Training, testing, and blind datasets used for machine learning algorithms for taxonomic classification of marine metagenomes: K12.kmers.txt - 12bp k-mer vocabulary constructed by Jellyfish v1.1.11 from 47,894 genomes in GTDB release 202 MarRef_1.6.tsv - Metadata file downloaded from MarRef v1.6 MarRef.genustrain.fasta - Training set from MarRef v1.6 (seed=808) used for genus classification MarRef.genustest.fasta - Testing set from MarRef v1.6 (seed=747) used for genus classification MarRef.speciestrain.fasta - Training set from MarRef v1.6 (seed=808) used for species classification MarRef.speciestest.fasta - Testing set from MarRef v1.6 (seed=747) used for species classification MarRef.traintest.key.tsv - Table containing MarRef accession, GenBank accession, GenBank taxonomy ID, taxonomic information, and labels used for species and genus testing and training anonymous_reads_*.fq - Blind datasets (1-10) in interleaved fastq format reads_mapping_*.tsv - Key for blind datasets 1-10. Each sequence header is mapped to its corresponding MarRef accession and NCBI taxonomic ID.