admetshiny

An open-source R package and Shiny application for the management, calculation, filtering, visualization and exploratory analysis of molecular descriptors and ADMET properties of small molecules.

Overview

admetshiny integrates cheminformatics and bioinformatics workflows with an intuitive dashboard to support the prioritization of compounds in early-stage drug discovery. The application is organised in two complementary modules:

• CDK & webchem — retrieve canonical SMILES from PubChem (via the webchem package), enter them manually, or upload them as a CSV, then compute nine physicochemical descriptors locally with the Chemistry Development Kit (CDK).
• ADMET Master Manager — upload any CSV or Excel (.xlsx) ADMET dataset and manually map its columns to the application’s 20-field standard schema. Missing descriptors are back-filled from SMILES via CDK when available.

Both modules share the same drug-likeness filters, BOILED-Egg model, P-gp substrate classifier and 14-chart catalogue.

Features

• Drug-likeness filters with the original publication thresholds: Lipinski (1997), Veber (2002), Ghose (1999), Egan (2000), Muegge (2001).
• BOILED-Egg model (Daina & Zoete, 2016) for gastrointestinal absorption and blood-brain barrier permeability, with dual WLOGP / ALogP polygon systems.
• P-glycoprotein (ABCB1) substrate prediction via an embedded Random Forest classifier (100 trees, 5-fold CV: 69.4% accuracy, MCC = 0.383).
• Chemical-space visualizations: PCA, t-SNE, UMAP, parallel coordinates, violin, radar, correlation heatmap, cluster heatmap and Tanimoto / AGNES structural-similarity clustering.
• Dark mode, custom colour palettes and high-resolution figure export (PNG, PDF, SVG, TIFF up to 1200 DPI).
• Comprehensive HTML / PDF / Word report generation.

Installation

You can install the development version from GitHub:

# install.packages("remotes")
remotes::install_github("xavierclementegarcia/admetshiny")

The optional CDK-based descriptor calculation requires the rcdk package, which in turn requires Java JDK (a JRE alone is not sufficient).

Usage

Launch the interactive application:

admetshiny::run_app()

Use the computational functions programmatically:

library(admetshiny)

# Compute CDK descriptors for a few SMILES and apply the Lipinski filter
smiles <- c("CCO", "CC(=O)OC1=CC=CC=C1C(=O)O", "CN1C=NC2=C1C(=O)N(C(=O)N2C)C")
desc  <- calcCDKDescriptors(smiles)
desc  <- mapCDKDescriptors(desc)               # adds #violations and ADMET cols
filtered <- applyFilters(desc, filters = c("Lipinski", "Veber", "Ghose"))

# Plot the BOILED-Egg
plotBoiledEgg(filtered)

# Or normalize any external ADMET dataset via manual column mapping
# (returns the standard schema with #violations + ADMET properties):
# d <- read.csv("my_admet.csv", check.names = FALSE)
# mapping <- setNames(c("SMILES", "MW", "LogP", "TPSA"),
#                     c("CanonicalSMILES", "MW", "iLOGP", "TPSA"))
# d <- mapADMETColumns(d, mapping, calculate_cdk = TRUE)

References

• Lipinski, C. A., et al. (1997). Advanced Drug Delivery Reviews, 23(1-3), 3-25.
• Ghose, A. K., et al. (1999). J. Combinatorial Chemistry, 1(1), 55-68.
• Veber, D. F., et al. (2002). J. Medicinal Chemistry, 45(12), 2615-2623.
• Egan, W. J., et al. (2000). J. Medicinal Chemistry, 43(21), 3867-3877.
• Muegge, I., et al. (2001). J. Medicinal Chemistry, 44(12), 1841-1846.
• Daina, A., & Zoete, V. (2016). ChemMedChem, 11(11), 1117-1121. (BOILED-Egg model)

License

MIT LICENSE