The matched pair analyzer now supports the ability to be saved and restored
to disk or stream. This allows checkpoint backups during a long analysis run
or a long analysis run to be restored from disk for subsequent use without
the need to re-run the analysis (see OEReadMatchedPairAnalyzer
The ChEMBL18 solubility data has been re-indexed with the latest version of the
matched pair analyzer to include hydrogen-member transformations. The result is
a larger number of transformations and potentially a larger number of transformed
and annotated structures returned from OEApplyChEMBL18SolubilityTransforms.
For a 10K test set of structures, using OEMatchedPairContext::Bond0
resulted in an average of three times more output structures and
OEMatchedPairContext::Bond2 produced an average of 4.5 times the
number of output structures as the previous release when applying all transformations
by overriding the default value of minMMPThreshold parameter with 0.
Using the default value (5) for the minMMPThreshold parameter, the number of annotated
structures from the 10K test set compared to the previous version is approximately the same
for the OEMatchedPairContext::Bond0 context and approximately 25% less
structures for the OEMatchedPairContext::Bond2 context. The number of
structures produced is input dataset-dependent, so tuning of the minMMPThreshold
parameter is highly encouraged.
Structures returned from the application of matched pair transformations will now have
their coordinates cleared. Previously, newly added atoms were simply placed at the
origin, possibly affecting stereochemistry and generating odd depiction output.