Welcome to ReMODE
ReMODE (Receptor-based MOlecular DEsign) server is a freely accessed webserver for de novo drug design. In the ReMODE server, the BiTL (bidirectional transfer learning) algorithm was specifically designed to extract and transfer the desired geometric features of the protein-ligand complexes to a latent space for generation. The pharmacophore conditioning and docking-based Bayesian sampling were applied to efficiently explore the vast chemical space for the design of molecules with desired geometric properties and/or pharmacophores. Based on these frameworks, users can select a target in the list to create the generative task and generate molecules rapidly, and the conditional models allows users to create the multi-property constraints task and fragment-based design task. Other key modules of ReMODE are the pharmacophore constraints and docking conformation optimization. Users using these two modules will get molecules with favorable pharmacophore matching coefficients and docking conformations, and these molecules will have higher potential as drug candidates.
ReMODE server has the following main functions:
Target-specified generation: Users can select a protein target in the list for de novo inhibitor design. All the protein targets in the list have the pretrained generative models. If the switch 'Optimization' is off, users can select a target from the list, then enter the task name, email address, and the number of molecules to be generated to quickly generate the target-focused molecules.
Multi-property constraint design: Users can generate a number of molecules with the specific values of MW and LogP within given ranges, and with the optimized values of QED and SA in the 'Physicochemical properties' module. It was also possible to adjust a single target property without changing the others.
Fragment-based generation: Users can perform fragment-based design by uploading a pre-defined scaffold/fragment/structure in the 'Structure features' module. The uploaded SMILES string is used as an anchor in the latent space to find the molecules with high structural similarity to the uploaded structure.
Docking-based generation: Users can generate molecules with favorable docking conformations by using Bayesian optimization. The 'Bayesian Optimization' module allows users to generate the target-focused molecules with high docking scores and favorable docking conformations using a specified Autodock Vina docking function.
Pharmacophore-based generation: Users can generate molecules according to the 3D pharmacophore models. The pharmacophore properties in CVAE were set to Relative Pharmacophore-Fit to a pharmacophore model.