Kmer preclustering

[AroneyS]

• Allow independent clustering
• Add sourmash to Ibis env
• Filter target taxa
• Generate kmers from unbinned window sequences
• Kmer cluster into <1000 sample clusters
• Minimum cluster size: all clusters smaller get combined? - No, since should be cutting quite high on dendrogram so anything in a small cluster at that level is very distinct from other clusters.
• --max-precluster-size default 1000
• --kmer-precluster [never, large, always] - large default with >1000 samples
• Iterative clustering, removing largest cluster <1000 each iteration (continue when remaining # samples <1000)
• Update evaluate to use separate edges/targets outputs

Test

• 10,000 samples
• 200,000 samples

Final cutoff precision: 0.01