Biomedical Engineering Reference
In-Depth Information
input fi le. Assuming that this contains the pre-existing structures, all that is
left after this node are the new structures. They are fi nally written into an
SD fi le. The middle row of the workfl ow likewise saves all duplicate
structures, whereas the bottom row saves all problematic structures that
were gathered during the various preparation steps into another fi le
(removing and resorting some properties for convenience before they are
actually written to the fi le).
6.4.2 Enumerating amides with RDKit
The next workfl ow deals with reaction simulation, in particular the
formation of amides by a two component reaction of amines and acids.
Similar workfl ows are in use at Novartis and other pharmaceutical
companies. This workfl ow makes use of the RDKit [5] nodes. A lot of
functionality is provided from this well-known open source toolkit in
KNIME. The concepts of quickforms and variables are also introduced
by this workfl ow. The workfl ow is shown in Figure 6.9.
First, an SD fi le containing several amines and acids (in any order) is
read in. Next, a meta-node encapsulates some preprocessing of the
workfl ow. Specifi cally, it splits the input fi le into amines and acids and
also reduces their number based on the user selection (e.g. three amines
and three acids). Each group is then output to one of the two ports and
passed to the Two Component Reaction node. This node simulates
pairwise reactions between all molecules from both input ports and
outputs a table with the resulting products. As the products are stored in
the internal RDKit format, they need to be converted back to SDF before
they can be written into a fi le (upper branch). A PDF fi le containing 2D
depictions of all products is created in the lower branch. In the process,
2D coordinates are generated fi rst and then the column with the 2D
depictions is transformed into a grid. This means that all cells in the
Figure 6.9
Amide enumeration workfl ow
Search WWH ::
Custom Search