ITsFlexible: Predicting the Conformational Flexibility of Antigen Receptors

ITsFlexible is a deep learning tool that solves a critical challenge in antibody and T cell receptor (TCR) engineering: reliably predicting the structural flexibility of Complementarity-Determining Region (CDR) loops. While models like AlphaFold predict a single, static structure, ITsFlexible provides a crucial binary classification of whether a CDR loop is 'rigid' (single stable state) or 'flexible' (multiple conformational states).

This predictive power, backed by rigorous experimental validation using cryogenic electron microscopy (cryo-EM), allows researchers to enhance the investigation of antibody function, affinity, and specificity, fundamentally impacting the design of therapeutics.

Scientific Foundation: The ALL-conformations Dataset

The core innovation behind ITsFlexible is the creation of a massive training dataset, ALL-conformations, which addresses the long-standing problem of data scarcity in conformation prediction.

• Massive Data Scale: ALL-conformations systematically mines the Protein Data Bank (PDB) to collect 1.2 million loop structures, representing more than 100,000 unique sequences.

• Targeted Motifs: The dataset focuses on CDR3 loops—the most critical and flexible region—and CDR3-like loop motifs (loops between two antiparallel β-strands) across all proteins.

• High-Confidence Labeling: Loops were classified based on structural similarity (using a pair-wise RMSD threshold of 1.25 A˚ to define a conformation). To ensure high confidence in the output, a loop is only labeled as rigid if it adopts a single conformation in at least five separate PDB structures.

Performance and Predictive Power

ITsFlexible is a graph neural network (GNN) architecture trained on the ALL-conformations dataset, demonstrating state-of-the-art performance compared to traditional biophysical models and existing uncertainty metrics from structure prediction tools.

• Superior Benchmarking: ITsFlexible significantly outperforms all alternative approaches on crystal structure datasets, including baselines based on loop length and solvent exposure, and zero-shot methods based on AlphaFold2 (AF2) predicted local distance difference test (pLDDT).

• Effective Generalization: The model successfully generalizes to complex biological systems by achieving high predictive accuracy when tested on conformational ensembles derived from molecular dynamics (MD) simulations of unbound antigen receptors.

• Experimentally Confirmed: The predictions were validated using cryo-EM experiments on three novel antibodies, with the model correctly identifying both flexible and rigid states in two out of three challenging cases.

Applications in Therapeutic Antibody Engineering

The ability to predict CDR flexibility is a powerful pre-screening tool in drug design and antibody engineering, allowing researchers to tune critical therapeutic properties:

• Affinity and Specificity: Flexibility impacts the entropic cost of antigen binding. Highly rigid loops are often associated with improved binding affinity, while flexible loops can lead to polyspecificity (binding multiple targets). ITsFlexible helps identify receptors with the desired flexibility profile.

• Rational Design: Enables the rapid screening of candidate molecules to identify those with favorable therapeutic profiles, helping to prioritize which designs move forward to expensive MD simulations or experimental validation.

• Input Flexibility: ITsFlexible achieves similar performance when using inputs derived from predicted structural models (like ImmuneBuilder or AF2) rather than crystal structures, making it applicable even to sequences without solved structures.

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. The Tamarind team hold information/data security as a top priority, as detailed in our Trust Center & Terms of Service, ensuring your data is safe on the platform.

How to Use ITsFlexible on Tamarind Bio

To leverage ITsFlexible's power for CDR flexibility prediction, a researcher could follow this streamlined workflow on the Tamarind Bio platform:

• Access the Platform: Begin by logging in to the tamarind.bio website.

• Select ITsFlexible: From the list of available computational models, choose the ITsFlexible tool.

• Input Target Information: Provide the structural model of the antibody or TCR Fv domain (e.g., a PDB or mmCIF file from an experimental structure or a structure prediction tool).

• Run Flexibility Classification: The platform executes the GNN architecture to perform a binary classification of the CDR3 loops based on the input structure and its contextual residues.

• Analyze Prediction Score: The output provides a score (ranging from 0 to 1), where scores closer to 0 indicate rigid loops and scores closer to 1 indicate flexible loops, allowing leads to be prioritized for further antibody engineering or simulation.

Source