You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When dereplicating sequences, it is crucial to cluster genomes that significantly overlap each other (i.e., high bidirectional coverage) to avoid clustering sequences where one is entirely contained within the other. For example:
In the example above, it is there's no combination of --cov and --ani that prevent these two sequences from being clustered together. Although setting a high --tani cutoff could resolve this specific case, it would not be effective in more complex scenarios. For instance, if the goal is to connect pairs of genomes with a minimum bidirectional coverage of 75% and no minimum ANI, using --tani wouldn't work.
A related question: how does Clusty handle multiple edges between node pairs? For instance, with --metric ani --ani 0.1, both edges in the example would pass the filters. In such cases, how does Clusty determine the clustering weight? Does it use the maximum ANI value, the mean, the first/last value in the file, something else?
The text was updated successfully, but these errors were encountered:
Thank you for the great suggestion! In Vclust v1.1.1, we've introduced query and reference coverage (qcov and rcov) in the output, allowing for bidirectional coverage filtering to handle cases like the one you described.
When it comes to clustering with multiple edges between node pairs, Clusty selects the maximum value for clustering (e.g., ANI if --metric ani). All clustering algorithms in Clusty, except for the Leiden algorithm, are threshold-based and do not rely on weights (i.e., a genome is clustered if it meets the similarity threshold with a centroid, closest, or furthest member).
For the Leiden algorithm, if you want to include coverage information in the edge weight, you can use the gani metric. This is calculated as the number of identical nucleotides divided by the length of the query sequence (i.e., ANI multiplied by query coverage), which aligns with the approach used by IMG/PR.
When dereplicating sequences, it is crucial to cluster genomes that significantly overlap each other (i.e., high bidirectional coverage) to avoid clustering sequences where one is entirely contained within the other. For example:
In the example above, it is there's no combination of
--cov
and--ani
that prevent these two sequences from being clustered together. Although setting a high--tani
cutoff could resolve this specific case, it would not be effective in more complex scenarios. For instance, if the goal is to connect pairs of genomes with a minimum bidirectional coverage of 75% and no minimum ANI, using--tani
wouldn't work.A related question: how does Clusty handle multiple edges between node pairs? For instance, with
--metric ani --ani 0.1
, both edges in the example would pass the filters. In such cases, how does Clusty determine the clustering weight? Does it use the maximum ANI value, the mean, the first/last value in the file, something else?The text was updated successfully, but these errors were encountered: