EMP Release 1 — SILVA 138.2 NR99 remapped BIOM table and OTU cross-reference mapping

The Earth Microbiome Project (EMP) Release 1 closed-reference OTU table (Thompson et al. 2017, Nature 551:457) was built against SILVA 123 97% OTU centroids. SILVA 138.2 NR99 (released 2022) is the current recommended 16S reference, but OTU IDs changed substantially between releases. A direct ID join fails for ~53% of OTUs. This deposit provides a BLAST-derived cross-reference mapping from SILVA 123 OTU IDs to SILVA 138.2 NR99 IDs (≥97% identity, ≥80% query coverage), along with the remapped EMP Release 1 BIOM table. Dataset: emp_silva138_map.tsv : BLAST-derived cross-reference mapping from SILVA 123 OTU IDs to SILVA 138.2 NR99 IDs (≥97% identity, ≥80% query coverage). Coverage: 209,503 / 226,267 SILVA 123 OTUs (92.6%). emp_cr_silva_138.2_NR99.biom : EMP Release 1 BIOM table with feature IDs remapped to SILVA 138.2 NR99. Multiple SILVA 123 OTUs mapping to the same SILVA 138.2 centroid have their per-sample counts summed. 110,115 features across 97,902 samples; 89.9% read retention. emp_silva138_remap_report.json : Summary (feature counts, read retention, merge statistics). emp_silva138_remap.py : Standalone Python script to reproduce the remapping or apply it to other BIOM tables. Method: SILVA 123 97% centroid sequences were queried against the SILVA 138.2 SSURef NR99 database using blastn (-perc_identity 97 -qcov_hsp_perc 80). Limitations: Not an official EMP product. This is a community contribution. The EMP team has not validated or endorsed it. Unmapped OTUs (7.4%). 16,764 SILVA 123 OTUs could not be mapped at ≥97% identity / ≥80% coverage and are dropped by default. Downstream diversity metrics will be slightly deflated. Threshold choices are opinionated. The 97%/80% cutoffs are biologically motivated but not the only valid choice. Users with stricter requirements should re-run with --min-pident / --min-qcovs. Many-to-one merging is lossy. When multiple SILVA 123 OTUs collapse into one SILVA 138.2 centroid, per-sample counts are summed. Fine-scale diversity present in the original table may be lost. BLAST mapping, not curated alignment. This is sequence-similarity based, not a hand-curated taxonomy crosswalk. Edge cases exist.