deletions | additions       diff --git a/Discussion.tex b/Discussion.tex  index e95eb28..b04bbc1 100644  --- a/Discussion.tex  +++ b/Discussion.tex 
...
\section*{Discussion}  The MUV dataset, with its focus on eliminating analogue bias, is particularly resistant to single-query shape-based virtual screens \cite{Tiikkainen_2009}. This is reflected in our overall results, shown in Figure~\ref{aucs}, where only two targets (Rho and PKA) achieve AUCs where the 95\% confidence interval does not overlap with 0.5 (random performance). The remaining targets likely lack meaningful whole-molecule shape complementary between the query ligand and the active compounds of the benchmark. One exception may be HIV-rt, where this is clear early enrichment, indicates that a subset of the actives may be compatible with the query molecule.  FOMS dramatically outperformed other methods for the Rho and PKA targets due to correct positioning of a fragment with key, conserved interactions.FOMS provides a rapid means of template docking \cite{Ruiz_Carmona_2014,abagyan2015icm} with shape-based scoring. Choice of fragment critical  Shape constraints generally tracked or improved upon the performance of FOMS similarity ranking. For cathg, shape constraints provided the most meaningful virtual screening result as they could select for essentially provides  a subshape rapid means of template docking \cite{Ruiz_Carmona_2014,abagyan2015icm} using shape-based scoring. The disadvantage  of the query ligand while the other methods must rank the full shape. fragment-oriented approaches  The primary advantages of shape-based fragment alignment search are threefold. Biasing the alignment to the desired fragment position. This is consistent with other results that demonstrate the importance of adding pharmacophoric properties (or `color') to shape similarity. Fragment alignment introduces a hard bias toward the specific fragment alignment without introducing any additional computation or calculation, as with color methods. In fact, as we have shown, pre-alignment substantially improves performance, which is the second main advantage. Prealignment, whether to fragments (FOMS) or canonical internal coordinates (VAMS) is orders of magnitude faster. This holds true even if the cost to create the search database is taken into account. The time to create the databases scales with the number of molecular shapes (about 10 shapes a second on our system) and compares favorable with RDKit search (2 molecules a second).  Choice of fragment critical  Shape constraints generally tracked or improved upon the performance of FOMS similarity ranking. For cathg, shape constraints provided the most meaningful virtual screening result as they could select for a subshape of the query ligand while the other methods must rank the full shape.  The expected use case for these algorithms is for scenarios where a number for fragment-oriented databases are created to enable repeated searches by multiple users investigating multiple targets, in which case the cost of database creation essentially gets amortized into insignificance.  Shape constraints provide a novel and unique method for specifying molecular shape queries. Since they are fragment oriented, they can be used to perform partial shape similarity search, a generaly challenging problem (cite). Shape constraint searches generally took well under a second making them perfect for interactive applications. Although we have investigated automatically generated shape constraints, our assumption is that intelligently designed constraints created by human experts guided by interactive analysis of a virtual screening database would substantially outperform our interaction point based constraints. This hypothesis remains untested as intuitive interfaces that allow expert users to sculpt their desired molecule need to be developed and integrated into a comprehensive virtual screening evaluation environment.