### srd

#### SYNOPSIS

`srd [pc=<number of PCs; defaults to the number of PCs of the current PLS model>] \`

[type={WEIGHTS | LOADINGS; defaults to WEIGHTS}] \

[seeds=<number of seeds to be extracted; defaults to the smallest \

among the following three parameters:

1) 1/10 of the total number of X variables;

2) 1/2 of the number of active X variables;

3) 3000 > \

[critical_distance=<distance in Å
below which variables are assigned to the nearest seed>; defaults to 1.0 Å] \

[collapse={YES | NO; defaults to YES}] \

[collapse_distance=<distance in Å between two seeds
below which the respective groups are collapsed into one>; defaults to 2.0 Å]

#### DESCRIPTION

The`srd`

keyword is used to carry out a variable grouping procedure called
Smart Region Definition by Pastor et al. who originally implemented
it in GOLPE [1]. While a brief explanation of the meaning of the
different keywords is reported below, full details about the methodology
can be found in the original literature.`pc`

: number of PCs which will be taken into account to obtain PLS partial weights or PLS loadings (according to the value of the`type`

parameter). The initial picking of seeds, in fact, takes place on the basis of a D-optimal variable selection carried out either in the space of PLS partial weights or PLS loadings

`type`

: it indicates whether the D-optimal variable selection is accomplished in the space of PLS partial`WEIGHTS`

or`LOADINGS`

`seeds`

: number of variables initially picked by the D-optimal algorithm; it defaults to the smallest of the following criteria:

- 1/10 of the total
number of
*X*variables - 1/2
of the number of active
*X*variables - 3000

`critical_distance`

: it is the minimum distance in Å used to build Voronoi polyhedra;*X*variables are assigned to the nearest neighboring seed if their euclidean distance from the latter is smaller than the`critical_distance`

parameter

`collapse`

: according to the value of this parameter, Voronoi polyhedra built in the first stage of the algorithm are merged into larger ones (`collapse=YES`

) or left as they are (`collapse=NO`

).

`collapse_distance`

: if`collapse=YES`

, this is the smallest distance between two seeds below which the respective groups are merged into one

#### EXAMPLES

```
# the following commands
perform a Smart Region Definition variable grouping, operating the
initial D-optimal selection to identify the default number of seeds in
the space of PLS partial weights; 3 principal components are taken into
account, using the default critical distance (1.0 Å). No collapsing is
carried out after assigning variables to the respective seeds
```

srd
pc=3 type=WEIGHTS collapse=NO

# the following commands
performs a Smart Region Definition variable grouping, operating the
initial D-optimal selection to identify 1500 seeds in the space of
PLS loadings; 5 principal components are taken into account, using the
default critical distance (1.0 Å). Collapsing is carried out using a
2.5 Å collapse distance

srd pc=5 type=LOADINGS
seeds=1500 collapse=YES \

collapse_distance=2.5

#### REFERENCES

- Pastor, M.; Cruciani, G.; Clementi, S.
*J. Med. Chem.***1997**,*40*, 1455-1464. DOI