get_descriptor: get a descriptor (fingerprints and/or CDK physical...

In iqspr: Inverse Molecular Design

Description

Get a descriptor (fingerprints and/or CDK physical descriptors) from SMILES strings with possbility to request the scaling (for continuous descriptors, e.g. physical) or re-casting (for binary descriptors, e.g. fingerprints) of the output descriptors.

Usage

get_descriptor(smis = c("C1=CC=C(C=C1)O"), desctypes = c("standard"),
  scale = F, scale_init = F, mdesc = 0, sddesc = 1, quiet = F)

Arguments

smis	is a SMILES strings vector ("C1=CC=C(C=C1)O", canonical SMILES of a phenol by default).
desctypes	is a vector of characters defining the fingerprints and/or physical descriptors types to compute ("standard" by default). The actual entire list of available fingerprints: "standard", "extended", "graph", "hybridization", "maccs", "estate", "pubchem", "kr", "shortestpath" and "circular", and physical descriptors: "constitutional","topological","electronic" can be computed.
scale	sets to TRUE (FALSE by default) for scaling the physical descriptors only (i.e. continuous features) - mean = 0, s.d. = 1.
scale_init	sets to TRUE (FALSE by default) to keep in memory the means and s.d. related to each descriptor after a first scaling. Indeed, after the descriptors on a training set have been first computed, the mean and s.d. have to be kept fixed for future descriptors computation on test and/or validation sets. In this last case, the scale_init variable is set to FALSE.
mdesc	is a scalar (0 by default) or vector of means for a post-scaling of physical descriptors.
sddesc	is a scalar (1 by default) or vector of standard deviations for a post-scaling of physical descriptors.
quiet	keeps the console's outputs quiet if sets to TRUE (FALSE by default).

Value

the descriptor(s) with the associated means and standard deviations for scaling.

Examples

## Not run: 

descriptors <- get_descriptor(smis = "C1=CC=C(C=C1)O", desctypes = c("standard","topological"))


## End(Not run)

iqspr documentation built on Aug. 1, 2017, 9:02 a.m.