FTrees is a highly efficient software tool for fuzzy similarity searching facilitating virtual HTS. The ability to detect novel molecular scaffolds is one of several features that charaterizes FTrees capabilities.
Its underlying topological descriptor (the Feature Tree) captures connectivity and physico-chemical properties of functional groups. The optimum similarity of two descriptors is defined by an alignment, so an SAR may be readily detected.
What puts FTrees one step ahead of other software in a typical similarity-based virtual screen?
The first thing about the results obtained with FTrees from a chemist's point of view is that the molecules are aligned. That means the chemist can understand why the molecules are awarded a certain similarity. Also, the chemist can influence the alignment — do I want to pick out molecules with a common substructure or molecules with common end groups, similar sized molecules or fragments?
OK, but I want to carry out more complicated experiments for example, I have several active molecules displaying some diversity, what now?
FTrees is ideal for this scenario. We have already seen how it is possible to align molecules. Let's align the actives; they must have something in common to make them all active. This is possible with the FTrees model building facility. A model is built from a set of several aligned molecules and is in fact itself just another Feature Tree.
So my model tree describes the unified features of my actives but what is this good for?
You can screen with the model Feature Tree just as easily as with a Feature Tree representing a single molecule. That will give just one result set instead of several. The results will highlight compounds that best match the common description of the set of actives.
I work on a daily basis with the in-house database — this contains millions of compounds these days, where do I start?!
Millions of compounds represent no problem to FTrees — 2 million Feature Trees can be handled at once (in memory) by the software. In fact, FTrees is perfect for carrying out data reduction of large datasets. Starting with a small set of seed compounds, you can use FTrees to filter the dataset down to a size more manageable for more intensive modeling methods.
What about fragment based design?
This is possible with an add-on called FTrees-FS.