MAG set dereplication

Depending on the application, it may be necessary to dereplicate the set of MAGs to remove redundancy and retain only unique genome representatives. Our workflow includes a dereplication step that can use any genome distance matrix to find clusters of similar genomes (based on a specific similarity threshold) and identify the most representative MAG (in our case, it will be the longest genome in the cluster). Here we use Sourmash to generate the distance matrix but any other tool could also be used.

Compute MinHash signatures with Sourmash

In this step, Sourmash is used to compute MinHash signatures for the filtered MAGs. MinHash is a method used to compare large datasets efficiently by creating compressed representations of genomes.

mosh sourmash compute \
    --i-sequence-file ./cache:mags_filtered_50 \           
    --p-ksizes 35 \                                  
    --p-scaled 10 \                                                             
    --o-min-hash-signature ./cache:mags_minhash_50 \
    --verbose

Compare MinHash signatures

Here we compare the computed MinHash signatures to evaluate the similarity between the genomes. This step will allow for dereplication by identifying highly similar genomes.

mosh sourmash compare \
    --i-min-hash-signature ./cache:mags_minhash_50 \ 
    --p-ksize 35 \                                                          
    --o-compare-output ./cache:mags_dist_matrix_50 \
    --verbose         

Dereplicate MAGs

This step dereplicates the filtered MAGs, ensuring that only unique MAGs are retained. Dereplication reduces redundancy by merging similar genomes based on a similarity threshold.

mosh annotate dereplicate-mags \
    --i-mags ./cache:mags_filtered_50 \                    
    --i-distance-matrix ./cache:mags_dist_matrix_50 \       
    --p-threshold 0.99 \
    --o-dereplicated-mags ./cache:mags_derep_50 \   
    --o-feature-table ./cache:mags_ft_50  \
    --verbose