New An advanced version for virtual screening, SEARCH-ML, is now available. Try it here

RASPD +

Rapid Screening with Physicochemical Descriptors + Machine Learning

About RASPD+

RApid Screening with Physicochemical Descriptors + Machine Learning

RASPD⁺ is a computationally fast protocol for identifying lead-like molecules based on predicted binding free energy against a target protein with a 3D structure and a defined ligand binding pocket.

Originally developed at the Supercomputing Facility for Bioinformatics and Computational Biology (SCFBio), IIT Delhi by Goutam Mukherjee and B. Jayaram, development continued at the Heidelberg Institute for Theoretical Studies (HITS).

In version 1.0, we introduced scaffold search capabilities and integrated advanced machine learning algorithms. The model was rigorously trained on approximately 4000 non-metallo protein-ligand complexes retrieved from the PDBBIND refined data set.

Key Capabilities

  • Screen 1 Million molecules in minutes.
  • No rigorous docking required.
  • Pearson correlation: 0.74
  • RMSE: ±1.86 kcal mol⁻¹

Model Performance

Validation results on a test set of 493 unseen protein–ligand complexes.

RMSE Plot

Root Mean Square Error

[Image of Pearson Correlation Plot] Pearson Plot

Pearson Correlation

Q-Squared Plot

Predictive Ability (q²)

R-Squared Plot

Coeff. of Determination (R²)

Developed In Collaboration With