How to Use LibInvent Online
Try LibInvent
Commercially Available Online Web Server
LibInvent: Reaction-Based Generative Scaffold Decoration for In Silico Library Design
LibInvent, a novel generative deep learning tool for designing focused chemical libraries. The software is an extension of the REINVENT framework and is capable of proposing libraries of compounds that share a common molecular core while optimizing for a range of desirable properties. A key innovation is the ability to incorporate specific chemical reactions, which ensures that the generated libraries are more likely to be synthetically accessible under the same or similar conditions.
How LibInvent Works
LibInvent uses a recurrent neural network (RNN) with an encoder-decoder architecture to model the generation of molecules based on SMILES strings. The workflow consists of two main stages:
Prior Model Training: A general model, or "prior," is trained once on a large database of molecules, such as ChEMBL, to learn the syntax of the SMILES language. This prior model understands how to decorate a given scaffold with chemical groups.
Reinforcement Learning (Finetuning): This is the core of the tool's application. The pre-trained prior model is efficiently fine-tuned using a reinforcement learning loop. The model iteratively proposes decorations for a user-provided scaffold and receives rewards based on a defined scoring function. This guides the model to produce compounds with specific properties and ensures they adhere to user-specified reaction filters.
The framework also includes a Diversity Filter to prevent mode collapse by penalizing the generation of identical compounds, which promotes a broader exploration of the chemical space.
What is Tamarind Bio?
Tamarind Bio is a pioneering no-code bioinformatics platform built to democratize access to powerful computational tools for life scientists and researchers. Recognizing that many cutting-edge machine learning models are often difficult to deploy and use, Tamarind provides an intuitive, web-based environment that completely abstracts away the complexities of high-performance computing, software dependencies, and command-line interfaces.
The platform is designed provide easy access to biologists, chemists, and other researchers who may not have a background in programming or cloud infrastructure but want to run experimental models with their data. Key features include a user-friendly graphical interface for setting up and launching experiments, a robust API for integration into existing research pipelines, and an automated system for managing and scaling computational resources. By handling the technical heavy lifting, Tamarind empowers researchers to concentrate on their scientific questions and accelerate the pace of discovery.
Accelerating Discovery with LibInvent on Tamarind Bio
Using LibInvent on a platform like Tamarind would accelerate lead optimization and drug discovery by providing a fast, controlled, and synthesis-aware design workflow.
Customizable Library Design: Researchers can define a molecular scaffold, a set of desired properties, and even specific chemical reactions to be used for synthesis. The platform uses LibInvent to generate a focused library of thousands of compounds that meet all these criteria, which are more suitable for high-throughput synthesis.
Enhanced Synthetic Feasibility: By training on data sliced according to chemical reactions, LibInvent ensures that the generated compounds are more likely to be synthetically accessible. This reduces the friction between computational design and wet lab execution, accelerating the overall design-make-test-analyze (DMTA) cycle.
Rapid Finetuning: The pre-trained prior model can be rapidly adapted to new tasks using reinforcement learning. This means that researchers can quickly pivot to new design projects without needing to retrain the entire model, significantly improving productivity.
How to Use LibInvent on Tamarind Bio
To leverage LibInvent's power on a platform like Tamarind, a researcher could follow this streamlined workflow:
Access the Platform: Begin by logging in to the tamarind.bio website.
Select LibInvent: From the list of available computational models, choose the LibInvent tool.
Select a Scaffold: Input the SMILES string of a molecular scaffold with a variable number of attachment points you wish to decorate.
Define a Scoring Profile: Use the platform's interface to set up a scoring function that includes molecular descriptors, QSAR models, and other criteria to guide the generation toward desired properties.
Apply Reaction Filters: (Optional) Specify the desired chemical reactions for each attachment point to ensure the resulting library is synthetically feasible.
Generate Library: The platform runs the reinforcement learning loop, which generates a diverse library of compounds that are optimized for your criteria. The high-scoring compounds are saved in a "scaffold memory" in real time.