Filter placements by placement distance + Deal with multiple placement locations for a single query

[Robaina]

Implement filter to remove placements located too far away from insertion point / closest reference sequences in the tree

[Robaina]

Currently considering these options to deal with multiple placement locations for a single query:

1. Check whether all placement locations lay within a predefined cluster. In that case, results won't be affected since they are cluster-dependent (taxonomy assignment by gappa examine assgin would prob. have to be re-evaluated)

2. Test gappa's subcommand accumulate (https://github.com/lczech/gappa/wiki/Subcommand:-accumulate)

NOTE: these measures are independent of filtering placements by placement distance

[Robaina]

SOME NOTES:

1. gappa examine assign takes into account pendant length when passing "--distant-label" in version v.0.8.0 (default in previous versions). DISTANT is a new label that gets LWR assign proportionally based on the pendant length. Thus we could filter out placements based on the LWR assigned to DISTANT (above a given percentage threshold would imply query is not well defined/removed from counting)
2. We could filter out placements based on pendant length directly in the .jplace file. We could set a threshold based on ratio to tree diameter (max distance between two leaves)

[Robaina]

Alright, so filtering by pendant length already implemented... there are three options:

1. Filter directly my maximum pendant length
2. Filter by maximum pendant to distal length ratio
3. Filter by maximum pendant to tree diameter ratio

Tree diameter: computed as maximum distance between any two pairs of leaves (tips) in the tree

For a description of distal and pendant length see: https://github.com/lczech/gappa/wiki/Subcommand%3A-assign

[Robaina]

ANOTHER NOTE

gappa version >= 0.8.0 for newest code to work. Latest gappa version available in conda: https://anaconda.org/bioconda/gappa.

However, I couldn't get the latest version from the install command. Instead, I had to download the file from here: https://anaconda.org/bioconda/gappa/files and install directly (conda install "filename")

[Robaina]

Ok, this issue fixed by PR #73

[Robaina]

Reopening this issue because we realized that filtering out placements which were assigned to more than one cluster should be a default behavior of labelplacements.py, otherwise cluster scores are not properly assigned in those cases (at least impose this behavior when cluster_scores are provided.

The next PR will include this filter by default