LSP Cheminformatics Toolkit

• LSP Cheminformatics Toolkit
  • Components
    • LSP Cheminformatics Server
    • R client
  • Installation
    • Installing Anaconda
    • Creating the environment and installing dependencies
    • Installing the LSP Cheminformatics tools
  • Running the server
  • Querying the server
    • Installing the R client
    • Query for similarities
    • Query for substructure matches
  • Additional functionality
  • Funding

Components

LSP Cheminformatics Server

The server runs in the background and performs the requested computations. It can be queried using a JSON API.

LSP Cheminformatics Server JSON API documentation

R client

The R client provides easy access to the functionality provided by the Cheminformatics server.

R client documentation

Installation

Installing Anaconda

We recommend installing this package in a separate Anaconda environment. If Ananconda is not available, first install Anaconda.

Creating the environment and installing dependencies

Once anaconda is installed, create a new environment for this package and install RDKit (see here for more help).

conda create -c rdkit -n lspcheminf_env python=3.7 rdkit click flask pandas gunicorn marshmallow apispec
conda activate lspcheminf_env
conda install -c conda-forge molvs

Installing the LSP Cheminformatics tools

conda activate lspcheminf_env
pip install --no-deps 'git+https://github.com/labsyspharm/lsp-cheminformatics.git#egg=lspcheminf&subdirectory=lsp_cheminf_server'

To upgrade to a newer version if the LSP Cheminformatics tools are already installed:

conda activate lspcheminf_env
pip install --no-deps --upgrade 'git+https://github.com/labsyspharm/lsp-cheminformatics.git#egg=lspcheminf&subdirectory=lsp_cheminf_server'

Running the server

The JSON API is exposed by running the server in the background. The following command runs the server on port 8000.

conda activate lspcheminf_env
gunicorn --workers=4 -b 127.0.0.1:8000 -t 600 lspcheminf

Once this command is issued the server will continue running until it is manually stopped using the Ctrl-C key combination.

While the server is running the documentation for the JSON API is available at http://127.0.0.1:8000/doc.

Querying the server

Any software capable of sending and receiving JSON can be used to query the server. For convenience, there is an R client that implements the JSON API in simple functions. In this example we use the R client to query for the chemical similarities between a number of compounds.

Installing the R client

devtools::install_github("labsyspharm/lsp-cheminformatics", subdir = "lsp_cheminf_rclient")

The client only functions while the lspcheminf server is running.

Query for similarities

The R client can be used to query the chemical similarity between compounds. By default the Morgan fingerprinting algorithm is used.

library(lspcheminf)
chemical_similarity(
  c("resveratrol" = "InChI=1S/C14H12O3/c15-12-5-3-10(4-6-12)1-2-11-7-13(16)9-14(17)8-11/h1-9,15-17H/b2-1+"),
  c(
    "tofacitnib" = "InChI=1S/C16H20N6O/c1-11-5-8-22(14(23)3-6-17)9-13(11)21(2)16-12-4-7-18-15(12)19-10-20-16/h4,7,10-11,13H,3,5,8-9H2,1-2H3,(H,18,19,20)/t11-,13+/m1/s1",
    "aspirin" = "InChI=1S/C9H8O4/c1-6(10)13-8-5-3-2-4-7(8)9(11)12/h2-5H,1H3,(H,11,12)"
  )
)
#> # A tibble: 2 x 3
#>   query        score target
#>   <chr>        <dbl> <chr>
#> 1 resveratrol 0.0571 tofacitnib
#> 2 resveratrol 0.132  aspirin

If we want to compute the similarity based on topological fingerprints instead and pass some custom arguments to the RDKit function we can do this:

chemical_similarity(
  c("resveratrol" = "InChI=1S/C14H12O3/c15-12-5-3-10(4-6-12)1-2-11-7-13(16)9-14(17)8-11/h1-9,15-17H/b2-1+"),
  c(
    "tofacitnib" = "InChI=1S/C16H20N6O/c1-11-5-8-22(14(23)3-6-17)9-13(11)21(2)16-12-4-7-18-15(12)19-10-20-16/h4,7,10-11,13H,3,5,8-9H2,1-2H3,(H,18,19,20)/t11-,13+/m1/s1",
    "aspirin" = "InChI=1S/C9H8O4/c1-6(10)13-8-5-3-2-4-7(8)9(11)12/h2-5H,1H3,(H,11,12)"
  ),
  fingerprint_type = "topological",
  fingerprint_args = list("minPath" =  2, "useHs" = FALSE)
)
#> # A tibble: 2 x 3
#>   query        score target
#>   <chr>        <dbl> <chr>
#> 1 resveratrol 0.0967 tofacitnib
#> 2 resveratrol 0.198  aspirin

Query for substructure matches

match_substructure(
  c("secondary_amine" = "[H]N(C)C"),
  c(
    "tofacitnib" = "InChI=1S/C16H20N6O/c1-11-5-8-22(14(23)3-6-17)9-13(11)21(2)16-12-4-7-18-15(12)19-10-20-16/h4,7,10-11,13H,3,5,8-9H2,1-2H3,(H,18,19,20)/t11-,13+/m1/s1",
    "aspirin" = "InChI=1S/C9H8O4/c1-6(10)13-8-5-3-2-4-7(8)9(11)12/h2-5H,1H3,(H,11,12)"
  ),
  query_identifier = "smiles"
)
#> # A tibble: 1 x 3
#>   match      query           target
#>   <list>     <chr>           <chr>
#> 1 <list [6]> secondary_amine tofacitnib

Six matches for secondary amine groups where found in tofacitnib. The atom indices for each match are stored in the match list column.

Additional functionality

At the moment there is no R client code yet for the additional functionalities of the JSON API, like the molecule drawing, ID conversion etc.

Funding

This work was supported by NIH grants U54-HL127365, U24-DK116204 and U54-HL127624.