two (nonexclusive) methods for dealing with singletons:
–writing them out to a separate file.
–If singleton rescue is used (option -RESCUE_SIZE),
rescued singletons will appear
in clusters to which they are rescued.
–If a singleton file is generated (option
-SINGLETON_FILE), it may be fed
back to nearneighbors and then reclustered.
an additional processing option which is not part of the original Jarvis-Patrick algorithm.
–This option (-NN_BEST_THRESHOLD) allows the
preprocessing of the neighbours
lists as follows:
•the best neighbour (excluding itself) for each structure
is compared with the threshold
value. If the best neighbour has a similarity lower than the specified threshold, then the structure is marked as a
singleton and is excluded
from the clustering. This is a useful way to discover very tight(?) clusters within a dataset.