Overview
FEgrow is a molecular-design workflow for constructing user-defined congeneric ligand series in protein binding pockets. The README describes the earlier FEgrow (1.*) workflow as preparing ligands for subsequent free energy calculations, rather than presenting FEgrow itself as a complete free-energy calculation engine. Its stated starting context is a user-defined ligand series and a protein binding pocket; its intended output is constructed ligand structures for downstream preparation and calculation. The source excerpts do not specify accepted structure-file formats or exported artifacts.
The README introduces FEgrow 2.0.0 as adding active learning and acceleration through Dask across multiple CPUs, nodes, or clusters. It directs users to notebooks covering basic interactive molecular design, chemspace functionality, and active learning for inhibitor design. These examples place FEgrow both in hands-on ligand elaboration and in compound-prioritisation workflows. The active-learning notebooks are linked to a 2024 ChemRxiv study concerning on-demand libraries targeting the SARS-CoV-2 main protease; a separate repository contains scripts used to create that study’s Figures 2–6.
The linked documentation is the stated destination for installation instructions and fuller workflow details. The excerpts provide no numerical acceleration results, predictive-accuracy measurements, or compatibility matrix, so suitability for a particular target, compute environment, or downstream free-energy package remains an evaluation question. The repository licence is MIT; that finding concerns repository code, not independently linked datasets or other resources.
Key Features
- Interactive construction of user-defined congeneric ligand series within protein binding pockets.
- A ligand-building workflow intended to supply structures for downstream free energy calculations.
- Active-learning functionality introduced in the README’s FEgrow 2.0.0 section, with an inhibitor-design notebook example.
- Dask-based acceleration described for multi-CPU, multi-node, and cluster execution.
- Tutorial notebooks covering basic interactive molecular design and chemspace functionality.
Use Cases
- Suggested evaluation: construct a congeneric ligand series for a selected protein pocket and assess whether the resulting structures fit an existing free-energy preparation workflow.
- Suggested evaluation: use the chemspace tutorial to explore how FEgrow’s documented functionality could support a molecular-design project.
- Suggested evaluation: study the inhibitor-design active-learning example as a starting point for compound prioritisation, without assuming its results transfer to another target.
- Suggested evaluation: assess Dask-based execution in a chosen compute environment before planning a larger molecular-design workload.
How to Use
- Read the repository README to distinguish the earlier ligand-building workflow from the active-learning and acceleration additions described under FEgrow 2.0.0.
- Follow the installation instructions at the official documentation. Consult those instructions for environment requirements rather than inferring commands or compatibility from the brief source excerpts.
- Begin with basic interactive molecular design in the tutorials folder. Use the notebook to determine required inputs and inspect its outputs before adapting it to a protein-pocket project.
- Continue with the chemspace and inhibitor-design notebooks in the same folder. Treat application to a new target or compound library as an evaluation, not an established result.
- For research context, consult the linked active-learning study and figure-script repository. Separately assess downstream free-energy requirements and Dask execution needs for your intended workflow.