Lecture 3: Chemical informatics primer

Introduction:

Tutorial inspured by this paper: Chemoinformatic Analysis of Combinatorial Libraries, Drugs, Natural Products and Molecular Libraries Small Molecule Repository: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2686115/

.. compared the properties of molecules found in four different chemical libraries.

The authors did several types of analysis. First, they generated a bunch of different features or “descriptors” for each molecule. A descriptor can be any property of the molecule, from trivial stuff like molecular weight and charge. Then they did a principal components analysis on the molecules and then scatterplotted the molecules in each library versus the set of FDA-approved drugs. Results: most of the molecules in these libraries occupy a chemical space quite distant from the approved drugs… therefore, not therapeutically active.

Important databases:

• PubChem

• DrugBank

• ChemBank

• ChEMBL

Getting started

R packages:rcdk and rpubchem. rJava is needed for visualization.

install.packages(c("rJava", "rcdk")) # huge, takes a few mins to download

library(rcdk) # docs at http://cran.r-project.org/web/packages/rcdk/rcdk.pdf, paper at http://www.jstatsoft.org/v18/i05/paper

Create a molecule object

Next I sought to manually create a Java object for a molecule from its SMILES. SMILES is a systematic, unambiguous plain-text representation of molecular structure. I found anle138b‘s SMILES in its PubChem entry.

# parse.smiles accepts a vector of SMILES strings and returns a list of type AtomContainer,
# containing items of type IAtomContainer
# if you have just one molecule of interest, just grab the first item with [[1]]
anle138b <- parse.smiles("C1OC2=C(O1)C=C(C=C2)C3=CC(=NN3)C4=CC(=CC=C4)Br")[[1]]

Exercise: Parse acetaminophen

Search for acetaminophen in PubChem, get the SMILES format: https://pubchem.ncbi.nlm.nih.gov/compound/acetaminophen Parse it in rcdk

ac <- parse.smiles("CC(=O)NC1=CC=C(C=C1)O")[[1]]

View/draw it

The next thing I wanted to do is plot it. The draw.molecule.2dfunction draws it in a separate Java window:

# draws it in a separate Java window
view.molecule.2d(anle138b)

Exercise: Draw acetaminophen

Get descriptors

The next thing I learned about rcdk is how to get these descriptors of molecules. rcdk comes with access to a whole list of descriptors which you can get like so:

```r
descriptors <- get.desc.names(type = "all")

This gave me a vector of character strings, the names of fifty descriptors you can calculate on your molecule. I figured a good one to try out would be the number of strikes a molecule has against it as a drug according to the Rule of Five:

# get one descriptor, for instance Rule of Five descriptors http://en.wikipedia.org/wiki/Lipinski's_rule_of_five
eval.desc(anle138b, "org.openscience.cdk.qsar.descriptors.molecular.RuleOfFiveDescriptor")

Exercise: Other properties

Try to get some properties of a molecule using rcdk’s base functions, like get.total.charge, get.volume, get.exact.mass

get.exact.mass(ac)

Another way of getting drugs

Load it as sdf

# Download the chemical structure in SDF format from PubChem (Diclofenac)
dic <- load.molecules("data/Structure2D_CID_3033.sdf")

Evaluate the new target drug

sapply(descriptors, eval.desc, molecules = anle138b)