human-associated or environmental datasets. We then investigated which of these tools would give classifications closest to the real composition of metagenomic samples using a range of simulated and mock samples and examined the combined impact of tool–parameter–database choice on the taxonomic classifications given. This revealed that there may not be a one-size-fits-all ‘best’ choice. While Kraken2 can achieve better overall performance, with higher precision, recall and F1 scores, as well as alpha- and beta-diversity measures closer to the known composition than MetaPhlAn 3, the computational resources required for this may be prohibitive for many researchers, and the default database and parameters should not be used. We therefore conclude that the best tool–parameter–database choice for a particular application depends on the scientific question of interest, which performance metric is most important for this question and the limit of available computational resources.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it